doi
stringlengths 28
28
| title
stringlengths 19
311
| abstract
stringlengths 217
5.08k
| plain language summary
stringlengths 115
4.83k
| article
stringlengths 3.87k
161k
|
|---|---|---|---|---|
10.1371/journal.pcbi.1006550 | Comprehensive computational modelling of the development of mammalian cortical connectivity underlying an architectonic type principle | The architectonic type principle relates patterns of cortico-cortical connectivity to the relative architectonic differentiation of cortical regions. One mechanism through which the observed close relation between cortical architecture and connectivity may be established is the joint development of cortical areas and their connections in developmental time windows. Here, we describe a theoretical exploration of the possible mechanistic underpinnings of the architectonic type principle, by performing systematic computational simulations of cortical development. The main component of our in silico model was a developing two-dimensional cortical sheet, which was gradually populated by neurons that formed cortico-cortical connections. To assess different explanatory mechanisms, we varied the spatiotemporal trajectory of the simulated neurogenesis. By keeping the rules governing axon outgrowth and connection formation constant across all variants of simulated development, we were able to create model variants which differed exclusively by the specifics of when and where neurons were generated. Thus, all differences in the resulting connectivity were due to the variations in spatiotemporal growth trajectories. Our results demonstrated that a prescribed targeting of interareal connection sites was not necessary for obtaining a realistic replication of the experimentally observed relation between connection patterns and architectonic differentiation. Instead, we found that spatiotemporal interactions within the forming cortical sheet were sufficient if a small number of empirically well-grounded assumptions were met, namely planar, expansive growth of the cortical sheet around two points of origin as neurogenesis progressed, stronger architectonic differentiation of cortical areas for later neurogenetic time windows, and stochastic connection formation. Thus, our study highlights a potential mechanism of how relative architectonic differentiation and cortical connectivity become linked during development. We successfully predicted connectivity in two species, cat and macaque, from simulated cortico-cortical connection networks, which further underscored the general applicability of mechanisms through which the architectonic type principle can explain cortical connectivity in terms of the relative architectonic differentiation of cortical regions.
| The mechanisms that govern the establishment of cortico-cortical connections during the development of the mammalian brain are poorly understood. In computational simulation experiments reported here, we explored the foundations of an architectonic type principle, which attributes adult cortical connectivity to the differences in architectonic differentiation between cortical areas. Architectonic differentiation refers, among other characteristics, to the cellular composition of cortical areas. This architectonic type principle has been found to account for diverse properties of cortical connectivity across mammalian species. Our in silico model generated connectivity patterns that were consistent with the architectonic type principle, as they are typically observed in mammalian cortices, if model settings were chosen such that they corresponded to empirical observations of cortical development. Our computational experiments systematically evaluated previously proposed mechanisms of cortical development and showed that connectivity consistent with the architectonic type principle arose only from realistic assumptions about the growth of the cortical sheet.
| Axonal connections among brain areas are the structural substrate of information transfer throughout the brain. Cortico-cortical connections form networks that are neither regular nor random, but characteristically link specific brain regions, and exhibit large-scale topological features, such as modules and hubs [1, 2], rich-clubs [3, 4] and diverse-clubs [5], that have been the subject of wide-ranging investigations [6–19]. Moreover, there exist noteworthy regularities in the laminar patterns of cortical projection origins and terminations [20–23].
Many structural features of the cortex have been probed for their relationship to axonal connections between brain regions. For example, aspects of cell morphology have been shown to correlate with properties such as area degree (i.e., the number of projections maintained by an area) in the macaque monkey [24, 25] and humans [26].
One potent explanatory framework that imposes order onto the tangle of cortico-cortical connections is the so-called structural model [27] (reviewed in [28, 29]), also termed architectonic type principle (ATP). This principle describes the patterns of cortical projections and their laminar origins and terminations in terms of the relative architectonic differentiation of brain areas. Briefly, graded differences in cortical architecture have been found to account for the graded patterns observed in the distribution of projection origins and targets across cortical layers [27, 30–36]. Moreover, greater similarity in the architectonic differentiation of cortical areas has been found to be associated with higher connection frequency between them, above and beyond the explanatory power of spatial proximity [34, 36, 37] (see [29, 38] for reviews). Originally described for ipsilateral connections of the macaque prefrontal cortex [27], the ATP has since been confirmed for a considerable number of brain systems and species, as well as contralateral connections [30–37, 39–42], suggesting a mammalian-general organisational principle. The general applicability of this principle was further supported in a recent study which performed prediction analyses that transferred information across mammalian species [43]. Specifically, by training a classifier on the relationship between cortical structure and connections in a first species, area-to-area connectivity in a second species could be reliably predicted from structural variations of cortical areas in the second species without making changes to the classifier. Moreover, in the human brain, a similar association has been observed, whereby less differentiated agranular or dysgranular areas have the highest amount of functional connectivity [44].
The architectonic type principle, thus, allows the prediction of cortico-cortical connectivity from brain architecture regularities. Further substantiation of the ATP calls for a mechanistic explanation of how the described relationships between brain architecture and connectivity may emerge. From early on, the origin of this relationship has been hypothesized to be linked to developmental events [27]. Specifically, the observed close relationship between variations in cortical structure and axonal connections may arise from an interplay between the ontogenetic time course of neurogenesis and concurrent connection formation [29, 35, 45]. Areas which develop during different time windows were suggested to be afforded distinct opportunities to connect, with self-organisation rather than precisely targeted connection formation leading to the strikingly regular final connectivity patterns (cf. [46]). Put differently, it has been hypothesized that spatiotemporal interactions in the forming tissue, and specifically the relative timing of neurogenesis across the cortex, determine the connectivity patterns between cortical areas. Empirically, such a relationship has, for example, been observed in the olfactory system of the rat [47].
Here, using systematic computational simulation experiments, we explored whether this suggested mechanism may be capable of generating cortico-cortical connectivity consistent with empirical observations and the architectonic type principle (Fig 1). To this end, we implemented an in silico model of the growing two-dimensional cortical sheet of a single cerebral hemisphere that was progressively populated by neurons and divided into cortical areas. Model neurons randomly grew their axons across the cortical sheet and stochastically formed connections with potential postsynaptic targets (similar, for example, to simulation experiments in [48] and [49]). We assessed the resulting network of simulated structural connections between cortical areas in the same way as in previous experimental studies (e.g., [34, 36]) and compared the results to the empirical observations. Since we constrained the model to a single hemisphere, the simulated connections represent ipsilateral connectivity. Following this general approach, we characterized a number of variants of the in silico model of the growing cortical sheet, which differed in their adherence to empirical observations about developmental processes, specifically the spatiotemporal sequence of neurogenesis across the cortex. By comparing the networks generated from these variants, we could infer which aspects of the proposed mechanistic underpinnings of the ATP, particularly, which neurodevelopmental assumptions, were necessary to approximate empirical ipsilateral cortical connectivity.
We explicitly incorporated three aspects of corticogenesis in our simulations which are briefly described here.
First, the cortical sheet is established through neurogenesis spreading out from spatial origins, or primordial points (where the earliest neuronal populations are observed on the developing cortex), so that the surface of the cortex expands over time. This expansion is accompanied by a gradient in the time of onset of neurogenesis across the cortical sheet, which we refer to as the planar gradient of time of neurogenesis [50–60]. Developmental studies indicate that neurogenesis proceeds from at least two points of origin [57, 60, 61], with new neurons successively increasing the extent of cortical tissue between these neurogenetic origins. This progression entails that areas formed earlier become further separated on the cortical sheet as new areas are generated. Moreover, there is a superimposed radial gradient in the progression of neurogenesis [50, 51, 53, 62, 63] (which was not included in our in simulations), resulting in the characteristic inside-out generation sequence of neurons across layers (meaning that, with the exception of neurons in layer I, neurons in lower cortical layers are generated before neurons in upper cortical layers). In contradistinction to the findings outlining a planar gradient in the onset of neurogenesis, as described above, it has also been suggested that the onset of neurogenesis is simultaneous across the cortex [64, 65]. To contrast these two interpretations, we included both alternatives in our simulation experiments, as described in more detail below.
Second, cortical areas that are generated later are generally more architectonically differentiated [45, 60, 66, 67] (also briefly reviewed in [35]). Gradual changes in cortical architecture along two trends were described already several decades ago [68–72] (reviewed in [29, 38]). In brief, the two foci of least differentiated cortex are the allocortical three-layered archicortex (hippocampus) and paleocortex (olfactory cortex). These cortices are surrounded by periallocortex, where additional layers can be discerned, but without the clear laminar organisation found in the isocortex. Proisocortex, the next stage of differentiation, has a definite laminar organisation, but is missing a well-developed layer 4. Finally, there are different levels of isocortex with increasing demarcation of laminar boundaries and prominence of layer 4. More recently, changes in cell cycle kinetics across the forming cortical sheet and genetic correlates of the neurogenetic gradients have been described [58, 59, 73–75], which elucidate how gradual changes in cortical architecture are effected and provide an association between time of origin and architectonic differentiation. Particularly, a lengthening in the cell cycle along the planar neurogenetic gradient is accompanied by a successive increase in the proportion of progenitor cells differentiating into neurons with each cell cycle. In combination with the mentioned relation between time of origin and final laminar position of neurons, this mechanism results in a relatively increased number of supragranular layer neurons in later generated sections of the cortical sheet. Thus, a positive correlation can be observed between time of origin and neuron density across the cortex [67]. This link has been corroborated by findings in the human cortex, which directly traced systematic architectonic variation of the cortex to the timing of development [45]. A lengthening of the overall developmental time period, and with it the neurogenetic interval, appears to be responsible for increased neuron numbers both within the cortex of a given species, as well as across species which differ in their overall level of architectonic differentiation [66, 67, 76]. In fact, it has been suggested that cortical architecture correlates not only with neurogenetic time windows during ontogenesis, but also with the succession of architectural differentiation observed during brain evolution [60, 71]. This finding suggests that phylogenetic age has a bearing on architectural gradients. As mentioned above, it has repeatedly been reported that areas at similar points in the architectonic differentiation spectrum, as well as within the two described trends of architectonic progression, are preferentially linked, even if they are dispersed throughout the brain (also reviewed in [38]). The link to phylogeny, added to this correlation between architectonic progression and associated connectivity, thus, further points towards a developmental origin of the interrelations captured by the architectonic type principle.
The third aspect of neurogenesis which we incorporated into our simulations is that axon outgrowth starts concurrently with, or immediately after, neuronal migration [74, 77–80], and appears to be largely unspecific spatially [81]. We, therefore, assumed that connection formation starts as soon as neurons were placed in the cortical sheet. Further assumptions derived from these observations were that axons grow randomly across the cortical sheet (i.e., with no particular spatial orientation) and that they indiscriminately form connections once they are close enough to a potential target neuron, a mechanism that has been named Peters’s Rule [82, 83]. Thus, the process of connection formation can be described as stochastic, and has been simulated in this way in previous computational models of connection development, such as [49]. This mechanism entails that the probability of a neuron forming a connection is only dependent on the probability of its axon finding a target neuron. Since neurons that are far apart are separated by a larger number of neurons that could accommodate the axon, the probability of connecting to a target neuron is lower, the larger the distance between two neurons is. In effect, there is a positive correlation between the spatial proximity and connection probability of different neurons.
The spatiotemporal dynamics of corticogenesis that emerge from the combination of these empirically grounded assumptions were hypothesized to result in the establishment of realistic cortico-cortical connectivity. In particular, we expected interactions between the spatial and temporal aspects of neurogenesis to lead to the formation of connections which are consistent with the predictions of the architectonic type principle concerning the relationship between areas’ relative architectonic differentiation and connection frequencies. Our simulation experiments, thus, contribute the first systematic exploration of the neurodevelopmental mechanisms that have been hypothesized to underlie the ATP [27, 29, 35, 40].
In summary, we implemented several aspects of neurogenesis in an in silico model of the growing mammalian cerebral cortex. These aspects were then modified in some variants of the model, so that they either corresponded to, or violated, empirically observed phenomena. This strategy allowed us to compare the cortico-cortical connectivity resulting from hypothetical variants that differed in their assumptions, where some of these assumptions were empirically grounded and others were not. The approach enabled us to assess the merits of mechanisms which have been proposed to link cortical structure and connectivity through the ATP.
We simulated the growth of cortico-cortical connections between areas of different neuron density according to a constant set of growth rules. We evaluated how closely the simulated connectivity corresponded to empirical observations made in mammalian connectomes when the physical substrate of the connections, that is, the simulated cortical sheet, developed along different spatiotemporal trajectories. To this end, we systematically varied the settings of our in silico model to construct a number of variants, which we refer to as spatiotemporal growth layouts. We considered five sets of growth layouts: (1: realistically oriented density gradient) planar growth of the cortical sheet, such that cortical areas were added around neurogenetic origins, with new areas having an increasingly higher neuron density (i.e., neuron density increased with distance from a point of origin); (2: inverse density gradient) planar growth of the cortical sheet, such that cortical areas were added around neurogenetic origins, but with new areas having increasingly lower neuron density (i.e., neuron density decreased with distance from a point of origin); (3: radial) no planar growth of cortical areas on the fringes of the cortical sheet, but gradual addition of neurons at a constant rate across the cortical sheet, which resulted in an ordered gradient of area neuron density that was the same as in sets 1 and 4; (4: static) no planar growth of cortical areas, but the same final gradient of area neuron density as in sets 1 and 3; (5: random) planar growth of the cortical sheet, such that cortical areas were added around neurogenetic origins, but with no ordered gradient of area neuron density, instead neuron density varied randomly between locations on the cortical sheet. For all five sets, we implemented three growth modes: (1D 1row) one-dimensional growth implemented with one row of areas; (1D 2rows) one-dimensional growth implemented with two rows of areas; and (2D) two-dimensional growth. For all five sets, all three growth modes were implemented with planar growth around two neurogenetic origins. For set 1 (realistically oriented density gradient), we additionally implemented each growth mode with one neurogenetic origin as well as three (1D growth) or four (2D growth) neurogenetic origins. Thus, in total, we considered 21 growth layouts, grouped into five sets according to the spatiotemporal trajectory the cortical sheet traversed (see Fig 2 and Table 1 for an overview).
We first present some general statistics of the simulated connectivity and then go on to characterize how well the relationship between connectivity and the two factors of (relative) neuron density and spatial distance corresponded to previously published empirical observations for the different growth layouts. Finally, we assess how well the different growth layouts predicted empirical connectivity, as an indication of how realistic the simulated connectivity was for a given growth layout. Fig 3 provides an outline of this procedure. Table 2 gives an overview of all results.
The cortico-cortical networks resulting from the simulations showed realistic levels of overall connectivity, with between 39% and 66% of possible connections present (Fig 4A, Table 3). Previously, between 50% and 77% of connections were reported to be present in the macaque and cat cortex [22, 34, 84]. Some 2D growth layouts reached higher levels of connectivity, with up to 87% of possible connections present. This connection density translated into several hundreds of inter-areal connections (Fig 4B, Table 3), with between 250 and 400 connections for growth mode 1D 1row and between 900 and 1500 connections for growth mode 1D 2rows. Due to the large number of areas, connection numbers were much higher for 2D growth layouts, between 8000 and 18600.
We first checked how well the simulated networks corresponded to the empirical observations that a larger fraction of connections is present between regions that are more similar in neuronal density, as suggested by the architectonic type principle, and spatially closer to each other. To this end, we computed the relative frequency of present connections (Fig 5, Table 3). We then examined how well both factors, absolute density difference and distance, enabled the reconstruction of the simulated networks using logistic regression. Specifically, we assessed these relations by computing McFadden’s Pseudo R2 statistic, which provides a measure of the increase in the model log-likelihood with inclusion of either or both factors compared to a null model (Fig 6, Table 3).
Another property of the simulated networks that we compared to empirical observations was area degree (i.e., the number of connections per area). We previously reported that, in biological cortical networks, the number of connections maintained by an area is negatively correlated with the area’s cytoarchitectonic differentiation [34, 36]. Here, we show an analogous analysis for the simulated networks (Fig 7, Table 3, S3 Fig). For random, static and radial growth layouts, area degree was not significantly correlated with neuron density, with the exception of 2D growth layouts, which showed a positive and significant correlation in each case.
Growth layouts with realistically oriented density gradients showed a strongly negative, significant correlation between area degree and neuron density, with median correlation coefficients between -0.42 and -0.79 for both 1D growth modes. Conversely, for growth layouts with an inverse density gradient, area degree was strongly positively correlated with neuron density. For 2D growth along a realistically oriented density gradient, the observed effect was more variable. Correlation coefficients were of weak to moderate magnitude, and the correlation was not significant for 2D growth around one origin (2D 1origin: median ρ = 0.03, median p > 0.05; 2D 2origins: median ρ = 0.17, median p < 0.05; 2D 4origins: median ρ = 0.34, median p < 0.05). This observation was in contrast to the strongly positive and significant correlations observed for the 2D growth layouts without oriented growth, where median correlation coefficients were larger than 0.50 (random 2D: median ρ = 0.54; static 2D: median ρ = 0.62; radial 2D: median ρ = 0.59). We, therefore, concluded that the effect of oriented growth along a realistically oriented density gradient on area degree, as observed for both 1D growth modes, persisted in the 2D growth mode, but that it was not strong enough to completely abolish the tendency for a positive correlation between area degree and neuron density inherent to the 2D growth layouts, instead only diminishing the positive correlation.
In summary, the empirically observed negative correlation between area degree and neuron density was only reproduced for the growth layouts with a realistically oriented density gradient. We cannot rule out that there existed a minor contribution of geometric centrality to this relationship. However, taking into account the results for the radial and static growth layouts made clear that such an effect, if there was any in the realistically oriented gradient growth layouts, could only be secondary. Without expansive, planar growth, there is no temporal advantage helping earlier-formed areas to accrue more connections. Any negative correlation between neuron density and area degree would, thus, be caused by geometric centrality. Fig 7 illustrates that no such correlation arises for the radial and static growth layouts, where instead area degree appears to vary randomly with neuron density.
In the previous sections, we showed that empirically observed regularities, particularly a close relationship between connection existence and the two factors of relative neuron density and spatial distance, could be reproduced in silico. We further characterized how well the simulation captured this phenomenon by predicting empirical connectivity using classifiers trained on the simulated networks. Classification performance was used as a measure of how well the properties of the artificially generated networks reflected the characteristics of empirical brain networks, in particular, the macaque and cat cortical connectomes. We report two measures of classification performance, accuracy and the Youden index, J. Accuracy was calculated as the percentage of predictions that were correct, while the Youden index is a summary measure that takes into account both the rate of true positives and the rate of true negatives and indicates divergence from chance performance.
As seen from Figs 8 and 9, classification performance was generally higher for the macaque connectome than for the cat connectome. However, the described differences between growth layouts held for both species. We also provide the fraction of the available empirical connections that were classified in each species (Fig 10, Table 4). Generally, between 30% and 60% of the empirical connections were classified, with some growth layouts reaching up to 86% (Fig 10). However, for some growth layouts, nearly no empirical connections reached posterior probabilities of at least 0.75 (the minimal threshold applied for assigning a predicted label), and, thus, very low fractions of the available empirical connections were classified. Specifically, this applied to random growth layouts (median fraction classified between <0.01 and 0.14) and the inverse 2D growth layout (median fraction classified macaque: 0.08, cat: 0.05). The overall low posterior probabilities for these growth layouts and the resulting small fraction of classified empirical connections already suggested that the properties of those layouts did not correspond well to the properties of the empirical networks. This impression was corroborated by the classification performance measures (see below).
By performing comprehensive computational simulation experiments of how the network of interareal connections may develop during ontogenesis, we addressed the question of how cortico-cortical structural connections come to be closely related to the underlying substrate’s cytoarchitectural differentiation, an empirical observation made in multiple species [27, 30–37, 39–42]. The main component of our in silico model was a developing two-dimensional cortical sheet, gradually populated by neurons. To assess potential explanatory mechanisms, we varied the spatiotemporal trajectory of this simulated histogenesis. The rules governing axon outgrowth and connection formation, by contrast, were kept fixed across all variants of simulated histogenesis, so that the differences in outcome measures between spatiotemporal growth trajectories were introduced exclusively by the specifics of when and where neurons were generated.
To allow for straightforward interpretation of the simulation results, we applied network measures that were used in previous empirical studies, which allowed us to perform analyses on the simulated connectomes in the same way as we did on the empirical connectomes. Accordingly, the two characteristics of areas that were considered in the analyses of the final simulated network of interareal connections were their final position on the two-dimensional cortical sheet relative to other areas, measured as Euclidean distance, and their neuron density, which functioned as a surrogate for overall architectonic differentiation. Neuron density was expressed relative to the densities of other areas, that is, as density difference, for most analyses. We treated the existence of connections between areas as binary, thus, connections were considered as either absent or present.
We considered different spatiotemporal trajectories of how neurons populated the simulated cortical sheet. To recapitulate, simulated histogenesis proceeded according to five different sets of growth rules, with three to nine specific implementations per set (a total of 21 different growth layouts). These five sets were (1: realistically oriented density gradient) planar, expansive growth of the cortical sheet, with newer areas having successively higher neuron density; (2: inverse gradient) planar, expansive growth of the cortical sheet, with newer areas having successively lower neuron density; (3: radial) instead of planar growth, neurons started to populate all areas simultaneously and were added at a constant rate across the whole cortical sheet until each area reached its predetermined complement of neurons, with a final neuron density gradient identical to sets 1 and 4; (4: static) all neurons of the cortical sheet formed simultaneously, with a neuron density gradient identical to the final gradient of sets 1 and 3; (5: random) planar, expansive growth of the cortical sheet, with no ordered gradient of area neuron density around the two origins. To exclude effects specific to any particular implementation of these sets of growth rules, we considered three growth modes for each set: one-dimensional growth with one row of areas, one-dimensional growth with two rows of areas, and two-dimensional growth. For set 1, with a realistically oriented density gradient, we considered growth around one origin and three or four origins (for one-dimensional and two-dimensional growth modes, respectively) additionally to the growth around two origins that was used in all five sets.
These distinct spatiotemporal trajectories of cortical sheet growth led to considerable differences in the properties of the generated networks of structural connections. See Table 2 for an overall assessment of the results. While all growth layouts exhibited a clear decline in the relative frequency of present projections across larger distances, this measure correlated with absolute density difference only for a subset of growth layouts (Fig 5). Particularly, there was no consistent relationship for the random, static and radial growth layouts, while for oriented growth, both along a realistically oriented density gradient and along an inverse gradient, the relative frequency of present connections decreased with larger absolute density differences between areas.
A more precise assessment of the extent to which distance and density difference determined connection existence was obtained by predicting simulated connectivity using binary logistic regression. Here, a similar picture as for relative connection frequency emerged from comparing McFadden’s Pseudo R2 values across growth layouts (Fig 6). Distance was a better-than-chance predictor of connection existence for most growth layouts, as shown by the performance increase compared to a constant-only null model that is measured by McFadden’s Pseudo R2. In contrast, inclusion of absolute density difference increased prediction performance only for the layouts with oriented growth (both along realistically oriented and inverse density gradients), but not for the random, static or radial growth layouts.
Finally, the growth layouts differed in whether neuron density correlated with area degree (Fig 7). As before, for random, static and radial growth layouts, there was no consistent effect of neuron density on the measure of interest, in this case area degree. In contrast, there was a significant correlation with neuron density for layouts with oriented growth. This correlation was negative, as observed empirically, for growth layouts with a realistically oriented density gradient, but positive for growth layouts with an inverse density gradient.
In combination, these results demonstrate that the relation between neuron density, which is one crucial feature of the physical substrate in which connections are embedded, and cortico-cortical connections is strongly influenced by the precise spatiotemporal trajectory of cortex growth, which coincides with the time of connection formation. By manipulating where and when areas of varying neuron density were generated, we could observe a change in the extent to which connections of the simulated network were accounted for by the two factors of spatial proximity on the fully formed cortical sheet and the relative neuron density, indicating relative architectonic differentiation of areas.
As described above, the extent to which spatial proximity and relative neuron density determined simulated connectivity strongly depended on the specific spatiotemporal trajectory of the simulated growth of the cortical sheet. Growth layouts that more closely mirrored the biological developmental trajectory of the mammalian cortical sheet led to closer correspondence of the simulation results with empirical observations on adult connectivity. This finding became particularly apparent when we predicted empirical connectivity in two different mammalian species, cat and macaque, from regularities that were extracted from the simulated connectivity generated by the different growth layouts. Applying the regularities that emerged in our simulations to empirical data afforded a strong test of whether the simulations adequately captured ontogenetic processes and produced realistic networks. Our results showed that both of the aspects that were manipulated across growth layouts, the temporal trajectory of area growth as well as the direction of the neuron density gradient, were relevant for how well simulated connectivity allowed to predict empirical connectivity (Figs 8 and 9). First, it could be observed that growth layouts in which areas appeared successively around origins of neurogenesis (i.e., the realistically oriented density gradient growth layouts), were much better able to predict empirical connectivity than growth layouts with the same final neuron density gradient, but without the observed link between time of origin and neuron density (i.e., static and radial growth layouts). Second, in the presence of planar growth around origins, the direction of the neuron density gradient was crucial. This finding was indicated by the large decrease in prediction performance when comparing the realistically oriented gradient growth layouts with the random and inverse density gradient layouts. These sets of growth layouts both followed the same time course of cortical sheet expansion as the realistically oriented gradient, but with no relationship between time of origin and neuron density or a negative correlation between time of origin and neuron density, which contradicts the empirically observed positive correlation of time of origin with neuron density. Hence, the extent to which neuron density is well suited as a predictor of connectivity could be due to it reflecting neurodevelopmental time windows.
Third, our analyses revealed that the number of neurogenetic origins, around which new areas grew, influenced the correspondence to empirical connectivity (Tables 5 and 6). Growth around two origins arguably led to the best prediction performance: it was superior to growth around one origin for both accuracy and Youden index, and performed better than growth around three or four origins in terms of accuracy. For the Youden index, this performance difference was present, but too small to be meaningful or statistically significant. Thus, while correspondence between simulated and empirical connectivity clearly increased with the addition of a second origin of neurogenesis, there was at the very least no further performance increase with the addition of a third or fourth origin. Fourth, we observed that the overall level of prediction performance for the realistically oriented density gradient growth layouts was quite high, indicating that they afforded a good correspondence with empirical connectivity not only relative to the other growth layouts, but also in absolute terms. Therefore, a dual origin of neurogenesis and the resulting cytoarchitectonic gradients arguably are necessary components of the developmental mechanism for the empirically observed relations to hold (Fig 11). These findings stress the importance of the theory of the dual origin of the cerebral cortex [38, 71] and the presence of multiple gradients of neurogenesis [57, 85], for the configuration of connectivity in the adult cortex.
Collectively, the presented results suggest that planar cortical sheet growth around two origins of neurogenesis and a systematic increase in neuron density with later time of origin are crucial determinants of the development of realistic cortico-cortical structural connections. Conversely, assuming that connection formation is a stochastic process with few constraints, as simulated here, the assumptions underlying the spatiotemporal growth trajectories of the random, static, radial and inverse growth layouts were shown not to mirror actual principles of cortical organisation.
With the postulation of the architectonic type principle it was suggested that a close relationship between cortico-cortical connections and architectonic differentiation of the cortex might arise from the timing of neurogenesis [27], a process that occurs in close temporal proximity to the formation of connections. Specifically, it has been hypothesized that the relative time of generation of areas of different neuron densities affords them with different opportunities to connect with each other, thus imposing constraints on stochastically formed connections [29, 35]. This mechanism would be in line with findings in Caenorhabditis elegans [86] and rat cortex [47]. Moreover, a previous computational study demonstrated that topological features, such as modular connectivity, may arise from the growth of connectivity within developmental time windows [87]. Thus, the main premise of this study, that spatiotemporal interactions in the forming cortex determine connectivity, has long been under consideration. Here, we provide the first systematic exploration of the possible mechanistic underpinnings of the ATP. We simulated multiple combinations of spatiotemporal growth trajectories of the cortical sheet and neuron density gradients, to probe from which of the combinations realistic connectivity emerged. Our results showed that, indeed, of the wide variety of examined spatiotemporal growth trajectories, the variant of the in silico model that led to the best correspondence with empirical observations was the one that was based on the same assumptions as the mechanism proposed to underlie the realization of the ATP. Hence, the underlying assumption that differences in neuronal density correspond to distinct time windows was not refuted in the model, and neuron density carried predictive power with respect to connectivity features only if such a relation between density and neurogenetic timing held. Our systematic simulation experiments, thus, distinctly corroborate the previously hypothesized mechanistic underpinnings of the ATP and contribute a conceptualization that can be scrutinized for similarities with, and distinctions from, actual ontogenetic processes. This approach opens up the possibility of characterizing in more detail how correlations between the structure of the cortex and cortical connections emerge, because all aspects of the process are observable. Further refinement of the simulation, for example by introducing species-specific histogenetic time courses, will enable the exploration of species differences or potentially the demonstration of invariance to changes in some aspects of ontogenesis. Another factor that could be probed is how robust the emergence of realistic connectivity is against changes in absolute neuron density, which varies considerably across species [66, 88]. From our simulations, it appears that temporal proximity of areas during neurogenesis underlies the positive relationship between similar neuron density and high connection probability. The close correlation between time of origin and architectonic differentiation described empirically (see Introduction) leads to a derivative correlation between temporal proximity of neurogenetic time windows and relative differentiation of cortical areas. Independent of this correlation, on a cortical sheet that expands around the origins of neurogenesis, areas with closer neurogenetic time windows tend to be spatially closer as well. Assuming that connection formation is a stochastic process, which implies that connection probability declines with spatial distance, this process leads to a higher connection probability between areas that are generated during nearby time windows. Temporal proximity during neurogenesis would, thus, be the common antecedent determining both relative architectonic differentiation and connection probability, while those two factors would only be indirectly related. Temporal proximity, however, is difficult to measure, and it is, therefore, no surprise that the correlation between its two direct consequences has been empirically observed first. This chain of reasoning reveals how our modulation of the relationship between temporal proximity during neurogenesis and relative architectonic differentiation in the considered growth layouts could have caused the vastly different outcomes in connectivity described here.
In our simulations, we observed a relationship between the spatial proximity of areas and their likelihood to be connected, which appears to be an epiphenomenon of stochastic connection growth within a physically embedded system (c.f. [49, 89]). Distance is an inherent property of a spatially embedded system that cannot be removed from the implementation of spatial growth. However, in our simulation of cortical growth, the final distance between areas was not always an accurate measure of their distance during the time period of connection formation, which would be the factor that mattered principally for determining the likelihood by which two areas became connected. Since this distance during cortical sheet growth is correlated with the areas’ final distance, there was also a correlation between final spatial proximity and connection probability. But this correlation does not genuinely describe the dependency of the stochastic growth process on distance, because inter-areal distance was not static, as implied by this measure. The distance measure relevant here, namely distance at the time of connection formation, would be challenging to measure empirically. Therefore, relying on measures of final, adult distance and assuming a strong correlation between the two distance measures appears as a pragmatic strategy for empirical analyses.
The present simulation experiments were designed to allow for the analysis of connection existence, that means, whether a possible connection between a pair of areas is present or absent in the final network. Naturally, axonal connections have many further properties beyond their simple existence; one prominent feature being the laminar distribution of the projection neurons’ somata and axon terminals in the areas of origin and termination, respectively. Laminar patterns of projection origin and termination are very well explained by the architectonic type principle (reviewed in [28, 29]), as has been demonstrated extensively in different species and cortical systems [27, 30–33, 35, 36, 39, 41, 42]. These conspicuous regularities most likely arise from fundamental developmental mechanisms, since they are ubiquitous and quite robust. This aspect becomes strikingly apparent in reeler mutant mice, where laminar connectivity patterns are largely correct [90–92] (shortly reviewed in [74]), despite a systematic inversion (to ‘inside-out’) of neurons’ final laminar positions relative to the regular order that neurons typically assume according to their time of origin (‘outside-in’) [53, 90, 93–96]. However, the precise mechanisms through which laminar projection patterns become established are still under investigation. Further simulation experiments could, therefore, be helpful in evaluating potential candidate mechanisms. Expanding the simulation of cortical sheet growth to take into account the radial distribution of neurons across layers, as would be required for assessing laminar projection patterns, will afford the introduction of spatially and temporally more fine-grained features of neurogenesis and final architectonic differentiation. In addition to the planar gradient in neurogenetic time windows, which was taken into account in the present simulation experiments, this could mean to include the radial gradient in neurogenetic time windows that characterises neurons populating different layers [50, 51, 53, 62, 63]. Beyond the density of neurons in any given area, as considered here, structural variation could include a number of cellular morphological measures which have been shown to change systematically with overall density (e.g., [25]). In the human brain, cellular morphological measures have also been shown to associate with how densely connected a cortical area is, both in the healthy brain and in the context of brain disorders [97, 98].
Another feature that could prove to be relevant for the establishment of laminar projection patterns is the relative neuron density of cortical layers. As overall neuron density increases across the cortex, layer 2/3 becomes successively more pronounced [38, 99] and neuron density increases more in the supragranular layers than in the infragranular layers [67]. Thus, there is a shift in when, and in which layers, the majority of neurons is generated for areas of different overall density, which could affect laminar patterns, especially in interaction with the sequential growth of areas across the cortical sheet. Further, there is systematic variation in the size of pyramidal cell somata across the cortex, a phenomenon termed externopyramidization [71, 100]. Specifically, the ratio of soma size in supragranular to infragranular layers is larger in areas of strong architectonic differentiation than in weakly differentiated areas (i.e., in weakly differentiated areas, infragranular pyramidal cells tend to be larger than supragranular cells, while the reverse is true for strongly differentiated areas). Hence, it seems that the laminar position of projection origins is aligned with relatively larger cell size in the candidate population for cortico-cortical connections, pyramidal cells. Since maintenance of long-distance connections between cortical areas is metabolically expensive [101, 102], relatively larger cell size conceivably is advantageous for their maintenance. Thus, as hypothesized before [43,103], externopyramidization might be linked to a shift in projection origins.
In addition to the above-mentioned properties of the cortical sheet itself, there are potential modifications of the stochastic formation of connections to be considered. First, the pruning of connections during later stages of development [104] was not taken into account in the present simulations. Laminar projection patterns may conceivably be affected by selective elimination of some axon branches but not others [105, 106]. Moreover, it has been observed that the time course of connection formation is not the same for all types of cells. Callosal projection neurons can reach their target areas without actually invading the gray matter, instead remaining in the white matter for a waiting period of days [107–109]. Similarly, waiting periods below the gray matter have been described for infragranular neurons projecting to area V4 from multiple areas in the ipisilateral hemisphere in macaques [110]. In contrast, supragranular neurons in the same tract-tracing experiments were found to invade the gray matter early, but many of them formed only transient projections that were subsequently eliminated. More generally, these and similar tract-tracing experiments have been interpreted to demonstrate different developmental profiles for axon outgrowth and connection formation in infra- and supragranular neurons [110–113]. In ‘feedback’ pathways, which according to the ATP can be conceptualised as projecting towards a relatively more differentiated area, extensive remodelling of laminar projection patterns until long after birth has been observed in a number of species (mouse, cat, macaque, human) and target areas [110–120]. This remodelling has been linked to activity-dependent maturation of pathways and the emergence of more refined perceptual capabilities [111, 120, 121] (for example, reviewed in [122, 123]). This observation suggests that not all factors contributing to adult laminar projection patterns may be accessible in simulation experiments with time frames that are restricted to corticogenesis and initial axon outgrowth.
A further potential determining factor in the establishment of laminar projection patterns that warrants exploration is the possibility of genetic specification. The laminar position of projection targets might be regulated by genetically encoded factors. Numerous layer-specific transcription factors and neurotrophins have been described, which afford a precise targeting of specific layers or even cell types and cellular compartments (reviewed, e.g., in [124, 125]). Co-culture experiments using cortical explants have shown that appropriate laminar position of axon terminals was retained outside of the ontogenetic growth environment, that is, in the absence of regular temporal and spatial relationships. Accurate laminar specificity has been demonstrated, for example, for thalamo-cortical, geniculo-cortical, and cortico-spinal connections in co-culture (e.g., reviewed in [124]). Similarly, connections formed in co-culture of rat visual cortex explants were shown to conform to organotypic laminar distributions [126, 127]. Castellani and Bolz [128] elegantly demonstrated that organotypic and cell type specific projection patterns could be induced by membrane-associated factors through both induction and prevention of axon ingrowth and branching. Moreover, it has been shown that transcription factors can have population-specific effects, enlarging the range of potential interactions. For example, Castellani and colleagues [129] found that the membrane-bound protein Ephrin-A5 functioned as a repulsive axonal guidance signal in neurons destined to migrate to layer 2/3, while acting as a ‘branch-promoting’ signal in neurons destined for layer 6. These observations suggest that laminar patterns of projection terminations may not be entirely explicable by spatiotemporal interactions in the forming tissue, but are regulated by more prescriptive determinants.
Our results illustrate how a mechanism linking the temporal order of neurogenesis across the cortex with the architectonic differentiation of areas could come to shape cortico-cortical connectivity such that it resembles the empirically observed connectivity of mammalian connectomes. However, simulation experiments, as performed here, can only assess whether a suggested mechanism is principally feasible, and explore what its essential components might be. That is, such computational experiments put a candidate mechanism to the test and allow drawing some inferences about possible (and, importantly, impossible) ingredients, but they do not establish biological facts by themselves. Ultimately, only empirical observation of the ontogenesis of the cortex can establish how this developmental process unfolds. The possibility cannot be excluded that there may exist an unrelated mechanism working through features not considered here, which could cause the phenotype of interest, in our case the close relation between architectonic differentiation and connectivity. Generally, incorporating more empirical anchor points in a model gives the conclusions of a simulation study more significance. To triangulate a likely solution to the developmental puzzle of how axonal connections are organized, it is necessary to constrain potential mechanisms by as many observable features as possible. As discussed above, more processes that shape connectivity could be included in our in silico model of neural development, such as waiting periods for connection formation, a differential ability of cortical layers to retain connections (possibly linked to externopyramidization), the pruning of established connections, or the action of signalling molecules in attracting and repelling axons during connection formation. By integrating such processes, new insights could be gained into the emergence of further connection features such as laminar projection patterns and projection strengths.
We constrained our in silico model to represent a single cerebral hemisphere, hence our results only apply to ipsilateral, intra-hemispheric connections. Contralateral, inter-hemispheric axonal connections have also been reported to be well represented by the architectonic type principle [31, 37], although at generally lower connection strengths. The in silico model could be expanded by a second hemisphere which develops simultaneously. Since similar types of cortex in the two hemispheres would be formed at nearby points in time, but further apart in space, this setup would be expected to lead to the observed pattern of ATP-consistent, but weaker connectivity, if the principle holds that spatiotemporal interactions govern connectivity patterns.
We modelled the developing cortex as a two-dimensional sheet, across which axons grew until they met a target soma and formed a connection. In reality, the mammalian cortical sheet is not flat, but becomes at least curved, and often intricately folded, during corticogenesis. Moreover, axons are not positioned exclusively within the grey matter, but instead cover large distances through the white matter. These shortcuts between distant points on the cortical sheet imply that representing projection length as Euclidean distance between points on a flat cortical sheet is not accurate. Yet, regardless of how the concurrent processes of neurogenesis, axon formation and cortical folding affect each other [130, 131], measuring the precise lengths of projections in the adult cortex has so far not been straightforward. Hence, approximate measures have been employed, such as border distance on a cortical parcellation, Euclidean distance in three-dimensional space, or geodesic distance which accounts for some of projections’ confinement to white matter tracts. Euclidean distance on the simulated two-dimensional cortical sheet may, therefore, be a suitable surrogate measure for these approximate empirical measures. In line with this assumption, if cortical folding had a strong impact on our prediction of empirical data, it would be expected that performance in the less folded cat cortex would be better than in the more strongly folded macaque cortex. As this was not the case, we suspect that cortical folding and the resulting changes in projection lengths do not dramatically alter the spatiotemporal interactions which we hypothesize link architectonic differentiation and cortical connectivity. To further test this expectation, it would be interesting to predict connectivity data from a wider range of species, such as lissencephalic rodents and humans, whose cortex is even more strongly folded than the macaque cortex.
Lastly, applying the classifier which was trained on simulated network data to predict empirical connectivity data resulted in better prediction performance for the macaque cortex than the cat cortex. Ultimately, there might be two reasons for this finding: Either the architectonic type principle characterises connectivity better in one of these species than the other, or the empirical measures that were used more faithfully capture the true structure in one of the species. Conceivably, adherence to the ATP might not be as pronounced in the smaller cat cortex, where both distances are shorter and therefore less distinctive, and there is less variation in total neuron number within the cortex due to a shortened neurogenetic interval [67]. Regarding the second possible reason, the structural measures from which we predicted connectivity were more detailed in the macaque cortex (neuron density and Euclidean distance) than in the cat cortex (structural type and border distance). Further experiments are therefore required to distinguish between these two explanations. Indeed, it would be intriguing to expand the prediction of empirical connectivity data from simulated networks to other species, preferably to mammals whose cortex is on either side of cat and macaque on the scales of size and degree of architectonic differentiation. Just as for assessing the impact of cortical folding, rodents and humans would be good candidates to identify the source of the observed difference in prediction performance.
We performed simulations of cortical sheet growth and the concurrent formation of cortico-cortical connections, systematically varying the spatiotemporal trajectory of neurogenesis as well as the relation between architectonic differentiation and time of origin of neural populations. Our results showed that, for realistic assumptions about neurogenesis, successive tissue growth and stochastic connection formation interacted to produce realistic cortico-cortical connectivity. This finding illustrated the fact that precise targeting of interareal connection terminations was not necessary for obtaining a realistic replication of connection existence within a cortical hemisphere. Instead, spatiotemporal interactions within the structural substrate were sufficient if a small number of empirically well-grounded assumptions were met, namely (i) planar, expansive growth of the cortical sheet as neurogenesis progressed, (ii) stronger architectonic differentiation for later neurogenetic time windows, and (iii) stochastic connection formation. We, thus, demonstrated a possible mechanism of how relative architectonic differentiation and connectivity become linked during development. These findings support hypotheses advanced in previous reports about the mechanistic underpinnings of the architectonic type principle [27, 29, 35, 40]. The successful prediction of connectivity in two species, cat and macaque, from our simulated cortico-cortical connection networks further underscores the generality of the ATP and the wide applicability of its explanation of connectivity in terms of relative architectonic differentiation.
We first describe the variants of the in silico model we considered and how we simulated the formation of cortico-cortical connections on a forming cortical sheet, representing a single hemisphere. We then detail how we analysed the resulting simulated networks.
Connection formation was simulated to take place on a two-dimensional, rectangular cortical sheet, where neuron somata and axon terminals were assigned two-dimensional coordinates without spatial extent. Somata were arranged in rectangular cortical areas which differed in their surface density of neurons. Neuron density has been shown to be a good indicator of a cortical area’s overall degree of architectonic differentiation [40] and has been used previously to relate differentiation to connectivity in the macaque brain (e.g. [35, 36]). Hence, we used neuron density as a central marker for architectonic differentiation, with larger neuron density corresponding to a stronger degree of differentiation. We did not adjust the absolute magnitude of neuron density to empirical values, but did choose the range of neuron densities such that it was similar to empirically observed variation in neuron densities across the cortex, with about a five-fold increase between areas of lowest and highest neuron density (cf. [35]). We implemented neuron density as number of somata per unit area of cortical sheet (#/arbitrary unit2). All cortical areas were defined to be of the same size. From these two constraints on neuron density and area size, it followed that areas of different densities contained different numbers of neurons. Within an area, somata were spaced equidistantly.
The generation of the cortical sheet across time was simulated in a number of different settings of the in silico model, here called variants or growth layouts. These growth layouts systematically differed in where and when neurons were generated on the forming cortical sheet, that is, they had different spatiotemporal growth trajectories. Below, we describe all growth layouts and their correspondence to neurodevelopmental findings in detail. An overview is provided in Table 1, and Fig 2 as well as S1 Fig give a visualisation of cortical sheet development over time for the different growth layouts.
All considered spatiotemporal growth trajectories were grouped into five sets of growth layouts. These sets differed with respect to whether cortical areas were generated by planar, expansive growth, whether there was radial growth, and in the final gradient of neuron density around neurogenetic origins.
In growth layouts with planar growth, the cortical sheet expanded, as, with each growth event, new cortical areas emerged around neurogenetic origins. Each new cortical area was grown within one time step, thus all constituent neurons appeared on the cortical sheet simultaneously. Neurogenesis occurred on the outer fringes of the portion of the cortical sheet already generated around each origin of neurogenesis. For more than one neurogenetic origin, this process entailed that newly generated areas moved previously generated areas apart on the cortical sheet, increasing the spatial distance in between them. Thus, planar growth mimicked the empirically observed planar gradient in onset of neurogenesis (see Introduction).
Radial growth, in contrast, did not expand the cortical sheet over time. Here, the cortical sheet had its final dimension already at the start of corticogenesis and cortical areas did not differ with respect to the time of onset of neurogenesis, but instead in the length of their neurogenetic interval. During each growth event, neurons were added at a constant rate across the entire cortical sheet. Areas with lower neuron density finished generating their complement of neurons earlier in time than areas with a higher neuron density, which needed to generate a larger number of neurons. Radial growth thus reproduced an alternative interpretation of studies of neurogenetic timing (see Introduction).
Growth events, during which the cortical sheet was generated, were distributed across the fixed simulated length of time. For both planar and radial growth, they were timed in such a manner that all neurons had grown after one third of the simulation length, and the remaining time steps could be used for connection formation by all neurons.
These three main properties of spatiotemporal growth of the cortical sheet were combined in the five sets of growth layouts, with each set containing three (or in one case nine) growth layouts, as follows: The first set, the realistically oriented density gradient growth layouts, grew by planar growth. Here, newly generated areas were of higher neuron density than previously grown areas. That is, there was a positive correlation between time of origin and neuron density, which appeared as a distinct gradient in neuron density around the neurogenetic origins on the final cortical sheet. The second set, the inverse neuron density gradient growth layouts, grew by planar growth like sets 1 and 5. However, in these inverse gradient growth layouts, newly generated areas were of lower neuron density than previously grown areas, that is, there was a negative correlation between time of origin and neuron density. The third set, the radial growth layouts, grew by radial growth. The final density gradient was identical to sets 1 and 4, but for the radial growth layouts, this pattern was caused by a positive correlation between length of the neurogenetic interval and neuron density, instead of a correlation between the time of onset of neurogenesis and neuron density. The fourth set, static growth layouts, did not in fact grow at all. All neurons were grown during the first growth event, thus the cortical sheet was fully formed from the beginning of the simulation. The final density gradient was identical to sets 1 and 3. Finally, in the fifth set, the random growth layouts, the cortical sheet grew by planar growth. The resulting final cortical sheet had no directed gradient of neuron density around the neurogenetic origins. Instead, each newly generated area was randomly assigned a neuron density. Possible density values were drawn from the neuron densities found on the final cortical sheet of the first set, realistically oriented neuron density gradient.
For each of these five sets, we implemented three different growth modes to mitigate influences of any specific choice of spatial implementation. Each growth mode was implemented around two neurogenetic origins. The three growth modes were as follows: First, one-dimensional growth with one row of areas (1D 1row growth layouts), where new areas grew to the left and right of neurogenetic origins (i.e., along the x-dimension of the cortical sheet) and there was only one row of cortical areas. Second, we implemented one-dimensional growth with two rows of areas (1D 2rows growth layouts), where, again, areas were added to the left and right of neurogenetic origins, but there were two rows of areas stacked in the y-dimension of the cortical sheet. Third, we implemented two-dimensional growth (2D growth layouts), where new areas were added on all sides of neurogenetic origins (i.e., in both the x- and y-direction of the cortical sheet). In this growth mode, each successive growth event led to an exponentially increasing number of added areas, and for set 1, realistically oriented density gradient, an unproportionally high number of areas of the highest neuron density, which did not accurately reflect the composition of the mammalian cerebral cortex. However, as stated above, we simulated the different growth modes to alleviate side-effects that might unintentionally arise from any particular spatial layout. Considering results across these specific implementations vastly reduced the risk of misinterpretation. We therefore included the two-dimensional growth mode despite its unrealistic rendering of the cortical sheet as a further control.
As mentioned before, each of the 15 growth layouts that were described so far was implemented around two origins of neurogenesis (5 sets x 3 growth modes x 1 number of origins). For set 1, realistically oriented neuron density gradient, we additionally considered two different numbers of origins for each growth mode. Specifically, we included growth around one neurogenetic origin and growth around three or four neurogenetic origins for 1D and 2D growth modes, respectively. These further six growth layouts allowed us to test whether the exact number of neurogenetic origins meaningfully influenced final connectivity.
Thus, we considered a total of 21 growth layouts (5 sets x 3 growth modes x 1 number of origins + 1 set x 3 growth modes x 2 numbers of origins). We simulated 100 instances of the spatiotemporal development of each of these 21 growth layouts.
Axons randomly grew across the cortical sheet and stochastically formed synaptic connections (similar to, e.g., [49]; also see [46]). Each neuron was assigned one axon terminal, which was initially located at the respective soma position. With each time step of the simulation, the axon extended by a fixed length at a random angle, and the position of the axon terminal changed accordingly. Once axon terminals left the cortical area their parent soma was located in, they were free to form a synapse with any neuron soma they encountered. Since both terminals and somata were defined by point-coordinates, a synapse was formed once the axon terminal approached a soma closer than a defined maximal distance. Upon synaptic contact, an axon stopped growing and the now occupied axon terminal remained at the location of the contacted soma for the remainder of time steps. To further increase stochasticity, we imposed a connection probability of 90% on potential synaptic contacts. Thus, in 90% of cases, a synapse successfully formed once the terminal was close enough to a soma, but in a randomly chosen 10% of cases, no synapse formed at this time step and the axon continued to grow. If soma positions changed because the cortical sheet grew, axon terminals (both occupied and unoccupied) were shifted with the cortical area they found themselves in at the time, and synaptic contacts were retained. This procedure of axon growth and synapse formation was not modified across variants of the in silico model.
Different parameters of the axon growth process interacted to determine how fast axon terminals made synaptic contacts. This included for example the increase in axon length per time step and the maximal distance for synapse formation. In pilot runs of the simulation, we calibrated the relevant parameters such that after the fixed simulated length of time, most axon terminals (>99.9%) had made synaptic contact and final interareal connectivity fell into a range comparable to empirical reports [22, 34, 84]. This calibration resulted in slightly different parameter values for 1D and 2D growth modes, but the same values were used in all simulation instances within these growth modes.
From the final state of the simulated cortical sheet, we extracted a number of features that were analogous to measures used in previous analyses of the mammalian cortex.
First, we collapsed the axonal connections between individual neurons into a simulated connectome, which contained information about the existence of all possible area-wise connections. Thus, we constructed a complete binary connectivity matrix where connections were coded as either absent or present.
Second, we extracted the two relevant structural measures from the final cortical sheet. The first measure was each area’s neuron density, and derived from that the difference in neuron density between area pairs, where density difference = densityarea of origin − densityarea of termination. For most analyses, we considered the undirected equivalent, the absolute value of density difference, which indicates the magnitude of the difference in neuron density between two areas. These two measures were equivalent to measures of architectonic differentiation previously used in studies examining mammalian cortical connectivity, such as neuron density difference (e.g., [35]), the log-ratio of neuron densities [36], or difference in cortical type, which is an ordinal measure of architectonic differentiation (e.g., [33–35]). The second measure was the spatial proximity between pairs of areas, which we calculated as the Euclidean distance between areas’ centres of mass. This measure was equivalent to measures of spatial proximity we used in previous empirical studies (e.g., [35, 36]). Since distance is an undirected measure, each analysis that included distance required the use of the undirected measure of neuron density difference, its absolute value.
We performed the described analyses for each of the 100 instances that were simulated for each growth layout and aggregated results across instances. For the simulations and analyses we used Matlab R2016a (The MathWorks, Inc., Natick, MA, USA).
|
10.1371/journal.ppat.1003292 | Illumination of Murine Gammaherpesvirus-68 Cycle Reveals a Sexual Transmission Route from Females to Males in Laboratory Mice | Transmission is a matter of life or death for pathogen lineages and can therefore be considered as the main motor of their evolution. Gammaherpesviruses are archetypal pathogenic persistent viruses which have evolved to be transmitted in presence of specific immune response. Identifying their mode of transmission and their mechanisms of immune evasion is therefore essential to develop prophylactic and therapeutic strategies against these infections. As the known human gammaherpesviruses, Epstein-Barr virus and Kaposi's Sarcoma-associated Herpesvirus are host-specific and lack a convenient in vivo infection model; related animal gammaherpesviruses, such as murine gammaherpesvirus-68 (MHV-68), are commonly used as general models of gammaherpesvirus infections in vivo. To date, it has however never been possible to monitor viral excretion or virus transmission of MHV-68 in laboratory mice population. In this study, we have used MHV-68 associated with global luciferase imaging to investigate potential excretion sites of this virus in laboratory mice. This allowed us to identify a genital excretion site of MHV-68 following intranasal infection and latency establishment in female mice. This excretion occurred at the external border of the vagina and was dependent on the presence of estrogens. However, MHV-68 vaginal excretion was not associated with vertical transmission to the litter or with horizontal transmission to female mice. In contrast, we observed efficient virus transmission to naïve males after sexual contact. In vivo imaging allowed us to show that MHV-68 firstly replicated in penis epithelium and corpus cavernosum before spreading to draining lymph nodes and spleen. All together, those results revealed the first experimental transmission model for MHV-68 in laboratory mice. In the future, this model could help us to better understand the biology of gammaherpesviruses and could also allow the development of strategies that could prevent the spread of these viruses in natural populations.
| Epstein-Barr virus and the Kaposi's Sarcoma-associated Herpesvirus are two human gammaherpesviruses which are linked to the development of several cancers. Efficient control of these infections is therefore of major interest, particularly in some epidemiological circumstances. These viruses are however host-specific and cannot be experimentally studied in vivo. The identification of a closely related viral species, called Murid herpesvirus 4 with the main strain called murine gammaherpesvirus-68 (MHV-68), in wild rodents opened new horizons to the study of gammaherpesvirus biology. Surprisingly, despite 30 years of research, MHV-68 transmission had never been observed in captivity. In this study, using in vivo imaging, we showed that MHV-68 is genitally excreted after latency establishment in intranasally infected female mice. This allowed us to observe, for the first time, sexual transmission of MHV-68 between laboratory mice. In the future, this model should be important to better understand the biology of gammaherpesviruses and should also allow the development of strategies that could prevent the spread of these viruses in natural populations.
| Herpesviruses are important pathogens which are ubiquitous in both human and animal populations. They establish persistent, productive infections, with virus carriers both making anti-viral immune responses that protect against disease and excreting infectious virions. Among herpesviruses, gammaherpesviruses establish a long-term latent infection of circulating lymphocytes. They drive lymphocyte proliferation as part of normal host colonization and consequently they can induce some lymphoproliferative disorders. In humans, Epstein-Barr virus (EBV) and the Kaposi's Sarcoma-associated Herpesvirus (KSHV) are associated with several human malignancies such as Burkitt's and Hodgkin's lymphomas, nasopharyngeal carcinoma, Kaposi's sarcoma and post-transplant lymphoproliferative disease [1], [2]. Human cancers associated with these two viruses are particularly prevalent in Africa where they are linked to malaria [3] and human immunodeficiency virus-1 (HIV-1) infection [4]. More generally, individuals with inherited or acquired immunodeficiency have an increased risk of developing a malignancy caused by one of these two viruses [5]. Efficient control of these infections is therefore of major interest, particularly in some epidemiological circumstances.
Knowledge and understanding of the mechanisms of virus transmission in populations are essential to implement large scale antiviral strategies. EBV is mainly shed from the oropharynx into saliva for horizontal spread of the infection to new hosts through mouth-to-mouth contact [6]–[8]. Similarly, horizontal transmission by saliva appears the most common route of KSHV spread in a population [9]. However, several studies in the past decades pointed to human gammaherpesviruses shedding through other routes such as the uterine cervix [10]–[13] or male genital tract [14], [15]. Thus, EBV and KSHV transmission could be more complex than previously thought.
Experimental studies are difficult to perform directly with human gammaherpesviruses because they show limited lytic growth in vitro and have no well-established in vivo infection model. However, the identification of a closely related virus, murine gammaherpesvirus-68 (MHV-68), in wild rodents offered the possibility of developing a mouse model of gammaherpesvirus pathogenesis [16]. MHV-68 readily infects laboratory mouse (Mus musculus) which is a valuable model for in vivo studies [17]. Experimental MHV-68 infection typically employs intranasal virus inoculation under general anaesthesia. This leads to a lytic infection of nose and of lung alveolar epithelial cells that is controlled within 2 weeks [18]. Virus meanwhile seeds to lymphoid tissue, mainly draining lymph nodes and spleen [19], and drives the proliferation of latently infected B cells. This peaks at 2 weeks post-infection (p.i.) and is controlled by 4 weeks. A predominantly latent infection of memory B cells then persists for life [20]–[22]. Macrophages and dendritic cells (DCs) also harbour latent MHV-68 infection [20].
Although MHV-68 has been studied for more than 30 years [16], attempts to demonstrate horizontal transmission in laboratory mice have been almost entirely unsuccessful [20], [21]. The only description of horizontal transmission of MHV-68 occurred in two uninfected mouse mothers which had eaten their diseased offspring previously inoculated with the virus [23]. This limited description leaves therefore unresolved how MHV-68 is spread in wild rodent hosts [20], [21]. Different hypotheses can be mounted to explain these poor results. Firstly, conventional animal caging could not allow physiological behaviours observed in the wild such as scent-marking or male fighting. Secondly, although the MHV-68 life cycle in mice following experimental infection is considered as well-known, unexplored inoculation routes could lead to important differences.
Methods available to follow viral infections are constantly evolving, becoming more sensitive and efficient. Recently, a bioluminescence imaging technique has been developed to measure the activity of luciferase reporters in living mice noninvasively and repetitively [24]. This technique has been successfully applied to MHV-68 [17], [19], [25]. In this study, we pursued this work. This allowed us to detect infectious virus in the genital tract of female mice after the time of latency establishment. This presence of infectious virus in the genital tract of latently infected females was transient and under the dependance of sexual steroid hormones. Strikingly, presence of infectious virus in female genital tract allowed us to observe sexual transmission of MHV-68 to naïve males.
The main advantage of whole body imaging of luciferase-expressing MHV-68 cycle in living mice is that it reveals novel sites of viral replication. Therefore, we infected 6 weeks-old female BALB/c mice intranasally under general anaesthesia with 104 PFU of luciferase+ MHV-68 and tracked infection daily by luciferin injection and charge-coupled-device camera scanning. Representative images are shown in Figure 1A. As previously described [17], [19], [25], we observed signals coming from the nose (d4 p.i.), the thoracic region (d7 p.i.), the neck (d14 p.i.) and the left abdominal region (d14 p.i.). Based on former descriptions [19], [25], [26], we considered the nose signals to come from the nasal turbinates [26]; thoracic signals from the lungs; neck signals from the superficial cervical lymph nodes (SCLNs); and the abdominal signals from the spleen. As previously described [17], [19], [25], the nose and lung signals peaked at 5–7 days after infection and were undetectable after day 14. On the opposite, signal appeared around day 7 in SCLNs and was maximal at day 14, the peak of latency amplification. SCLNs signal then disappeared over the two following weeks. Signals appeared in the spleens around day 10 but were more transient and less often observed than in the SCLNs.
Surprisingly, we randomly observed appearance of luciferase signal in the genital region of infected female mice (Figure 1A–B). This signal appeared after the initial clearance of acute lytic replication in nose and lungs. Moreover, the signal in the genital region was concomitant or appeared after disappearance of the SCLNs and spleen signals (Figure 1A–B and Figure S1). To further investigate MHV-68 replication in the female genital region, we followed it over time among different mice (Figure 1C). Interestingly, ∼80% of the mice displayed luciferase signal in the genital region during this period. This signal was transient (no more than 4 consecutive days) and recurrent. To confirm the sites of infection and to further investigate the origin of the signal, ex vivo imaging of individual organs was performed after euthanasia of luciferase+ MHV-68 infected mice. This approach revealed that the luciferase signal observed in the genital region was coming from small regions of the vagina (Figure 2A). Fragments of vagina identified as positive for light emission were dissociated from the rest of the organs (Figure 2A) and processed for histological analysis. Immunohistochemical staining for viral antigens identified focal sites of MHV-68 antigen expression in the superior layers of the vaginal epithelium (Figure 2B). This was associated with morphological changes of infected cells (Figure 2B, panel iii) and with the presence of leukocytic infiltrate in the lamina propria (Figure 2B, panels ii and v). These lesions were not observed every time, likely because of their restricted size.
In order to further investigate this observation, 12 mice were infected intranasally and light emission from the genital region was measured 23 days p.i. (Figure 3A). This allowed us to categorize mice into two groups: the first in which genital signal was observed was called IVIS+ and the other IVIS-, three uninfected mice were used as mock infected controls. Genital tracts of these mice were isolated as shown in Figure 2A and light emitting regions of the vagina were isolated. Equivalent regions were isolated in mock and IVIS- groups. These different samples were then analyzed by infectious center assays, infectious virus titration and viral genome quantification (Figure 3B–D). These experiments identified the presence of reactivable virus (Figure 3B) and infectious virions (Figure 3C) only in the IVIS+ group. Moreover, there were statistically more copies of MHV-68 genome in the IVIS+ samples than in the IVIS-. Finally, titration of vaginal lavage fluids, collected before euthanasia, revealed the presence of infectious virions in half of the IVIS+ samples (Figure 3E). The latter experiment was repeated on a higher number of mice between days 21 and 30 post-infection (Figure S2A). This revealed that excretion of infectious MHV-68 virions in female genital tracts occurred randomly and was limited in terms of number of PFUs. Finally, in order to establish that nonmanipulated WT MHV-68 parental strain exhibits properties similar to the tagged virus, we have compared viral shedding in the vagina of mice infected either by the WT or by the WT-LUC MHV-68 strains. Our results showed that both viral strains are excreted similarly in the vagina (Figure S2B) for similar latency loads in spleen (Figure S2C). All together, these experiments showed that MHV-68 luciferase signal in female genital tract is associated with the presence of infectious virus in the vaginal epithelium and in the vaginal fluids. Moreover, we observed similar results with the WT parental strain. This could therefore represent a potential portal of transmission of this virus.
Random and recurrent observations of MHV-68 associated luciferase signal in female genital tract suggest an association of this phenomenon with the estrous cycle. Indeed, female hormones influence susceptibility, reactivation and transmission of many viruses, including human herpesviruses [27]–[29]. To investigate this possibility, we compared occurrence of MHV-68 associated luciferase signal in genital tract among groups of control and ovariectomized female mice between days 14 and 32 post-infection (Figure 4A and S3). This revealed that ovariectomy greatly diminished observation of MHV-68 associated luciferase expression in the genital tract (Figure 4A and S3) although the normal progression of MHV-68 infection was not affected by the treatments (Figure S4). Indeed, no difference of luciferase signals was observed in lungs at day 7 post-infection between non-ovariectomized and ovariectomized mice (Figure S4A). Similarly, normal lymphoid infection was consistently observed from day 14 to day 21 post-infection (Figure S4B). In order to indentify if it was associated with specific hormonal deprivation, we implanted ovariectomized mice with slow-release progesterone and/or estrogen pellets (Figure 4A and S3). Estrogens alone or in combination with progesterone were sufficient to restore occurrence of genital luciferase signal to levels similar to the ones observed in the non ovariectomized group.
Estrogens could promote genital shedding of MHV-68 by either interacting directly with the infected cell or indirectly, for example by inhibiting the immune response against the virus. To determine whether estrogen treatment can directly trigger MHV-68 reactivation from latently infected cells, we used murine A20 B cells latently infected with MHV-68 (Figure 4B) or explanted splenocytes from MHV-68 infected mice, 14 days p.i. (Figure 4C). These cells were treated with increasing amounts of 17β-Estradiol and MHV-68 reactivation was analyzed by infectious center assays. The results obtained did not show that estrogen stimulation of latently infected cells induces MHV-68 reactivation. The observed effect of estrogens on occurrence of MHV-68 associated luciferase signal (Figure 4A and S3) could therefore be indirect or cell-type specific.
The presence of MHV-68 replication in latently infected females could affect gestation. To investigate this hypothesis, luciferase+ MHV-68 infected female mice were mated with uninfected males at the time of the first observation of genital signal. Mock infected female mice were used as controls. Effect of MHV-68 infection on litter size (Figure 5A), mortality/litter (Figure 5B) and gestation length (Figure 5C) was then monitored. We did not observe any effect of MHV-68 infection on any of these parameters (Figure 5A–C). Moreover, we also did not observe transmission to the progeny either at birth or after 3 or 6 weeks (Figure 5D). These results were confirmed by in vivo imaging. Briefly, we have imaged latently infected pregnant mothers (n = 30) during gestation (days 18–20 post-mating) and 2 weeks after delivery. Only one pregnant female displayed weak genital signal around delivery (data not shown) whereas all the others (29/30) had no detectable genital signal (Figure S5A and B). 2 weeks after delivery, none of these mice and their offspring displayed any detectable luciferase signal (Figure S5C). Similarly, we also did not observe seroconversion (Figure 5E) or detectable levels of MHV-68 DNA in the spleen of co-housed naïve female mice (data not shown).
To determine whether the presence of infectious virus in the vaginal epithelium and in the vaginal fluids can result in sexual transmission of MHV-68, we mated luciferase+ MHV-68 infected female mice with uninfected males at the time of the first observation of genital signal. We then tested transmission to males by serology at day 10 post-contact and more than 20 days post-contact. Interestingly, we observed seroconversion of 23 individuals among the 60 males that were tested (Figure 6A). As this seroconversion was moderate in comparison to the one observed after intranasal infection (Figure 6A), presence of MHV-68 DNA in spleens was tested. 24 out of these 60 males (among which the 23 that had seroconverted) displayed detectable levels of MHV-68 DNA in the spleen (Figure 6B). All together, these results therefore show that MHV-68 can be transmitted from infected female mice to naïve males.
To determine the route of MHV-68 transmission to naïve males, we repeated the previous experiment and tracked MHV-68 infection of males daily by luciferin injection and charge-coupled-device camera scanning (Figure 7). We observed that light emission appeared in the genital region around 4 days post-contact. This signal peaked around 10 days post-contact but was maintained for at least 3 weeks. To confirm the site of infection and to further investigate the origin of the signal, ex vivo imaging of individual organs was performed after euthanasia of luciferase+ MHV-68 infected males at different time points. This approach revealed that the luciferase signal observed in the genital region was coming from small regions of the penis (Figure 8A). Fragments of the penis identified as positive for light emission were dissociated from the rest of the organ (Figure 8A) and processed for histological analysis. Immunohistochemical staining for viral antigens identified focal sites of MHV-68 antigen expression in the superior layers of the penis epithelium and of the corpus cavernosum (Figure 8B). Viral antigens were also detected in deeper regions of the corpus cavernosum (Figure 8B, panel iii). Penis infection was associated with propagation of the infection to draining lymph nodes. Ex vivo imaging revealed that they were mainly lumbar aortic medial iliac lymph nodes (Figure 9). Light emitted by these lymph nodes had already been observed during imaging of living animals (Figure 7, days 13 to 17). Finally, colonization of the spleen was observed (Figure 7, days 15 to 19) as already showed by viral genome detection (Figure 6B). Interestingly, genital signal in males was never observed after intra-nasal infection (Figure S6).
We studied female to male transmission to validate the physiological significance of virus shedding from the female genital tract. We saw no virus shedding from the male genital tract after intranasal infection (Figure S6), so the requirements and routes of male to female transmission are less clear. To explore it we relied on the acute genital signal of infected males. Thus, we infected 30 female mice intranasally (104 PFU) with luciferase+ MHV-68, then imaged them each day by luciferin injection and CCD camera scanning. When genital signal was observed, infected females (3 per cage) were mixed with uninfected males (3 per cage). Male infection was then monitored by luciferase imaging and by serology at 10 and 20 days post-contact. We observed genital luciferase signal and positive serology in 13/30 males. These results are incorporated into Figure 6A. Luciferase+ males were then mixed with uninfected females (3 per male). We did not observe luciferase signal in these females, but one seroconverted (Figure S9) and infection was confirmed by Q-PCR of viral DNA from the spleen (10.4 viral genome copies/1000 spleen cells). Low dose intranasal infection (10 p.f.u.) often leads to seroconversion without detectable nasal luciferase signal (data not shown). Thus this was a bona fide infection, but by an undetermined route.
Transmission in host population is the main motor of viral evolution [30]–[32]. Herpesviruses have co-evolved with their host for millions of years and have therefore developed sophisticated mechanisms to persist and transmit in presence of protective immune response [33], [34]. This is particularly the case for gammaherpesviruses [20], [35], [36]. Until now, most of the immune evasion strategies of gammaherpesviruses have been studied in vitro or in animal models [20], [35], [36]. However, none has been investigated in the light of transmission mainly due to the lack of experimental transmission model. In this study, using in vivo imaging, we observed that MHV-68 is genitally excreted after latency establishment in intranasally infected female mice (Figures 1–3, S1 and S2). This allowed us to observe, for the first time, experimental transmission to naïve males after sexual contact (Figures 6–9).
The observation of vaginal shedding of MHV-68 is somewhat surprising as numerous people have been working on this model around the world for a long time without reporting such observations. However, several points can be mentioned. First, we have used a method of in vivo imaging that was recently developed and which is particularly sensitive, allowing the detection of low levels of replicative virus [17], [37]. Secondly, we have followed the infections daily and during a long period, generally between 14 and 32 days post infection. To our knowledge, such following of the infectious process has never been reported. Interestingly, the two previous studies using in vivo luciferase imaging of MHV-68 cycle suggested potential genital infection. Milho et al. showed that the female genital tract is a site of virus replication after intraperitoneal infection [19] and one of the mice used by Hwang et al. displayed light emission in the genital region after intranasal infection and latency establishment (Hwang et al., Figure 2A, day 18 p.i. [25]). The high frequency of genital signal observation in our study (∼80% of the infected mice) could reflect particular experimental conditions. For example, a potential co-infection with another pathogen could favour MHV-68 genital excretion. Such synergic relation has been demonstrated for others herpesviruses, notably HSV-2 and human cytomegalovirus, with HIV-1 [38]–[41]. If such a pre-existent infection exists, the causal agent remains to be identified. Another explanation could be related to our housing facility which homes both females and males mice. This can be an important element as male pheromones can modulate estrous cycle in mice [42].
The observation of genital signal in females was dependent on the estrous cycle as ovariectomy nearly abolished the phenomenon and as estrogens supplementation restored it (Figure 4). Steroid hormones influence susceptibility, replication as well as transmission of many viruses, including herpesviruses [43], [44]. Numerous studies have illustrated the influence of female sex hormones on both susceptibility and immune responses to sexually transmitted pathogens [43]–[45]. Thus, estrogen and progesterone influence the susceptibility to genital herpes infection [43], [44]. However, in those cases the presence of progesterone increased susceptibility to HSV-2 [46], whereas the presence of estrogen prevented or decreased the risk of HSV infection in the female genital tract [47]–[49]. While these hormones can directly influence the sensitivity of the cells of the genital tract, they could also attenuate or modulate the innate [50], [51] and/or the adaptive [44] immune response against the virus. For example, the abundance of antigen presenting cells, T cells and B cells has been shown to vary in uterus and vagina with the estrous cycle [52]–[54]. Replication and shedding of MHV-68 could therefore be a consequence of impaired immune surveillance. On another hand, as some of these immune cells harbour MHV-68 latent infection [21], [36], the transient observation of genital signal in females could reflect the variation of abundance of some particular cell types over time. In the future, these point will have to be tackled. Thus, the origin of vaginally excreted virions could be addressed by using cell-type specific Cre/Lox genetic labelling of MHV-68 to track the route of viral excretion in vivo as it has recently been done to explore the host colonization pathway [26]. Interestingly, the fact that the MHV-68-associated luciferase genital signal lasted for at most 3–4 days (Figure 1C) could be linked to the cyclic remodelling of the epithelium observed during the estrous cycle. Indeed, infected cells are located in the superior layers of the vaginal epithelium (Figure 2B) and could therefore be removed at each cycle.
Besides these indirect roles, steroid hormones have also been shown to directly induce herpesvirus reactivation from latency. Thus, 17β-estradiol promotes HSV-1 reactivation in latently infected neurons [29]. Similarly, several studies have shown that dexamethasone, a synthetic corticosteroid, induces Bovine Herpesvirus-1 reactivation from latency either in vitro or in vivo in calves [55] and rabbits [56]. This has been associated with the induction of cellular transcription factors and/or signalling pathways that stimulate viral lytic genes expression and subsequent reactivation [57]–[60]. In the present study, we did not observe any direct effect of estrogens on latently infected B-cells either in vitro or ex vivo (Figure 4 B and C). However, we have no evidence that these cells mimic what happens in vivo in the infected female genital tract. Further experiments are therefore required to identify the mechanism involved in estrogen-induced MHV-68 vaginal shedding.
In males, initial infection was localized in the superior layers of the penis epithelium and of the corpus cavernosum (Figure 8B). Infection then spreads to draining lymph nodes and spleen (Figure 7 and 9). Again, cell-type specific Cre/Lox genetic labelling of MHV-68 [26] will be helpful to track the route of viral infection after sexual transmission. As infectious virions were rarely detected in vaginal lavages although MHV-68 induced luciferase signal was frequent, we hypothesize that close contacts between genital organs of males and females are necessary to transmit infection. Indeed, the penis of the male mice is recovered of spines called filiform papilla. These structures could therefore induce abrasion of the vaginal epithelium and promote virus transmission. Interestingly, cells that were initially infected on penis were located around these filiform papilla (Figure 8B). Infection persisted at this site for at least three weeks (Figure 7). The importance of this observation for MHV-68 epidemiology will therefore have to be tested. For example, it has recently been shown that male circumcision significantly reduces the incidence of HSV-2 and HIV-1 infection and the prevalence of HPV infection [61], [62]. Our results suggest that it could also be the case for some gammaherpesviruses.
Until now, we did not manage to establish experimental conditions to repeatedly transmit the virus from genitally infected males to naïve females. Human herpesvirus transmission generally occurs at a low rate even between close contacts [63]. However, our results (Figure S7) can only be considered suggestive of male to female MuHV-4 transmission. Important aspects of mouse behaviour such as scent marking may not be properly reproduced with conventional housing. We conclude that under the experimental conditions used, male to female transmission is possible but inefficient, certainly much less so that female to male transmission. The normal mode of male to female MuHV-4 transmission remains to be determined.
Sexual transmission constitutes an easy way of spread for a virus in natural populations of wild animals. This is particularly the case for rodents. Indeed, rodents live generally in small groups spread on a relatively large territory. Sexual contact could therefore be a relatively efficient route of transmission. Interestingly, Telfer et al. showed that gammaherpesvirus (identified serologically as MuHV-4, though likely Wood Mouse Herpesvirus) infection in wood mice was more prevalent in heaviest, sexually active, males than in any other category of animal [64]. The fact that the viral shedding in the female genital tract is linked to sexual cycle and more precisely to the period of estrus (high rates of estrogens) would be very beneficial for transmission as re-excretion would occur during the periods of female receptivity for mating. Sexual transmission has also been proposed for EBV and KSHV [9], [65], [66] but is mainly important for HSV-1 and -2 [67]. The observation of MHV-68 sexual transmission from infected females to naïve males could therefore be particularly interesting in the general context of herpesvirus transmission.
Shedding of MHV-68 in the female genital tract could also have an effect on progeny. However, in contrast to what was reported by Stiglincova et al. [68], we did not observe premature termination of pregnancy, reduced number of newborns, vertical transmission or transmission through milk of MHV-68 in mice (Figure 5). We have no explanation for this discrepancy. However, mother to child transmission of human gammaherpesviruses, both transplacental or perinatal, seems also to be very limited [9], [69].
The identification of a route of transmission for MHV-68 in mice opens new fundamental research perspectives. Thus, it will allow testing the importance of various immune evasion strategies, such as those based on the gp150 glycoprotein [70], [71] in the light of transmission. It will also be interesting to test if transmission requires latency establishment and reactivation of the virus or, conversely, if it is enhanced by immunosuppression through the use of drugs like Cyclosporine A [25] or of depletion of specific cell types such as CD8 [72]. Finally, it will be possible to test antiviral and/or vaccinal strategies in the context of infection epidemiology.
Altogether, in this study we identified for the first time a genital excretion site of MHV-68 after latency establishment in intranasally infected female mice. This has allowed us to observe sexual transmission of the virus from infected females to naïve males. These results open new perspectives for the study of gammaherpesvirus in particular but also for the study of sexually transmitted infections in general.
The experiments, maintenance and care of mice complied with the guidelines of the European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes (CETS n° 123). The protocol was approved by the Committee on the Ethics of Animal Experiments of the University of Liège, Belgium (Permit Number: 1051). All efforts were made to minimize suffering.
Females and males BALB/c mice were purchased from Charles River Laboratories. All the animals were housed in the University of Liège, Department of infectious diseases. The animals were infected with MHV-68 when 6–12 weeks old. Intranasal infections with anaesthesia were in 30 µl aliquots. For luciferase imaging, animals were anaesthetized with isoflurane, injected intraperitoneally with luciferin (150 mg/kg), then scanned with an IVIS Spectrum (Caliper Life Sciences). Animals were routinely imaged after 10 min. For quantitative comparisons, we used Living Image software (Caliper Life Sciences) to obtain the maximum radiance (photons per s per cm2 per steradian, i.e. photons s−1 cm−2 sr−1) over each region of interest.
We used the MHV-68 strain of MuHV-4 [16] and a MHV-68 strain expressing luciferase under control of the M3 promoter that was described previously (hereafter called WT-LUC strain) [19]. Briefly, a luciferase expression cassette was inserted between the polyadenylation signals of ORFs57 and 58. This viral strain did not show any growth deficit either in vitro or in vivo [19]. Viruses used in this study were propagated on BHK-21cells cultured in Dulbecco's modified Eagle's medium (Invitrogen) and supplemented with 2 mM glutamine, 100 U penicillin ml−1, 100 mg streptomycin ml−1 and 10% fetal calf serum. Virions were concentrated as described previously [70].
Virus stocks were titrated by plaque assay on BHK-21 cells [73]. Cell monolayers were incubated with virus (2 h, 37°C), overlaid with 0.3% carboxymethylcellulose (CMC, medium viscosity, Sigma), and 4 days later fixed and stained for plaque counting [74]. Infectious virus in organs was measured by homogenizing them after freezing (−80°C) in 6 ml complete medium prior to plaque assay.
Virus detection in genital organs cell suspension was assayed by infectious centre assay (ICA) as follows. 5.105 BHK-21 cells grown in 6 well cluster dishes (Becton Dickinson) were co-cultured for 5 days at 37°C with ex vivo cell suspension in MEM containing 10% FCS, 2% PS, 0.3% CMC and 5.10−5 M of β-mercaptoethanol (Merck). Cells were then fixed and stained for plaque counting.
Viral genome loads were measured by real-time PCR [17]. DNA from organs (100 ng) was used to amplify MHV-68 genomic co-ordinates 4166–4252 (iCycler, Biorad) (gene M2, forward primer 5′- GTCAGTCGAGCCAGAGTCCAACA-3′, reverse primer 5′-ATCTATGAAACTGCTAACAGTGAAC-3′). The PCR products were quantified by hybridization with a TaqMan probe (genomic co-ordinates 4218–4189, 5′ 6-FAM-TCCAGCCAATCTCTACGAGGTCCTTAATGA-BHQ1 3′) and converted to genome copies by comparison with a standard curve of cloned plasmid template serially diluted in control spleen DNA and amplified in parallel. Cellular DNA was quantified in parallel by amplifying part of the interstitial retinoid binding protein (IRBP) gene (forward primer 5′-ATCCCTATGTCATCTCCTACYTG-3′, reverse primer 5′-CCRCTGCCTTCCCATGTYTG-3′). The PCR products were quantified with Sybr green (Invitrogen), the copy number was calculated by comparison with standard curves of cloned mouse IRBP template amplified in parallel. Amplified products were distinguished from paired primers by melting curve analysis and the correct sizes of the amplified products confirmed by electrophoresis and staining with ethidium bromide.
Vaginal lavage fluids were obtained by gentle flushing of the mouse vagina with 100 µl of sterile PBS. Lavage fluids were then centrifuged and the supernatant was titrated as described above.
Ovariectomy were performed at 3 weeks of age under isoflurane anaesthesia. Hormonal treatment was started 3 weeks after ovariectomy. 60 days slow-release pellets (Innovative Research of America, Sarasota, FL, USA) containing 0.05 mg 17β-estradiol (SE-121), or 25 mg progesterone (SP-131) per pellet were implanted subcutaneously, giving a release of ∼0.8 µg 17β-estradiol or ∼400 µg progesterone per 24 hours.
17-β-estradiol (Sigma) stock solution was prepared in DMSO (1 mg/ml). For in vitro stimulation, A20-Syndecan-1 cells [75] were persistently infected with a MHV-68 strain expressing eGFP under an EF1a promoter, between the 3′ ends of ORFs 57 and 58 (Dr P.G. Stevenson, unpublished data). For ex vivo stimulation, spleen of WT MHV-68 intranasally infected mice were harvested 14 days post-infection, cells were dissociated and erythrocytes were lysed by using red blood cells lysis buffer. Cells were cultivated in RPMI medium without phenol red, to avoid the presence of steroids, supplemented with 2 mM glutamine, 100 U penicillin ml−1, 100 mg streptomycin ml−1, 5*10−5 M of β-mercaptoethanol (Merck) and 10% Charcoal Stripped Fetal Bovine Serum (CSFBS, Sigma). Stimulation of virus reactivation by 17-β-estradiol was performed as follows. Briefly, 3*105 BHK-21 cells grown in 6 well cluster dishes were co-cultured for 5 days at 37°C with 5*103 MHV-68 infected A20 cells or 5*105 infected spleen cells in RPMI containing 10% CSFBS, 2% PS, 0.3% CMC, 5.10−5 M of β-mercaptoethanol (Merck) and complemented with increasing doses of 17-β-estradiol. After 5 days, cells were fixed and stained for plaque counting.
Portions of genital organs were fixed in buffered formol saline, processed routinely to 5-mm paraffin wax-embedded sections, stained with hematoxylin and eosin, and examined by light microscopy. Immunohistochemistry was performed using EnVision Detection Systems (DAKO) with anti-MHV-68 rabbit hyperimmune serum against MHV-68 as primary antibody [17].
Nunc Maxisorp ELISA plates (Nalgene Nunc) were coated for 18 h at 37°C with 0.1% Triton X-100-disrupted MHV-68 virions (2.106 PFU/well), blocked in PBS/0.1% Tween-20/3% BSA, and incubated with mouse sera (diluted 1/200 in PBS/0.1% Tween-20/3% BSA). Bound antibodies were detected with Alkaline Phosphatase conjugated goat anti-mouse Ig polyclonal antibody (Sigma). Washing were performed with PBS/0.1% Tween-20/3% BSA. p-Nitrophenylphosphate (Sigma) was used as substrate and absorbance was read at 405 nm using a Benchmark ELISA plate reader (Thermo).
|
10.1371/journal.ppat.1003838 | Cell Tropism Predicts Long-term Nucleotide Substitution Rates of Mammalian RNA Viruses | The high rates of RNA virus evolution are generally attributed to replication with error-prone RNA-dependent RNA polymerases. However, these long-term nucleotide substitution rates span three orders of magnitude and do not correlate well with mutation rates or selection pressures. This substitution rate variation may be explained by differences in virus ecology or intrinsic genomic properties. We generated nucleotide substitution rate estimates for mammalian RNA viruses and compiled comparable published rates, yielding a dataset of 118 substitution rates of structural genes from 51 different species, as well as 40 rates of non-structural genes from 28 species. Through ANCOVA analyses, we evaluated the relationships between these rates and four ecological factors: target cell, transmission route, host range, infection duration; and three genomic properties: genome length, genome sense, genome segmentation. Of these seven factors, we found target cells to be the only significant predictors of viral substitution rates, with tropisms for epithelial cells or neurons (P<0.0001) as the most significant predictors. Further, one-tailed t-tests showed that viruses primarily infecting epithelial cells evolve significantly faster than neurotropic viruses (P<0.0001 and P<0.001 for the structural genes and non-structural genes, respectively). These results provide strong evidence that the fastest evolving mammalian RNA viruses infect cells with the highest turnover rates: the highly proliferative epithelial cells. Estimated viral generation times suggest that epithelial-infecting viruses replicate more quickly than viruses with different cell tropisms. Our results indicate that cell tropism is a key factor in viral evolvability.
| RNA viruses are the fastest evolving human pathogens, making their treatment and control difficult. Compared to DNA viruses, RNA viruses replicate with much lower fidelity, which can explain why RNA viruses evolve significantly faster than most DNA viruses. However, there is tremendous variation among the evolutionary rates of different RNA viruses, which is not explained by variation in mutation rates. Here we present a survey of mammalian RNA virus rates of evolution, and a comprehensive comparison of these rates to different properties of virus genomic architecture and ecology. We found that cell tropism is the most significant predictor of long-term rates of mammalian RNA virus evolution. For instance, viruses targeting epithelial cells evolve significantly faster than viruses that target neurons. Our results provide mechanistic insight into why viruses that infect respiratory and gastrointestinal epithelia have been difficult to control.
| RNA viruses are responsible for a disproportionate number of emerging human diseases, including influenza, ebola hemorrhagic fever, hantavirus pulmonary syndrome, and Middle East respiratory syndrome, which place tremendous health and economic burdens on both the developing and developed world [1], [2]. In 2008, rotavirus and measles virus caused the deaths of 570,000 children under the age of five, making them two of the leading killers of children worldwide [3]. In 2009, it was estimated that rotavirus infections alone result in $325 million in medical treatment costs and $423 million in societal costs each year [4]. Further, the implementation of many intervention strategies has either failed or been delayed as a result of the evolutionary dynamics of these pathogens [1], [5], [6], [7], [8], [9].
Differences in viral evolutionary dynamics, such as rates of evolution, can explain why certain viruses have the capacity to adapt to new host species, increase in virulence, or develop resistance to antivirals [7], [8], [9], [10], [11]. Therefore, understanding why some RNA viruses evolve more quickly can facilitate better prediction of their pathogenic and epidemiological potential [8], [10], [11], [12]. Though extremely high nucleotide substitution rates are a defining feature of RNA virus evolution [1], [13], [14], [15], there have been few attempts to comprehensively examine the driving genomic and ecological factors behind these rates.
Differences in the strength and direction of selection pressures on these viruses result in variation among their substitution rates [1], [5], [13]. However, while some general patterns have been observed in selection pressures, such as enhanced purifying selection on the structural proteins of arboviruses [16], there have been no attempts to quantify the relationship between selection pressures and long-term viral substitution rates.
The high rates of RNA virus evolution are most commonly attributed to their replication with error-prone RNA-dependent RNA polymerases (RdRps) [1], [17], but these nucleotide substitution rates are known to span at least three orders of magnitude [5], [17] and do not correlate well with experimentally measured viral mutation rates [5]. Further, the substitution rates of some DNA viruses, which replicate with high-fidelity DNA polymerases, are comparable to the high substitution rates of RNA viruses [13]. Therefore, the polymerase error rate alone cannot explain the substitution rate variation in RNA viruses.
Along with mutation rate, viral replication frequency directly impacts the rate at which mutations can be introduced, and ultimately fixed as substitutions [13]. Replication frequencies could be influenced by a variety of factors related to viral genomic architecture or ecology [13]. For example, weak negative correlations between viral genome lengths and substitution rates have been attributed to either enhanced replication frequencies or higher mutation rates in viruses with smaller genomes [15], [17], [18], [19]. It has also been suggested that different transmission and infection modes result in differences in generation time, ultimately causing variation among per-year rates of synonymous substitution of RNA virus structural genes [5].
In this modern survey of mammalian RNA virus evolution rates, we generated and compiled published substitution rates of structural and non-structural genes produced by Bayesian coalescent analyses [20]. We analyzed these rates as a function of seven factors related to virus genomic architecture (i.e., genome length, genome sense, and whether or not the genome is segmented) and virus ecology (i.e., target cell, transmission mode, host range, and whether the infection is acute or persistent). We also evaluated the relationships of viral substitution rates with dN/dS estimates, experimentally measured mutation rates, and estimated generation times. Though recombination undeniably plays a role in shaping viral evolutionary dynamics and could inflate substitution rate estimates [21], [22], we conservatively removed any potential recombinants from our datasets prior to analysis. Through this broad analysis, we were able to demonstrate that cell tropism, and its impact on viral generation time, has the greatest influence on rates of mammalian RNA virus evolution.
A review of the literature yielded 92 published Bayesian nucleotide substitution rate estimates for the structural genes of 35 different mammalian RNA viral species, and 21 published Bayesian rates for RdRps or a non-structural gene of 14 different viral species (referred to collectively as “non-structural,” Table S1). These rates were supplemented with 26 novel Bayesian substitution rates of structural genes of 19 different viral species, and 19 novel Bayesian rates of non-structural genes of 16 different viral species (Table S2). Collectively, these rates span three orders of magnitude, ranging from 3.0×10−5 to 1.5×10−2 nucleotide substitutions per site per year (ns/s/y) and 2.0×10−5 to 1.3×10−2 ns/s/y for the structural genes and non-structural genes, respectively (Table S1).
Plotting the levels of each variable by ascending mean substitution rate revealed similar patterns (i.e., the same ordering of levels) for both the structural (S) and non-structural (NS) datasets in three of these variables, excepting transmission route. Viral substitution rates grouped according to target cell (panels 1A and 1B), transmission route (panels 1C and 1D), infection type (panels 1E and 1F), and host range (panels 1G and 1H) are shown in Figure 1.
Substitution rates were also grouped by viral genomic architecture (genome sense/strandedness, Figure 2A and 2B, and genome segmentation, Figure 2C and 2D) and plotted against viral genome length (Figure 2E and 2F). There were no apparent relationships between genomic properties and substitution rates (Figure 2), including no linear relationship between substitution rates and genome lengths in either dataset (coefficient of determination, S: R2 = 0.06, NS: R2 = 0.08).
dN/dS estimates calculated in this study were compiled with published estimates also calculated using the Single Likelihood Ancestor Counting (SLAC) method (56 structural gene dN/dS estimates, 33 non-structural gene dN/dS estimates total, Table S1).
ANCOVA analyses were performed separately on the structural and non-structural gene datasets to determine which, if any, of seven factors (target cell, transmission route, infection mode, host range, genome length, genome sense, and genome segmentation) significantly predict the nucleotide substitution rates of mammalian RNA viruses. To explore the many dummy-coded categorical variables, three analyses were run using different variable levels as the base levels (see Methods for details, Tables 1 and 2). For all of the ANCOVA analyses, the adjusted coefficient of determination () was ≥0.73, indicating that over 70% of the substitution rate variability can be explained by the predictor variables included in this study. Standardized residual plots identified only six potential outliers of the 118 structural gene rates and one potential outlier of the 40 non-structural gene rates (Figure S1), indicating that the data are normally distributed and therefore amenable to a general linear model.
Regardless of the base levels, target cells were the only significant predictors of log-transformed substitution rates for both structural and non-structural genes (Tables 1 and 2), with cell tropism as the only significant predictor variable by type III sum of squares (SS) analyses (P<0.0001 and P = 0.003 for the structural and non-structural gene datasets, respectively). Targeting epithelial cells or neurons was found to be the most significant predictor of structural gene rates in each analysis where these were not the base levels (P<0.0001, Table 1, Figure 3), while targeting neurons was found to be the sole significant predictor of substitution rates for the smaller non-structural gene dataset (P = 0.009, Table 2, Figure 3). Further, there was a high correlation between each viral species' estimated structural gene substitution rate and its corresponding non-structural gene rate (33 viruses, Pearson r = 0.87, P<0.0001). This suggests that if it were possible to calculate more non-structural rates, we would likely see results similar to those from the structural gene dataset.
To minimize any potential bias introduced by using multiple published rates for a single viral strain or species, we conducted control analyses using datasets with only one rate per species. For species with multiple substitution rates in one of our datasets, we calculated the average log substitution rate and used that as the sole substitution rate for the species in the control analysis. These data were also normally distributed (Figure S2), but the for these analyses were slightly lower than for the full datasets (S: = 0.65, NS: = 0.70, Tables S3 and S4). These control results were consistent with those from the full dataset analyses: tropisms for epithelial cells or neurons were the most significant substitution rate predictors (Tables S3 and S4, Figure S3).
Because of the high correlation between the structural and non-structural gene rates, we combined the two datasets (Figure 4) and performed a final set of three ANCOVA analyses using this combined dataset. The results from these analyses were nearly identical to those from the structural gene analyses (Table S5). The exception was that, in addition to cell tropism, Type III SS analysis also identified transmission route as a significant predictor variable (P = 0.007), though it was still less significant than cell tropism (P<0.0001). More specifically, in addition to different cell tropisms, transmission through arthropod vectors was also found to be a significant rate predictor in one of the three analyses (P = 0.002, Table S5).
To ensure that any substitution rate variability attributed to a given predictor variable was not significantly dependent on other predictor variables, we examined collinearity in all datasets. With the exception of the persistent infection variable, which was nested with the endothelial target cell variable and thus excluded, the ANCOVA analyses for the structural gene rate datasets and the combined rate dataset showed no significant collinearity (no variance inflation factors (VIF) were greater than 10). For the non-structural gene rate datasets, many different predictor variables had VIF>10. However, subsequent analyses where each individual variable was removed did not significantly reduce collinearity in these datasets (data not shown). Due to the consistent results between the structural and non-structural gene datasets, as well as those from the combined rate dataset, we concluded that correlations among independent variables did not significantly impact our results.
Since target cells were found to be the only consistently significant predictors of substitution rates, a series of one-tailed t-tests was used to confirm which cell tropisms are associated with higher viral substitution rates than others. Viruses that target epithelial cells were found to have significantly higher structural gene substitution rates than viruses that target neurons, endothelial cells, or leukocytes (Table 3, P<0.0009). Similarly, viruses that target epithelial cells were found to have significantly higher non-structural gene substitution rates than viruses that target neurons, hepatocytes, or leukocytes (Table 4, P<0.0007). These results were recapitulated in the control datasets that only used one rate per viral species (Tables S6 and S7). It should be noted, however, that most of the viruses in this study that are classified as targeting leukocytes ultimately cause systemic infections and infect a wide variety of cell types. Consequently, viruses in the leukocyte target cell category had the most rate variation of all the target cell categories (Figure 1).
Because transmission through arthropod vectors was also found to be a significant rate predictor in the ANCOVA analyses based on the combined datasets and because of the correlation between epithelial cell tropism and fecal-oral/respiratory transmission, we evaluated any significant variation among substitution rates of viruses with different transmission routes. Using a series of one-tailed t-tests, we found that viruses that are transmitted through the fecal-oral/respiratory route have significantly higher substitution rates than those transmitted by arthropod vectors (P<0.0001). However, we also compared different cell tropisms within each of these transmission routes. We found that fecal-oral/respiratory transmitted viruses that target epithelial cells have significantly higher substitution rates than those that target other cell types (P<0.0001, Figure 5). Similarly, we found that neurotropic arboviruses have significantly lower substitution rates than arboviruses that target other cell types (P<0.001, Figure 5).
We also tested for linear relationships between viral substitution rates and other evolutionary parameters for which only smaller subsets of our datasets could be analyzed. Reliable experimentally measured mutation rates estimated as mutations per base per infectious cycle were only available for four different viruses included in this study (poliovirus 1 [11], [23], [24], hepatitis C virus [25], influenza A virus [26], [27], [28], influenza B virus [26]). Mutation rates measured as mutations per base per strand replication were only available for three viruses included in this study (poliovirus 1 [29], measles virus [30], [31], and influenza A virus [32]). These mutation rates were not significantly correlated with their corresponding substitution rate estimates (r = 0.69, P = 0.31 and r = −0.93, P = 0.25, for mutation rates measured as mutations per base per infection and mutation rates measured as mutations per base per replication, respectively). Similarly, there were no significant correlations between the estimated substitution rates and dN/dS estimates (ρ = −0.02, P = 0.88 and ρ = −0.07, P = 0.68, for the limited structural gene and non-structural gene datasets, respectively).
ANCOVA and t-tests consistently revealed epithelial cell tropism and neurotropism as the most significant viral substitution rate predictors. Since these two cell types have some of the highest and lowest turnover rates, respectively, of all mammalian cells [33], [34], [35], [36], we sought to determine if there were any associations between host cell turnover rate and viral generation time. Using the model proposed by Sanjuán (2012) that relates the long-term substitution rate, K, to the mutation rate, μ, correcting for transient deleterious mutations, we were able to estimate generation times for the few viruses with reliable mutation rate estimates. This model, , with , (G = genome length, g = generation time, sH = harmonic mean of the selection coefficient) [15], confirmed that influenza A virus, influenza B virus, and poliovirus, which target epithelial cells, have substantially shorter generation times (<40 hours) than hepatitis C virus, which targets hepatocytes (>200 hours). These results, while based on a very limited dataset, provide quantitative evidence for a link between cell tropism and generation time. Shorter average generation times lead to more rounds of replication per year, which could neatly explain higher per-year substitution rates.
A variety of intrinsic and ecological factors could plausibly alter the tempo of virus evolution by influencing the rate at which genetic diversity is generated, maintained, and fixed within viral populations. Others have focused on genomic properties as drivers of substitution rate variation [14], [15], [17], [18], demonstrating a weak negative correlation between the genome lengths and substitution rates of RNA viruses [15], [17] or suggesting that ssRNA viruses evolve faster than dsRNA viruses [15]. However, we did not find any significant relationship between genomic properties and substitution rates (Figures 2 and 3). While some have conducted more limited studies on the influence of ecological factors [5], [37], we performed a comprehensive analysis that revealed that cell tropism is a key factor in understanding mammalian RNA viral substitution rates.
It has been proposed that persistent viruses evolve more slowly than those that produce acute infections [1], [5], [15], [38]. Unfortunately, with the exception of latent viruses, which are most commonly retro- or DNA viruses and thus not within the scope in this study, it can be difficult to classify viruses as acute or persistent. The duration of persistence can vary; most persistent viral infections begin with an acute phase and may occasionally be resolved after only this acute phase (e.g., HCV), and many viruses that predominantly result in acute infections occasionally persist [39], [40]. By classifying the viruses in this study as accurately as possible, we found no significant association between infection mode and substitution rate. However, only three viruses in this study, all endothelial-infecting hantaviruses, were classified as strictly persistent. This causes the nesting of the persistent level with tropism for endothelial cells, and the persistent infection variable was therefore excluded from our analyses. Infection duration could be a factor explaining substitution rate variation across the Baltimore classifications of viruses, but there is no evidence that it affects the mammalian RNA virus substitution rates included in this study.
Transmission mode and, less explicitly, host range are frequently invoked as determinants of viral substitution rates [5], [41]. Specifically, plant or animal viruses that primarily rely on arthropod vectors for transmission, and therefore obligately infect very diverse hosts, are thought to evolve more slowly than viruses with other transmission modes [5], [41], [42], [43]. Surprisingly, only one of our 15 ANCOVA analyses implicated transmission route as a significant substitution rate predictor, and we found no significant relationship between substitution rate and host range.
The seven genomic and ecological factors examined are not necessarily independent. For example, 25% of the arboviruses in our study are neurotropic, the second-most common cell tropism of our arboviruses (Table S1). Therefore, the observation that vector-borne viruses tend to evolve more slowly is qualitatively consistent with our results. Cell tropism does appear to be the more significant factor, though, as our results show that arboviruses with other cell tropisms evolve significantly faster than those with neurotropism. Previous studies have also indicated that phylogenetic relationships are predictive – that sister taxa have similar rates of evolution [5]. We initially included virus families as an explanatory variable in our analyses, but we had to discard it due to high colinearity with these other seven variables (data not shown). Once the virus families were removed, there was no statistically significant colinearity within the structural gene dataset. Of these seven non-colinear factors, cell tropism was the best predictor of viral substitution rates. The smaller non-structural gene dataset, on the other hand, had significant collinearity among predictor variables that could not be resolved. The NS dataset also had only 1/3 of the taxa, inherently reducing its statistical power. It was not possible to expand the mammalian RNA virus NS dataset at this time; our novel rate analyses increased the number of reliable rates by 40% by exhaustively searching the available sequences in GenBank. The results of the combined dataset were nearly identical to those from the dataset of only S rates, again identifying target cells as the only consistent predictor variables. While many factors likely influence nucleotide substitution rates, and there may be inherent relationships among some of our seven variables, our results affirm that cell tropism is the most significant predictor of mammalian RNA virus substitution rate.
Though previously unexplored, cell tropism could influence viral substitution rates by the same mechanisms that have been suggested for the other ecological factors described above [44]. Infection of different host cells could expose viruses to different selection pressures, which could influence the rates at which mutations are fixed as substitutions. Additionally, it is possible that cell tropism influences the rate at which genetic diversity is generated by affecting viral mutation rates or generation times.
Variation in strength and/or direction of selection has frequently been invoked as a determinant of viral substitution rates [12], [13], [21]. While positive selection can certainly result in variation among very short-term substitution rates, purifying selection tends to dominate over longer timescales [21], [45], [46], [47]. However, variation is observed in the strength of purifying selection due to differences in host ranges. For instance, as previously mentioned, viruses vectored by arthropods have unique evolutionary constraints placed on them by their host diversity [41], [42], [43], [48]. While previous studies found that arboviruses are under stronger purifying selection than non-arboviruses [1], [41], [49], we found that the dN/dS estimates based on structural genes of arboviruses were not significantly lower than those for non-arboviruses (P = 0.19). The dN/dS estimates based on non-structural genes of arboviruses were only moderately lower than those for non-arboviruses (P = 0.04). Further, we found no significant correlation between the estimated dN/dS and substitution rates, suggesting that detectable differences in selection pressures do not explain the variation in substitution rates of mammalian RNA viruses. To date, there are no data supporting a link between cell tropism and sustained differences in selection pressures.
Compared to the slower evolution of DNA viruses, the evolution of RNA viruses is dominated by their high mutation rates [1], [13], [15]. Weak negative correlations between genome lengths and viral substitution rates have been attributed to a relationship between mutation rate and substitution rate, as smaller genomes could in theory withstand higher mutation rates than larger genomes [13], [15], [50]. However, while differences in spontaneous mutation rates appear to be significantly correlated to the long-term substitution rates of DNA viruses [15], this linear relationship disappears past a certain mutation rate threshold: around 10−6 mutations per site per infectious cycle, the lower end of the mutation rate range of RNA viruses [13], [15]. It is, therefore, not surprising that we found no significant correlation between substitution rates and the available, reliable mutation rate estimates. Additionally, a recent study of the retrovirus HIV-1 found that infection of different cell types did not lead to differences in mutation rate [51], providing some evidence that mutation rate is not correlated with cell tropism. Together, these data suggest that mutation rate variation among different cell types is not driving higher substitution rates in epithelial-infecting mammalian RNA viruses.
Ruling out selection, mutation rates, and recombination frequencies as drivers of RNA virus substitution rates implies that the rate variation is largely the result of variation in replication dynamics [5], [13]. Enhanced replication frequencies (shorter generation times) have been used to explain a variety of the previously suggested links between virus ecology and substitution rate. For example, viruses in the acute phase of an infection generally replicate more frequently than those in a persistent infection, and viruses in a latent phase do not replicate at all [39]. Further, as an alternative to differential selection pressures, the argument that transmission mode drives viral substitution rates assumes that viruses that can be transmitted more rapidly will have shorter generation times (e.g., horizontal transmission vs. vertical transmission [5], [52], [53]).
DNA viruses have shorter generation times in faster dividing cells [54], [55], but the associations between cell tropism and RNA virus generation time are less obvious, as RNA viruses do not depend on cellular replication machinery. However, there is evidence that for at least some RNA viruses, viral genome replication is highly dependent on host cell proliferation, with RNA synthesis occurring at much lower rates in poorly proliferating cells than in rapidly dividing cells [56], [57], [58], [59], [60]. For example, it has been repeatedly demonstrated that hepatitis C virus genome replication is enhanced in proliferating cells, perhaps due to higher levels of available nucleotides [59], or because of higher levels of viral protein synthesis facilitated by nuclear translation initiation factors that only become available in the cytoplasm during cell division [58]. Similar dependence on cell proliferation for viral replication efficiency has been demonstrated in a number of picornaviruses [57], [60], [61], [62]. Further, using the model proposed by Sanjuán (2012), we found that viruses that infect epithelial cells have generation times that may be as much as 40-fold shorter than a virus that infects non-epithelial cells. This offers a possible mechanistic basis for our finding that viruses that target the fastest-dividing cells in the body (intestinal and respiratory epithelial cells [34], [35], [36], [63]) have higher substitution rates than viruses that infect cells that turnover at very low rates, if at all (neurons [33], [35], [64]).
We are the first to provide statistical evidence that cell tropism predicts rates of mammalian RNA virus evolution, likely through its influence on virus generation time. These results offer a new perspective on why it has been difficult to create effective vaccines for viruses that infect epithelial tissue, such as rotavirus and enterovirus 71 [65], [66]. Further, as it has been shown that higher rates of viral evolution can result in increased genetic diversity and higher epidemiological fitness [26], [67], [68], the higher substitution rates of epithelial-infecting viruses predict increased evolvability and greater potential for emergence in novel host species [21].
Long-term nucleotide substitution rates of mammalian RNA viruses were collected from the literature, with a focus on finding rates for the outer structural gene containing the major antigenic site(s) and non-structural (preferably the RdRp) genes. While the RdRp genes of the (-)ssRNA and dsRNA viruses are classified as structural, or virion-associated, genes [69], they are generally thought to be more conserved and under very different selection pressures than the structural genes that interact with the host immune system [70], [71]. We excluded retroviruses from analysis because they are known to have highly variable substitution rates due to time spent integrated into DNA genomes, where they evolve at the rate of their hosts' genome [13], [72]. Viruses that predominately infect non-mammals, with mammals serving as incidental, dead-end hosts, were also excluded. Only rates estimated for individual viral species or strains were used, not those that aggregated multiple species into one analysis. Similarly, only rates from single gene analyses were included, not those based on full genomes or multiple gene alignments. In order to minimize any rate discrepancies that could result from variations among datasets (e.g., number of taxa, temporal range, portion of gene analyzed) and/or subtle methodological variations [45], [73], [74], [75], [76], [77], only rates produced by Bayesian coalescent analyses of datasets composed of at least 30 taxa, isolated over a minimum range of 15 years and spanning at least 40% of the analyzed gene were included. Bayesian coalescent analyses provide estimates of viral evolution that are calculated over a longer range than simply the date range over which the taxa were isolated. This is because they determine the likely phylogenetic relationship among the isolates and infer substitution rates over the entire evolutionary history of the sampled taxa: over decades, hundreds, even thousands of years. These rates can therefore be considered “long-term” nucleotide substitution rates.
Data regarding genomic architecture and ecology were obtained for all viruses with published substitution rates that met these criteria. We included multiple rates for a given virus when available, except when a single study examined multiple lineages and summarized the results in a single rate [78], [79], [80], [81]. Corresponding dN/dS estimates were collected when available.
These published substitution rates were supplemented with novel BEAST [20] rate analyses based on the sequence data available in GenBank (accessed through Taxonomy Browser, http://www.ncbi.nlm.nih.gov/Taxonomy). Sequences for structural and non-structural genes with years of isolation available in GenBank or the literature were manually aligned using Se-Al v2.0a11 [82]. Sequences with GenBank or published information that indicated they were genetically manipulated or extensively passaged in the lab prior to sequencing were eliminated from further analysis. The final datasets also adhered to the conservative criteria described above for published datasets.
As recombination events can lead to over-estimation of nucleotide substitution rates, each dataset was scanned for recombination using seven different algorithms (RDP, GENECONV, Bootscan, MaxChi, Chimaera, SiScan, and 3seq) implemented in RDP v3.44 [83]. Sequences implicated as recombinant by two or more algorithms were excluded from further analysis. These finalized alignments were deposited into Dryad (doi:10.5061/dryad.58ss8). Modeltest v3.7 [84] was used to determine the best-fit model of nucleotide substitution for each dataset (by AIC).
Long-term nucleotide substitution rates were estimated using BEAST v1.5.4 [20]. Each dataset was run for at least 50 million generations and until all parameters had stabilized (effective sampling size >200). Each dataset was run with two different clock models (strict and uncorrelated lognormal) and three different demographic models (constant, exponential, and Bayesian skyline). The best-fitting clock/demographic model combination for each dataset was determined using Bayes factors as implemented in Tracer v1.5 [85]. For each best set of priors, two independent runs were performed to ensure that the results were replicable, and a control analysis was run without the dataset to ensure that the priors were not controlling the outcome of the analysis.
The Single Likelihood Ancestor Counting (SLAC), codon-based maximum likelihood method available in the HYPHY package on the Datamonkey web server [86] was used to evaluate the strength of selection pressure on these datasets.
In order to determine which factors most significantly predict substitution rates of mammalian RNA viruses, ANCOVA analyses were run using SPSS Statistics v21 (IBM) with log-transformed mean substitution rates as the dependent variable and seven overarching predictor variables (target cell, transmission route, whether the infection is acute or persistent, host range, genome length, genome sense, and whether or not the genome is segmented). For each variable, different base levels were tested to ensure that the chosen base level did not significantly influence the results. Collinearity among the variables was also assessed, with variance inflation factors (VIF) greater than 10 indicating redundancy among variables. Separate ANCOVA analyses were run on the structural and non-structural gene datasets. As there were multiple published rates for some viral species and strains, additional analyses were run for both the S and NS datasets with only one substitution rate per virus species. When there were multiple rates for a given virus species, we calculated and used an average rate.
One-tailed t-tests were subsequently run in R v2.14.1 [87] to provide an additional measure of significant directional variation among the log-transformed mean rates of different levels for any categorical variable that was found to be a significant rate predictor (α = 0.01, adjusted by Bonferroni correction for multiple comparisons) in the ANCOVA analyses. Additional t-tests were also conducted using the control datasets with one rate per virus species.
Additionally, though there were no dN/dS or mutation rate estimates available for all viruses used in this study, the available data for each variable were compared to corresponding log-transformed mean substitution rate estimates using Spearman rank correlation (for dN/dS) or Pearson correlation coefficient (for mutation rates). Structural and non-structural gene rate estimates were also compared using Pearson correlation coefficient. All correlation analyses were performed in SPSS Statistics v21.
|
10.1371/journal.pntd.0004151 | Impact of an Ivermectin Mass Drug Administration on Scabies Prevalence in a Remote Australian Aboriginal Community | Scabies is endemic in many Aboriginal and Torres Strait Islander communities, with 69% of infants infected in the first year of life. We report the outcomes against scabies of two oral ivermectin mass drug administrations (MDAs) delivered 12 months apart in a remote Australian Aboriginal community.
Utilizing a before and after study design, we measured scabies prevalence through population census with sequential MDAs at baseline and month 12. Surveys at months 6 and 18 determined disease acquisition and treatment failures. Scabies infestations were diagnosed clinically with additional laboratory investigations for crusted scabies. Non-pregnant participants weighing ≥15 kg were administered a single 200 μg/kg ivermectin dose, repeated after 2–3 weeks if scabies was diagnosed, others followed a standard alternative algorithm.
We saw >1000 participants at each population census. Scabies prevalence fell from 4% at baseline to 1% at month 6. Prevalence rose to 9% at month 12 amongst the baseline cohort in association with an identified exposure to a presumptive crusted scabies case with a higher prevalence of 14% amongst new entries to the cohort. At month 18, scabies prevalence fell to 2%. Scabies acquisitions six months after each MDA were 1% and 2% whilst treatment failures were 6% and 5% respectively.
Scabies prevalence reduced in the six months after each MDA with a low risk of acquisition (1–2%). However, in a setting where living conditions are conducive to high scabies transmissibility, exposure to presumptive crusted scabies and population mobility, a sustained reduction in prevalence was not achieved.
Australian New Zealand Clinical Trial Register (ACTRN—12609000654257).
| Scabies is endemic in many Australian Aboriginal and Torres Strait Islander communities, with 69% of infants infected in the first year of life. Previous mass drug administration (MDA) programs using topical acaricides to decrease scabies prevalence have had varying degrees of success in Australia. We were invited by one community in eastern Arnhem Land to develop and deliver an oral-ivermectin MDA. Utilizing a before and after study design, we measured scabies prevalence through population census with sequential MDAs at baseline and month 12. Scabies prevalence fell from 4% at baseline to 1% at month 6, rising to 9% at month 12 in association with an identified exposure to a presumptive crusted scabies case. For new entries to the cohort at month 12 scabies prevalence was higher at 14%. We were able to demonstrate a reduction in scabies prevalence in the six months after each MDA with a low risk of acquisition (1–2%); however, a sustained reduction was not achieved.
| Scabies mites infect up to 300 million people worldwide, most of whom are children living in poverty and overcrowded conditions.[1–3] In remote Australian Aboriginal communities, scabies has been near universal during the first year of life (69%).[4] Secondary infections with highly pathogenic bacterial pathogens Streptococcus pyogenes and Staphylococcus aureus contribute to high rates of pyoderma in these communities.[5–8] Acute post-streptococcal glomerulonephritis (APSGN) and streptococcal and staphylococcal sepsis,[9],[10] are recognised complications of pyoderma, whereas rheumatic fever, rheumatic heart disease and chronic renal failure are postulated sequelae that all occur in Australian Aboriginal people at the highest rates in the world.[11,12] In contrast, scabies is infrequently seen in non-Indigenous Australians.[2,8,13]
Individuals with scabies classically present with profuse pruritus involving only 5–15 mites per person, whereas an individual with crusted scabies, a rare condition, can have thousands of mites.[14,15] Well documented to occur in immune compromised hosts, most Aboriginal people identified with crusted scabies have no definable immune defect.[16] People with crusted scabies are highly infectious and have been identified as core transmitters in scabies epidemic cycles and institutional outbreaks.[3,16,17] Prior to 1996 and the introduction of ivermectin in Northern Territory (NT) Australia, there was a 5-year mortality rate of up to 50% for people with crusted scabies.[16]
Mass drug administration (MDA) programs using topical acaricides to decrease scabies prevalence have had varying degrees of success in Australia.[5,8,13] Due to high endemicity, high transmissibility of infestations, low treatment uptake and limited regional coverage, the presence of crusted scabies in communities and mobility of regional populations, a sustained reduction in prevalence has not been achieved to date in remote Aboriginal communities.[1,8,18] Having an established collaboration through the East Arnhem Healthy Skin Program [1,19,20] which demonstrated poor uptake of topical acaricides in household contacts,[1] we were invited by one community in eastern Arnhem Land to develop a proposal for an oral-ivermectin MDA targeting both scabies and strongyloidiasis. Strongyloidiasis is an infection with the intestinal nematode parasite, Strongyloides stercoralis, for which ivermectin is the first-line treatment.[21] Here we report the outcomes against scabies of the MDA program designed in collaboration with the participating community.
The setting was a remote island community, 550km from Darwin, Australia with an estimated population of 2121.[22] Most residents lived in the main community; 200–400 lived in one of 10 associated homelands outside the community (five of which were accessible only by air/water).
In consultation with the community, we designed a staged roll-out of two MDAs, implemented 12 months apart for the respective households/homelands. MDAs are typically designed to be implemented within a short time frame to maximise reduction of infective stages. However, our consultations with the community stressed the need for a more extended roll-out period to encompass house to house consultation, screening and treatment involving locally trained workers. There were 159 houses in the main community at the start of the project and 165 houses at the second MDA. The program was evaluated in a before and after study design.
We conducted population censuses in 2010 (baseline) and 2011 (month 12) to screen for scabies and strongyloidiasis that all residents were eligible to participate in. The MDA was delivered at the same time using an allocated drug regimen (Table 1). Two surveys were conducted six months after each MDA (month 6 and 18) to: a) follow-up participants who were positive for scabies and/or had an equivocal/positive Strongyloides result in the census six months prior, b) screen a computer-generated random sample of participants who were negative for both scabies and strongyloidiasis in the census six months prior and c) follow-up contacts of scabies acquisitions diagnosed at month 6 or 18. Given the staged program roll-out, subsequent visits to households were scheduled to accommodate the planned 6–12 month follow-up timeline as per the study protocol [23].
An allocated drug regimen for both scabies and strongyloidiasis was delivered based on weight and pregnancy status (Table 1). All non-pregnant participants who weighed ≥15 kg were administered a single dose of ivermectin 200 μg/kg at baseline and at month 12. Those ineligible for ivermectin received either topical 5% permethrin or 10% crotamiton. Treatment was repeated after 2–3 weeks if scabies and/or strongyloidiasis were diagnosed. All household contacts of participants diagnosed with scabies were either treated as part of the MDA or referred to the clinic. At the month 6 and 18 surveys, those diagnosed with scabies and their household contacts were provided with treatment and follow-up. Strongyloidiasis cases were treated but not their family contacts.
Residents were excluded from the MDA if they had an allergy to any components of the allocated drug regimen or had received the eligible study medication in the previous seven days. All female study participants aged 12–45 years had the option of a urinary test to determine pregnancy status as ivermectin safety in pregnancy has not been established.[24] Those not tested were allocated to the same treatment regimen as pregnant women. Pregnancy testing and medication administration was undertaken in portable work stations ensuring individual privacy. Adherence with the allocated drug regimen was monitored by direct observation of oral therapy and through verbal discussions with those applying topical acaricides.
Scabies was diagnosed clinically from observation of exposed skin. We classified scabies as: scabies-like lesions in a person who had either an itch, lesions in a typical location, or a household member with an itch. We accepted typical scabies lesions as being burrows, erythematous papules and macules, scales, vesicles, bullae, crusts, pustules, nodules and/or excoriations located in the finger web spaces, flexor surfaces of the wrists and elbows, axillae, head, feet, palms or buttocks in children or male genitalia and female breasts where assessed. Flipcharts [25] were used by Aboriginal Health Practitioners, Registered Nurses and ACWs to assist with the diagnosis of scabies and pyoderma. Participants who had a clinical diagnosis of suspected crusted scabies were referred to the local health service for laboratory confirmation and medical care according to locally developed guidelines which have been adopted internationally.[26]
Data were analysed using Stata 13 (StataCorp LP). Scabies prevalence at baseline and month 12 was calculated as a proportion of those seen who were diagnosed with scabies. At month 6 and 18 surveys, prevalence was determined as a weighted average of (i) treatment failure rate—the prevalence for participants seen with scabies at the survey who had scabies at the population census six months prior, and (ii) acquisition rate—the prevalence for participants seen at the survey who did not have scabies at the census six months prior. In determination of scabies acquisition, we also included those who were Strongyloides positive/equivocal but scabies negative in the denominator along with the computer generated randomly selected negatives from six months prior, as there was no relation between scabies and strongyloidiasis at baseline or month 12. Pyoderma prevalence was reported at baseline and month 12. Per protocol treatment was calculated as a proportion of those eligible for the drug regimen who were administered medication as outlined in Table 1.
Data entry was validated by double entering 15% of the records. The data entry error rate for variables used in the analysis was <5%. Data is available from the Dryad Digital Repository. [27]
The project was registered with the Australian New Zealand Clinical Trial Register (ACTRN—12609000654257)[23] and received ethical approval from Human Research Ethics Committee of the Northern Territory Department of Health and Menzies School of Health Research (EC00153—project 09/34).
Study recruitment was conducted by Aboriginal Community Workers (ACWs) who had completed a nationally accredited training program (Certificate II in Child Health Research 70131NT). The ACWs visited each house to discuss the project with family members and establish a household occupancy list. Ascertainment of written informed consent was obtained using a pictorial flipchart that incorporated a culturally-appropriate process to explain the project.[28] Parents or registered caregivers provided written consent for children aged <18 years and additional written assent was obtained from children aged 12-<18 years.
At baseline, there were 1256 residents on the household occupancy lists in the population census (March-September 2010), of which 1013 (81%) consented to participate. Most participants (n = 960, 95%) were seen over a four month period (April-July). The median number of participants per house was nine (IQR 4–13) from 127 (80%) houses visited. Non-participating households were mostly those occupied by non-Aboriginal residents working in the community. Seven of the 10 homelands consented to participate; one refused whilst residents from the other two homelands were seen in houses in the main community. A total of 1002 participants had data recorded on scabies at baseline, with scabies data missing for the remaining 11 participants (1%).
At month 12, there were 1163 residents on the household occupancy lists in the second population census (April-October 2011), of whom 1060 (91%) participated (~150 per month). There were 700 (66%) whom had also been seen at baseline and 360 (34%) new participants not previously seen. The median number of participants per house was 8 (IQR 3–12) from 133 (81%) houses visited. The median time per person between the baseline and month 12 census was 14 months (IQR 12–17 months). Most participants (96%) received the per protocol MDA regimen at baseline and 12, whilst 72% of those diagnosed with scabies received their second treatment as per protocol. No adverse events following administration of medications were reported.
Scabies prevalence among the baseline cohort was 4% and remained relatively stable during the initial assessment period (2%, 6%, 3% and 5% per month from April-July 2010 respectively when 91% of the baseline cohort were seen). At the month 6 survey, prevalence was 1% but increased to 9% at month 12 (5% absolute increase from baseline to month 12 for the baseline cohort) (Fig 1). At month 18, prevalence fell to 2%. The median age of participants with scabies was 11 years (IQR 6–38 years) with more females at baseline diagnosed with scabies than males (Table 2). Of the 42 participants diagnosed with scabies, 8/35 (23%) had infected scabies.
Prevalence among the baseline cohort had increased from 4% to 9% at month 12, whereas prevalence among new entries to the cohort (those seen for the first time at month 12) was 14% (Fig 1). In addition to the new cohort entries, the increased prevalence at month 12 was influenced by a cluster of cases epidemiologically linked to a participant diagnosed with presumptive crusted scabies. Prevalence within the baseline cohort of those who were known contacts rose from 7% (7/96) at baseline to 18% (17/96) at month 12, whereas prevalence amongst others who were not known contacts within the baseline cohort rose from 4% (23/598) at baseline to 8% (46/604) at month 12. Of the 113 participants diagnosed with scabies, 34/105 (32%) had infected scabies.
The presumptive crusted scabies case was identified in May 2011, a school age participant who had been receiving topical acaricide treatment from the school nurse every two weeks for the previous two months. With support from nine public health personnel who joined the study team, we identified 13 priority houses for follow-up, three of which were houses where the presumptive crusted scabies participant had been living over the previous four weeks, and 10 other households that had school contacts with scabies.
There were 184 people identified as residing in these 13 houses of whom 141 (77%) were seen; a median of 13 (IQR 10–18) participants per house (Fig 2). Of the 141 participants seen, 91 (65%) were from the baseline cohort of whom 16 (18%) had scabies at month 12. Of the 50 new participants seen for the first time in the priority houses, eight (16%) had scabies. Scabies prevalence within these 13 households collectively was 17% (n = 24). Almost all (98%) of those seen received ivermectin or 5% permethrin at the first visit.
On follow-up, 77/184 residents (42%) from the 13 priority houses were seen again at visit 2, median 37 days (IQR 23–42) after visit 1, with an acquisition rate of 4%. Of the 24 participants observed with scabies lesions at visit 1, 18 (75%) were re-treated at visit 2 (12 had lesions present when reviewed). Follow-up of these priority houses was completed within two months.
The increase in scabies prevalence at month 12 was most evident among children <15 years of age and was highest amongst new entries to the cohort (Fig 3). Pyoderma prevalence where the sores were described as purulent or crusted also increased amongst these age groups at month 12 (Fig 4).
Scabies treatment failures and acquisition were low throughout the study period (S1 and S2 Tables). The treatment failure rate was 6% (2/35) at month 6 and 5% (5/91) at month 18. The acquisition rate was 1% (4/352) at month 6 and 2% (6/276) at month 18. The median time between participant visits from baseline to month 6 was six months (IQR 5–7 months) and from month 12 to 18, eight months (IQR 7–10).
In our study, MDA incorporating ivermectin had a demonstrable but relatively short-term impact on scabies prevalence. In the six-months following each MDA, both the low overall prevalence (1–3%) and the low acquisition rates (1–2%) suggest that transmission was substantially reduced. However, the rapid rise in prevalence at month 12 highlights that an MDA program, where utilised, needs to be incorporated with a multi-faceted control program and ongoing surveillance in the community.
MDAs have been used to control and eliminate diseases for more than 25 years.[29] Ivermectin is one of the most commonly used drugs worldwide in the treatment of strongyloidiasis, lymphatic filariasis, and onchocerciasis.[30] It is increasingly being used to treat other parasitic infections including scabies,[31] pediculosis capitis [32] and malaria.[33] In 2014, Merck Sharp and Dhome updated the indications for ivermectin use to include treatment of crusted scabies and classical scabies if topical treatment is ineffective.[24,34].
Scabies is a neglected tropical disease,[35] ubiquitous in Australian Aboriginal and Torres Strait Islander communities, despite repeated MDAs with topical acaricides.[8],[3] Infestations are highly transmissible,[3] and as this study shows, prevalence escalates in the presence of high exposure (prevalence amongst known contacts of the presumptive crusted scabies case rose from 7% at baseline to 18% at month 12) and a high proportion of mobility (36% new entries to the cohort at month 12, of whom 14% had scabies). Others have shown the impact of exposure to crusted scabies [36] and overcrowded living conditions [1] on scabies prevalence. Outbreaks [37] and high scabies prevalence [38] have previously been linked with epidemics of APSGN, the sequelae of a post streptococcal infection that is common in developing countries and Indigenous populations.[10]
The participation rate among residents within the community was noteworthy (80–95%) encompassing an informed consent process implemented with and by the community. Under the guidance of elders and key community stakeholders, the development of a pictorial flipchart that incorporated a culturally-appropriate process to explain the project was fundamental in obtaining informed consent.[28] The flipchart incorporated a local story well known in the community which we had gained specific approval to use and translate into local language. That some members declined participation is testament to the culturally appropriate processes enabled within this study. Moreover, the process of ongoing engagement and the culturally acceptable arrangements regarding pregnancy testing, screening and steady (as opposed to rapid) roll-out of the program were integral to the reach achieved over the course of the study.
A previous attempt to implement an ivermectin MDA for scabies control in Queensland Aboriginal communities in the early 1990s did not proceed due to administrative concerns about medication safety and informed consent.[39] Instead the team conducted a MDA with ivermectin in the Solomon Islands and showed ivermectin to be safe and effective with low scabies prevalence persisting for at least 32 months.[31] The longer-term duration of benefit however, is unclear as there was no ongoing active surveillance. In Fiji, no significant difference was found between MDAs with either ivermectin or benzyl benzoate after 24–28 days.[40]
To date, the use of ivermectin to treat scabies has not been associated with any serious adverse effects nor were any observed in our study. However, it is recommended that ivermectin not be administered to pregnant women or children who are younger than five years of age or in those who weigh less than 15 kg. This recommendation is due to theoretical concerns regarding potential neurotoxicity and a lack of safety data. Although there have been no reports of foetal problems when ivermectin has been administered in pregnancy to thousands of women, caution is still recommended.[41] While the safety of ivermectin at the extremes of age remains to be conclusively established, there is increasing evidence suggesting that the use of ivermectin in children <5 years is safe.[26]
The high proportion of new entries to the cohort at the month 12 census (36%) coincided with a large funeral that was attended by visitors from other communities who were camping in tents in the house yards of relatives. At this time, many local residents were also displaced from their homes into tents or other people’s homes as their houses were being refurbished or demolished and rebuilt, as part of a government initiative to address housing shortages in Aboriginal communities.[42] This change in population dynamics is considered highly mobile by Australian mainstream standards, but does not reflect the stability reflected by the customary attachment of Aboriginal people to their home community and the regional area.[43]
The increased scabies prevalence at month 12 was notable in the 0–14 year age group and in particular for those new participants to the cohort. Young children are particularly susceptible to scabies infestations [2,4,44] and, as shown in this study, are more likely than adults to show a change in prevalence. For population surveillance of scabies it has previously been recommended that this is best achieved by monitoring the prevalence in young children,[4] a recommendation that is further supported by this study.
At baseline there was concern about inter-observer variation in the diagnosis of scabies as more females (n = 33) than males (n = 9) had been diagnosed with scabies. These concerns were dispelled after reviewing the names of the researchers screening the children (for whom the majority of scabies were diagnosed) and found that the female researchers, who at that time had more experience in diagnosing scabies than the male researchers, had been conducting most of the skin checks for male and female children. Thereafter we conducted regular reviews of screening processes in the field and from photographs taken, to improve consistency in diagnosis and reduce inter observer variation.
It was also apparent to our community-based research team, that the relationship built over the course of the team’s work meant that by month 12 it was relatively commonplace for households to seek out the research team to assist in making their homes scabies free, and to send family members who had not been present on the day the family were seen to the research office for screening and treatment. We acknowledge that this may have introduced a screening bias in the latter part of the study but the increased scabies prevalence at month 12 amongst those who had been seen at baseline indicates that the increase in prevalence was not an artefact of care-seeking behaviour.
The rise in scabies prevalence at month 12 coincided with: a cluster of cases epidemiologically-linked to an individual with presumptive crusted scabies, a high prevalence amongst new entries to the cohort (an indicator of the impact of high population mobility), and an increased prevalence amongst members of the baseline cohort who did not have a known exposure to the suspected crusted scabies case (4% to 8%). This demonstrated how readily scabies prevalence can increase. Control measures were able to be implemented promptly as the research team had commenced the second house to house population census and MDA and were able to coordinate the response with the local PHC services and community. Of note, was an outbreak of APSGN [45] occurring at the same time in another large NT community that public health personnel were responding to. Scabies prevalence in this community for children aged 1–17 years was 3% (n = 8) and 40.5% (n = 219) for purulent or crusted sores. Personal communication from the public health unit revealed there had been three cases of ARF and no cases of APSGN reported in the region in the four months following the outbreak.
Our study provides evidence that ivermectin based MDAs can have a role in reducing scabies prevalence but also highlights that maintaining a reduction requires ongoing surveillance,[4] diagnosis and chronic case management of individuals with crusted scabies,[18,34] and ongoing engagement with community members that has a particular focus on households and close contacts.[1] Due to the customary movements of Aboriginal people, regional approaches to decrease re-introduction of scabies from neighbouring communities needs to be considered.
|
10.1371/journal.pntd.0004875 | Spatiotemporal Dynamics of Scrub Typhus Transmission in Mainland China, 2006-2014 | Scrub typhus is endemic in the Asia-Pacific region including China, and the number of reported cases has increased dramatically in the past decade. However, the spatial-temporal dynamics and the potential risk factors in transmission of scrub typhus in mainland China have yet to be characterized.
This study aims to explore the spatiotemporal dynamics of reported scrub typhus cases in mainland China between January 2006 and December 2014, to detect the location of high risk spatiotemporal clusters of scrub typhus cases, and identify the potential risk factors affecting the re-emergence of the disease.
Monthly cases of scrub typhus reported at the county level between 2006 and 2014 were obtained from the Chinese Center for Diseases Control and Prevention. Time-series analyses, spatiotemporal cluster analyses, and spatial scan statistics were used to explore the characteristics of the scrub typhus incidence. To explore the association between scrub typhus incidence and environmental variables panel Poisson regression analysis was conducted.
During the time period between 2006 and 2014 a total of 54,558 scrub typhus cases were reported in mainland China, which grew exponentially. The majority of cases were reported each year between July and November, with peak incidence during October every year. The spatiotemporal dynamics of scrub typhus varied over the study period with high-risk clusters identified in southwest, southern, and middle-eastern part of China. Scrub typhus incidence was positively correlated with the percentage of shrub and meteorological variables including temperature and precipitation.
The results of this study demonstrate areas in China that could be targeted with public health interventions to mitigate the growing threat of scrub typhus in the country.
| Scrub typhus is a vector-borne disease carried by the chigger mite and is endemic in the Asia-Pacific region. Now scrub typhus causes a considerable burden on public health and the economy in China. We explored the spatiotemporal dynamics of scrub typhus cases in China between January 2006 and December 2014, and explored the potential risk factors affecting the spatial distribution of the disease. The majority of cases were reported between July and November, with peak incidence during October every year. Several high-risk clusters were identified in southwest, southern, and middle-east China. Scrub typhus incidence was positively correlated with the percentage of shrub, and temporal variation in temperature and precipitation in China.
| Scrub typhus, also known as tsutsugamushi disease, is endemic in the so-called “tsutsugamushi triangle” area that includes Pakistan and Afghanistan in the west, far-eastern Russia and Japan in the north, and northern Australia in the south [1]. The causative bacterium of this disease, Orientia tsutsugamushi (O. tsutsugamushi), is spread to humans bitten by infected species of trombiculid mites [2, 3]. The clinical presentation of scrub typhus is characterized by high fever and rash or typical eschar at the location of the bite, which can progress to multiple organ failure and even death in some cases [4–6]. It is estimated that over one billion people are currently living in at-risk areas and approximately one million cases occur around the world annually [2, 7]. In recent years, there has been a drastic increase in both the frequency and geographic distribution of scrub typhus cases, which could signal the re-emergence of this neglected tropical disease [8–11].
The first reported case of a human infected with scrub typhus in China was identified in the southern province of Guangdong in 1948 [12]. Until the 1980s, scrub typhus cases primarily occurred in the regions south of Yangtze River with established natural foci including Zhejiang in the east and Yunnan in the west part of China [13, 14]. However, with rapid societal development, changing environment, climate change, population movement, better recognition by health care professionals and ever-improving detection techniques, both sporadic cases and disease outbreaks began to be identified in the northern provinces of Shandong, Jiangsu, Tianjin and Beijing, as well as the emergence of new natural foci in the past three decades [13, 15–17]. Currently, the disease is widespread in most of the provinces in mainland China, where the incidence has increased rapidly in recent years. Despite the recent resurgence of illness, scrub typhus remains a neglected tropical disease that is able to simultaneously impact tourism and military activities in China [9, 18], with the potential to cause a significant burden on both public health and economy.
In the past decades, spatiotemporal analysis techniques have been widely applied in the surveillance of infectious disease and outbreak investigations [19–21]. Previous studies have identified clusters of reported cases of scrub typhus in different provinces of China and reported that the geographic distribution of the disease varied by year [22, 23]. Some studies have revealed that environmental variables were important drivers in the transmission of scrub typhus [7, 24]. However, there are few studies that examine both the spatiotemporal dynamics and potential risk factors in scrub typhus transmission across China. Thus, the objectives of this study were to describe the temporal trends in scrub typhus incidence, to detect spatiotemporal clusters of scrub typhus cases at the county level, and to identify the physical environmental variables associated with scrub typhus incidence, which would be helpful for the health administration officers and public health workers to the implementation of effective intervention measures targeted toward high-risk areas and populations.
This study was approved by the Ethics Committee of Beijing Institute of Disease Control and Prevention. All the data analyzed in this study were de-identified to protect patient confidentiality.
In China, scrub typhus is a vector-borne notifiable disease; attending physicians are required by law to report to the China Center for Disease Control and Prevention through the China Information System for Disease Control and Prevention (CISDCP). Scrub typhus case reports include basic demographic and clinical data including gender, age, occupation, residential address, date of onset of symptoms, laboratory diagnosis, and clinical outcome for each case. Data from January 2006 through December 2014 were obtained from CISDCP. All scrub typhus cases were confirmed according to the diagnostic criteria issued by the Ministry of Health of the People’s Republic of China. The criteria for a confirmed case of scrub typhus include epidemiological exposure histories (travelling to an endemic area and contact with chiggers or rodents within 3 weeks before the onset of illness), clinical manifestations (such as high fever, lymphadenopathy, skin rash and eschar or ulcers), and also at least one of the laboratory diagnosis: a 4-fold or greater rise in serum IgG antibody titers between acute and convalescent sera by using indirect immunofluorescence antibody assay (IFA), or detection of O. tsutsugamushi by polymerase chain reaction (PCR) in clinical specimens, or isolation of O. tsutsugamushi from clinical specimens [24–26].
Demographic data for each county was obtained from the National Bureau of Statistics of China. Environmental and meteorological data from 2006 to 2014 were collected. Land cover variables such as the percentage coverage of cropland, forest, shrub, grassland, built-up land and water bodies were collected from the data on land cover 2005 and 2009 released by European space agency (http://www.esa.int). Meteorological variables including temperature, relative humidity and precipitation were obtained from the Chinese Academy of Meteorological Sciences (www.cams.cma.gov.cn).
In order to perform spatial analysis, the data set of cases was aggregated at the county level as the spatial unit for analysis. In mainland China, there are 31 provinces (or municipalities) comprised of 2,922 counties, with population sizes ranging from 7,123 to 5,044,430, with geographic areas ranging in size from 5.4 to 197,346 square kilometres. All cases were geocoded and matched to the county-level administrative boundaries using the ArcGIS software (version 9.3, ESRI, Redlands, CA).
The cases of scrub typhus reported at the county-level were aggregated to provide a national data set of monthly cases for time-series analyses. The monthly incidence as well as the cumulative number of cases was tabulated for visualization, along with the graphical assessment of the cumulative annual cases with various trends including linear, polynomial, and exponential growth curves using Excel (Microsoft, Redmond, WA). For the assessment of the seasonal trend in scrub typhus incidence, both the annual and long-term trends were assessed. The average monthly incidence for every calendar month (January to December) was compared to the average incidence in January (the lowest monthly incidence and beginning of each year) using a categorical Poisson regression model to generate incident rate ratios (IRR) and 95% confidence intervals (CI) for the IRR. Temporal autocorrelation between the monthly reported cases of scrub typhus and seasonal trend in incidence was assessed using time lags between 0 and 60 months. Seasonal trends were classified by maximum autocorrelations of 12 months and minimum autocorrelations observed every 6 months, that also demonstrated a sinusoidal oscillation with respect to time.
Local Indicators of Spatial Association (LISA) were used to assess the spatial pattern of scrub typhus incidence at the county level during the study period. LISA was used to identify significant hotspots (High-High), coldspots (Low-Low), and outliers (High-Low and Low-High) by calculating local Moran’s I index between a given county and the neighbouring values in the surrounding counties [27]. The significance level of clusters was determined using a Z score generated by comparison of the Local Moran’s I statistic for the average incidence in each county. A high positive Z score indicated that the surroundings had spatial clusters (High-High: high-value spatial clusters or Low-Low: low-value spatial clusters) and a low negative Z score indicated the presence of spatial outliers (High-Low: high values surrounded with low values or Low-High: Low values surrounded with high values) [27]. Kulldorff’s space-time scan statistic (SaTScan software, version 9.1.1) was used to explore the location of high-risk space-time clusters. The space-time scan statistic was defined by a cylindrical window with a circular (or elliptic) geographic base and with height corresponding to time [28]. The base was defined exactly as for the purely spatial scan statistic, while the height reflected the time period of the potential clusters [28]. In this study, circular scan windows were selected and fit discrete Poisson models. The maximum spatial cluster size was set to 5% of the population at risk in the spatial window and a maximum temporal cluster size of 10% of the study period in the temporal window. Likelihood ratio tests were evaluated to determine the significance of identified clusters and P-values were obtained through Monte Carlo simulation after 999 replications. The null hypothesis of a spatiotemporally random distribution was rejected when the P-value was < 0.05.
We conducted panel Poisson regression analysis to examine the association between yearly scrub typhus incidence and potential environment risk factors. An autocorrelation term was included to account for spatial and temporal dependency in scrub typhus incidence. The autocorrelation term was calculated using the minimum distance from each county-center to the nearest cluster. After aggregating the yearly incidence into a panel dataset, the association between incidence and environmental factors was examined, with IRR and their corresponding 95% confidence intervals and p values estimated using maximum likelihood methods. The temporal analysis and the panel Poisson regression analysis were conducted in STATA software (Stata Crop Lp, College Station, TX, USA).
A total of 54,558 scrub typhus cases were reported from 1,031 counties during the period between 2006 and 2014. The monthly variations in the number of scrub typhus cases presented in Fig 1A suggested a seasonal relationship, whereas the rapid increase in the total number of annual cases in Fig 1B was explained by an exponential growth function (R2 = 0.98). Scrub typhus cases occurred throughout the year, however began to increase dramatically in April through September, before reaching a peak in October and returning to low, constant levels of transmission in the winter months of December through March. The average numbers of reported cases are presented by month in Fig 2A, along with the incidence rate ratios comparing each month to January of each year. It was estimated that the month of October, which consistently had the largest number of reported cases, had an incidence of scrub typhus that was approximately 25 times higher than the winter months of January, February, or March (P <0.001); IRR = 25.2; 95% CI: (13.1, 27.4). Given the seasonal patterns observed in Fig 1A, the autocorrelation of the monthly scrub typhus cases was compared using a time-series analysis. The pattern of autocorrelation in Fig 2B not only demonstrated that the monthly reported incidence in one month was significantly correlated with the previous months’ incidence, but also exhibited maximum correlations in every 12 months, minimum correlations in every 6 months, and followed a sinusoidal pattern of oscillation with an increasing correlation between 2006 and 2014. Though aggregated for the temporal analyses, the annual incidence rate of scrub typhus was highly variable at the county level, which ranged from zero reported cases to 66.21 cases per 100,000 residents.
Spatial analysis of scrub typhus incidence at the county level demonstrated the spatial autocorrelation was positive, indicating clustering of reported cases during the study period. The values of Moran’s I ranged from 0.02 to 0.08 with all P values < 0.05 (Table 1), indicating the presence of clusters of scrub typhus incidence in each year. The hotspots (High-High) and outliers of scrub typhus transmission in mainland China were identified through LISA analysis. Hotspots were primarily distributed in the southern and southwestern provinces of China, however variation of the location was observed. In 2006, the hotspots were distributed sporadically in Yunnan, Guangdong and Fujian. Later hotspots in those provinces expanded to include larger geographic areas over the next eight years (Fig 3). Hotspots also occurred and expanded in Guangxi and Hainan province between 2008 and 2014, with a short appearance in the northern provinces of Anhui and Jiangsu in 2011. High-Low outliers were sporadically distributed in the middle-eastern provinces of China including Shandong, Jiangsu and Anhui, while Low-High outliers were mainly concentrated in Yunnan province (Fig 3). During the study period, the proportion of counties and populations within High-High clusters increased from 1.51% to 6.19% and 1.44% to 5.95% respectively. Additionally, hotspot counties were responsible for between 48.96% of all reported cases in 2006 to 67.57% in 2013 (Table 2).
Fig 4 shows the distribution of annual average scrub typhus incidence and the location of spatial clusters identified by using Kulldorff’s space-time scan statistic for each year from 2006 to 2014. As can be seen, both the number of counties with increased scrub typhus incidence expanded persistently between 2006 and 2014, which resulted in the formation of a large, continuous geographic area of scrub typhus incidence in southern mainland China. The primary cluster of scrub typhus cases was originally located in Shandong and Jiangsu province, after which the area expanded between 2006 and 2008. Since that time, the primary cluster of increased scrub typhus incidence shifted to southwest, except for 2011, where the primary cluster was identified in the northwestern region that included Anhui province. Secondary clusters of scrub typhus cases were also identified in southern and southeastern China as well as in Shaanxi and Beijing, with five to eight clusters identified every year.
Additionally, spatiotemporal clusters across the entire study period between 2006 and 2014 were identified by using Kulldorff’s spatiotemporal scan statistic (Fig 5). The primary cluster was located in southwest China, including 103 counties in Yunnan, 11 counties in Sichuan, and even a county in Tibet, with a radius of 491.64 km. The time frame of the primary cluster was from July to October in 2014, which coincided with the largest annual scrub typhus outbreak identified by the time-series analyses and carried a RR of 64.88 and log likelihood ratio (LLR) of 10,460 (Table 3). Most importantly, the primary cluster accounted for only 2.63% of the total population, but included 29.28% of the total cases during that time (Table 4). In addition, there were eleven significant secondary clusters identified, also primarily located in southern and middle-eastern China, with the RR and LLR ranging from 2.94 to 800.71 and 32 to 7433, respectively (Table 3).
The results of panel Poisson regression analysis revealed that scrub typhus incidence was positively correlated with the percentage of forest (IRR = 1.17; 95% CI: 1.15, 1.10) and shrub (IRR = 7.52; 95% CI: 7.18, 7.87) as well as temperature (IRR = 1.06; 95% CI: 1.05, 1.07) and precipitation (IRR = 1.01; 95% CI: 1.00, 1.01). Our results also indicate that lower incidence of scrub typhus is associated with the percentage of cropland (IRR = 0.61; 95% CI: 0.60, 0.62), grassland (IRR = 0.23; 95% CI: 0.19, 0.28), built-up land (IRR = 0.64; 95% CI: 0.61, 0.67), water bodies (IRR = 0.25; 95% CI: 0.23, 0.27), and relative humidity (IRR = 0.90; 95% CI: 0.90, 0.91). Except for percentage of forest, which was positively correlated in the univariate analysis but negatively correlated in the multivariate analysis (Table 5), the IRR were similar, with the largest differences observed with the percentage of shrub (IRR: 7.52 vs 1.29) and the IRR for temperature (IRR: 1.06 vs 1.35) in the multivariate analysis.
The results of our study indicate that the spatiotemporal transmission of scrub typhus has increased exponentially between 2006 and 2014 and spread throughout much of mainland China. LISA and spatial scan statistics analyses identified significant clusters with respect to both space and time that indicated outbreaks of scrub typhus were primarily located in southwestern and southern China. Given the ability of spatiotemporal analyses based on geographic information systems to assist in the identification of counties with the highest risk of contracting scrub typhus, we suggest that these methods could have further application in both future disease surveillance and planning of mitigation strategies.
In a previous study, we identified the most significant cluster of scrub typhus in the southeastern provinces of Guangdong, Fujian, Jiangxi, and Guangxi [25]. In the current study, the counties at the highest risk were located in the southwestern provinces of Yunnan and Sichuan and accounted for nearly a quarter of total cases during the 2014 outbreak. More importantly, by analyzing annual spatiotemporal clustering, the transmission dynamics appeared to shift between middle-east, southeast, and southwest China. Therefore, we suggest that each of these three high-risk regions be considered for the implementation of targeted interventions such as environmental management, controlling and killing rodent and mites, strengthening personal protection. Given that there was a low incidence of scrub typhus in 2007 in north eastern Xinjiang Uygur Autonomous Region contiguous to Gansu province, we suggest that more investigations be performed to determine if novel cases are as yet unreported in those outlying provinces. Notably, high-high spots detected by LISA analysis were primarily concentrated in southern China and rarely identified in the middle-eastern regions. The high-low outliers that were identified in this region, suggest that scrub typhus incidence in this region was concentrated in a few counties, which could indicate that natural foci of scrub typhus are still forming in this region as they expand into northern China. In addition, our study identified clusters in provinces such as Beijing, Shaanxi, and Anhui, which further confirmed disease outbreaks in these provinces reported by other studies [14, 29, 30]. Thus, our findings will further assist health authorities and public health practitioners through the identification of established foci in southern China as well as the documentation of the emergence of new foci in the north.
The increasing number of reported cases and geographic expansion of scrub typhus in China is partly due to increasing quality of the surveillance system and availability of detection facilities as the increasing investment of health resources. Moreover, environment change and human activities could be important factors contributed to this increasing trend [31, 32]. In this study, our findings demonstrated the percentage of shrub, temperature and precipitation were risk factors associated to the spatiotemporal heterogeneity of scrub typhus notifications in China. A possible explanation is that temperature, precipitation, and shrub may affect the population dynamics and activity levels of chigger mites [7, 33]. Previous studies also suggested the migration of infested rodents or chiggers may have led to the formation new natural foci in provinces of Shandong, Henan, and Beijing since the meteorological and vegetation cover conditions are similar in these areas [34]. Additionally, socio-economic factors could have also served as important drivers for the transmission of scrub typhus in recent years. For instance, the urbanization and change of land use may contribute to the spread of scrub typhus into urban areas by providing suitable habitats such as clearings, grasslands, and riverbanks for vectors and small rodents [35]. Presently, we only explored the association between environmental variables (land cover, weather) and the incidence of scrub typhus. In future, a more well-coordinated and interdisciplinary approach is imperative and urgently needed to explore the relative effects of environmental and socio-economic factors on scrub typhus transmission in mainland China.
While this study brings the important new knowledge on the epidemiology of scrub typhus in China, there are also some limitations. Since the case data were obtained from a passive surveillance system, the reporting system might miss some cases due to lack of diagnostic facilities and/or misdiagnosed due to the co-occurrence of other febrile diseases such as leptospirosis, typhoid fever, or hemorrhagic fever in the absence of the characteristic eschar [36, 37]. Additionally, in our analysis we chose to use a circular scan window in space-time scan statistics. While the circular scan has been documented to perform better at detecting larger clusters compared to the elliptic window scan, it may also include insignificant zones and has been shown to have reduced performance when used with irregular shapes [38, 39].
In conclusion, our results show that the incidences of scrub typhus vary in different spatial settings, and the geographic distribution of scrub typhus appeared to have expanded over recent years, indicating the disease is emerging or re-emerging and remains an important public health problem in China. Meanwhile, the study also prove environmental factors such as temperature, precipitation and vegetation type are important drivers in the dynamics of scrub typhus. To the best of our knowledge, this is the most detailed study on spatiotemporal epidemiology of scrub typhus across the entire country, which provides a sound evidence base for future prevention and control programs and also lays a foundation for further investigation into the social and environmental factors responsible for changing disease patterns. Given the exponential growth and spatiotemporal features observed in this study, it is likely that the incidence of scrub typhus will increase in the future, and the disease may be spreading even to non-traditional foci where cases had rarely been reported. Moreover, based on the results of this study, it is recommended that immediate measures be taken in high-risk areas to increase health education and awareness of scrub typhus, enhance the availability of diagnostic and treatment practices, as well as continue surveillance of this emerging infectious disease.
|
10.1371/journal.ppat.1002539 | Novel Transmembrane Receptor Involved in Phagosome Transport of Lysozymes and β-Hexosaminidase in the Enteric Protozoan Entamoeba histolytica | Lysozymes and hexosaminidases are ubiquitous hydrolases in bacteria and eukaryotes. In phagocytic lower eukaryotes and professional phagocytes from higher eukaryotes, they are involved in the degradation of ingested bacteria in phagosomes. In Entamoeba histolytica, which is the intestinal protozoan parasite that causes amoebiasis, phagocytosis plays a pivotal role in the nutrient acquisition and the evasion from the host defense systems. While the content of phagosomes and biochemical and physiological roles of the major phagosomal proteins have been established in E. histolytica, the mechanisms of trafficking of these phagosomal proteins, in general, remain largely unknown. In this study, we identified and characterized for the first time the putative receptor/carrier involved in the transport of the above-mentioned hydrolases to phagosomes. We have shown that the receptor, designated as cysteine protease binding protein family 8 (CPBF8), is localized in lysosomes and mediates transport of lysozymes and β-hexosaminidase α-subunit to phagosomes when the amoeba ingests mammalian cells or Gram-positive bacillus Clostridium perfringens. We have also shown that the binding of CPBF8 to the cargos is mediated by the serine-rich domain, more specifically three serine residues of the domain, which likely contains trifluoroacetic acid-sensitive O-phosphodiester-linked glycan modifications, of CPBF8. We further showed that the repression of CPBF8 by gene silencing reduced the lysozyme and β-hexosaminidase activity in phagosomes and delayed the degradation of C. perfringens. Repression of CPBF8 also resulted in decrease in the cytopathy against the mammalian cells, suggesting that CPBF8 may also be involved in, besides the degradation of ingested bacteria, the pathogenesis against the mammalian hosts. This work represents the first case of the identification of a transport receptor of hydrolytic enzymes responsible for the degradation of microorganisms in phagosomes.
| Phagocytosis is the cellular process of engulfing solid particles to form an internal phagosome in protozoa, algae, and professional phagocytes of multicellular eukaryotic organisms. In phagocytic protozoa, phagocytosis is involved in the acquisition of nutrients, and the evasion from the host immune system and inflammation. While hydrolytic enzymes that are essential for the efficient and regulated degradation of phagocytosed particles, such as bacteria, fungi, and eukaryotic organisms, have been characterized, the mechanisms of the transport of these proteins are poorly understood. In the present study, we have demonstrated, for the first time, the molecular mechanisms of how the digestive enzymes are transported to phagosomes. Understanding of such mechanisms of the transport of phagosomal proteins at the molecular level may lead to the identification of a novel target for the development of new preventive measures against parasitic infections caused by phagocytic protozoa.
| Lysozymes (EC 3.2.1.17) are the antibacterial protein that has an ability to damage the cell wall of bacteria [1]. The enzyme acts by catalyzing the hydrolysis of 1,4-beta-linkages between N-acetylmuramic acid and N-acetyl-D-glucosamine in peptidoglycans and between the N-acetyl-D-glucosamine residues in chitodextrins. While biochemical [2], functional [3], and structural [4] features of lysozymes have been well established, the mechanisms for intracellular trafficking and secretion remain poorly characterized except for the report that showed that condroitin sulfate is involved in lysosomal targeting of lysozymes [5]. Hexosaminidase (EC 3.2.1.52) is involved in the hydrolysis of terminal N-acetyl-D-hexosamine residues in hexosaminides. Three dimeric isozymes of β-hexosaminidase are formed by the combination of α and β subunits, encoded by HEXA and HEXB genes, respectively. β-Hexosaminidase and the cofactor GM2 activator protein catalyze the degradation of the GM2 gangliosides containing terminal N-acetyl hexosamines [6]. Mutations in HEXA gene decrease the hydrolysis of GM2 gangliosides, which is the main cause of Tay-Sachs disease, whereas mutations in HEXB gene results in Sandhoff disease [7]. The trafficking mechanism of β-hexosaminidase via mannose-6-phosphate receptor has been well studied in mouse lymphoma and myeloma cell [8]–[10]. However, the mechanisms of trafficking of β-hexosaminidase in eukaryotes besides mammals remain to be discovered.
Lysozyme and β-hexosaminidase are abundant components found in phagosomes from Entamoeba histolytica [11], [12], which is the anaerobic or microaerophilic protozoan parasite, causing amebic dysentery and amebic liver abscesses in an estimated 10 million cases annually [13]. However, the role and intracellular trafficking of these enzymes remain unknown. Phagocytosis and phagosome biogenesis seems to play a pivotal role in pathogenesis in E. histolytica [14]. E. histolytica is capable of internalizing extracellular particles by phagocytosis. The amebic trophozoites ingest microorganisms in the large intestine [15], [16], and host cells including non-immune cells [17], and immune cells [18] during tissue invasion. It has been well-established that in vitro and in vivo virulence correlates well with the ability of phagocytosis [14], [19], [20]. Furthermore, phagosomes contain a panel of proteins that were shown to be crucial in pathogenesis such as cysteine proteases (CPs) [21], amoeba pores [22], and galactose/N-acetylgalactosamine-specific lectin [23], [24], proteins involved in cytoskeletal reorganization [25], [26], vesicular trafficking [27]–[29], and signal transduction [30], [31]. Therefore, understanding the molecular mechanisms of phagocytosis and phagosome biogenesis as well as the role and trafficking of individual phagosomal proteins in phagosomes, should help to understand underlying links between phagocytosis and pathogenicity.
Recently, the proteins and mechanisms involved in phagocytosis have been demonstrated. For instance, the surface Ca2+-binding kinase (C2PK) has shown to be involved in the initiation of phagocytosis [31]. The antisense inhibition of C2PK caused inhibition of the initiation of erythrophagocytosis. It has also been shown that surface transmembrane kinase (TMK96) and p21-activated kinase (PAK) play an important role in phagocytosis of human erythrocytes [32], [33]. The unconventional myosin, myosin IB, was shown to be involved in cytoskeleton rearrangement during phagocytosis [25], [26]. Furthermore, phosphatidylinositides also play critical roles during phagocytosis [34], [35]. Our previous proteomic studies, where 159 proteins were identified from purified phagosomes [11], [12], also suggested a direct link between phagosome biogenesis and pathogenesis, as phagosomes contained a panel of proteins that were shown to be crucial in pathogenesis described above. Furthermore, the proteins that are implicated for degradation of phagocytosed bacteria, e.g. amoebapores [22], lysozymes, and β-hexosaminidase, as well as other hydrolytic enzymes such as amylase and ribonuclease were also demonstrated in phagosomes. While both the constituents of phagosomes and the kinetics of their recruitment are known, very little is known on how these proteins are transported to phagosomes. Recently, we discovered a putative transmembrane receptor for cysteine proteases from E. histolytica, which preferentially binds to CP5 (Nakada-Tsukui K, et al., unpublished data), which is directly implicated in the pathogenesis [36]–[38]. The E. histolytica genome contained a total of 11 members showing significant mutual identity and structural conservation to the transmembrane cysteine protease receptor: the signal peptide at the amino terminus, a single transmembrane domain close to the carboxyl terminus, and the YxxΦ motif at the carboxyl terminus. This family of proteins was designated as cysteine protease binding family proteins 1–11 (CPBF1-11). In the present study, we characterized one of the most highly expressed CPBF genes among the family, CPBF8. We showed that CPBF8 localizes to phagosomes during phagocytosis, while it is distributed to the acidic compartment in steady state. Affinity immunoprecipitation followed by LC-MS/MS analysis showed that CPBF8 specifically bound to lysozymes and β-hexosaminidase α-subunit. Repression of CPBF8 by gene silencing reduced lysozyme and β-hexosaminidase activities in phagosomes, and caused a defect of digestion of ingested bacteria.
We examined the localization of CPBF8 during phagocytosis of CHO cells. Trophozoites of CPBF8-HA-expressing strain were incubated with CellTracker-loaded CHO cells for 10 to 60 min to allow ingestion of CHO cells. Immunofluorescence assay using anti-HA antibody showed that CPBF8 was localized to phagosomes containing CHO cells at all time points (10, 30, and 60 mins) (Figure 1A). CPBF8 remained associated with phagosomes in the course of phagocytosis: the percentage of colocalization did not significantly changed (84, 92 and 82% at 10, 30, and 60 min, respectively). Immunofluorescence image of the amoeba undergoing engulfment revealed that CPBF8 localized to the basolateral portion of a phagosome, and excluded from the tunnel-like structure connecting a phagosome and the CHO cell being aspirated [35].
As immunofluorescence assay showed that CPBF8 was also distributed to a large number of vesicles and vacuoles under quiescent (i.e., non-phagocytic) conditions, we examined the nature of these compartments. CPBF8-HA was associated with the acidic organelles labeled with membrane-diffusible LysoTracker under steady-state conditions (60% of LysoTracker-positive vesicles/vacuoles was positive for CPBF8) (Figure 1B). CPBF8 colocalized nicely with a vacuolar membrane protein, pyridine nucleotide transhydrogenase, EhPNT, which converts NADPH and NADH using the proton gradient across the membrane [39] (Figure 1C). It has been shown that EhPNT is localized to the acidic compartment in steady state and transported to phagosomes upon phagocytosis [40].
To identify potential cargo proteins that CPBF8 binds and carries to phagosomes, we immunoprecipitated proteins that bind to CPBF8, from the lysates of the transformant where HA-tagged CPBF8 was ectopically expressed (Figure 2). Silver stained SDS-PAGE gel revealed three major bands of about >120, 60, and 20 kDa (bands C, E, and F) and three minor bands of about >300, >200, and 75 kDa (bands A, B, and D) exclusively found in the immunoprecipitated sample from CPBF8-HA strain, but not from HA control strain. These bands were excised and subjected to LC-MS/MS analysis (Table 1 and Table S1). Smeary band C, which showed an apparent molecular mass of ∼130 kDa on SDS-PAGE was identified as CPBF8 itself; the apparent size was larger than the predicted size (99.3 kDa), suggestive of post-translational modifications or aberrant structure (see below). Band E was identified as β-hexosaminidase α-subunit (XP_657529; EHI_148130) with 19.7% coverage. Band F was identified as a mixture of lysozyme 1 (XP_653294, EHI_199110) and lysozyme 2 (XP_656933; EHI_096570) with 22.7 and 30.2% coverage, respectively. Lysozyme 1 and 2 were previously demonstrated by our previous phagosome proteome analysis [11], [12]. Bands A, B, and D mostly corresponded to CPBF8 (Table S1). These data clearly indicate that β-hexosaminidase α-subunit and lysozymes are predominant proteins that bind CPBF8. β-hexosaminidase α-subunit was not previously detected by phagosome proteomics, whereas its β-subunit was detected.
To further demonstrate the role of CPBF8, we created a strain in which CPBF8 expession was repressed by long term transcriptional gene silencing [41] (“CPBF8gs strain”). Gene silencing is mediated by nuclear localized antisense small RNAs with 5′-polyphosphate termini [42], and observed only in G3 and its derived strains, in which amoebapore genes have been repressed. RT-PCR analysis showed that the mRNA level of CPBF8 gene in CPBF8gs strain was specifically reduced to the undetectable level (Figure 3A). DNA microarray analysis further verified that CPBF8 transcript was reduced by 326 fold, while the expression of other CPBF genes remained unchanged (Figure 3B). In in vitro cultivation CPBF8gs strain did not show any defect in growth compared to control pSAP2-Gunma-transfected strain (Supplemental information Figure S1). The doubling times of control and CPBF8gs strains were comparable (20.9 and 20.6 h, respectively). Thus, the defects in protein transport and the decrease in cytopathy against mammalian cells and bacteria digestion, described below, are not likely attributable to poor proliferation (growth) of CPBF8gs strain.
We examined β-hexosaminidase and lysozyme activities in CPBF8gs and control strains using a synthetic N-acetylglucosamine-related substrate (4-methylumbelliferyl-2-acetamido-2-deoxy-β-D-glucopyranoside, MUG) and its sulfo derivative (MUGS) (for β-hexosaminidase), and Bodipy-conjugated Micrococcus lysodeikticus cell wall (for lysozymes). The enzyme activity toward MUGS in the whole cells of CPBF8gs strain (0.045 U/g) decreased by 81%, compared to control (0.234) (Figure 4A), whereas that toward MUG reduced by 32% (44.4 and 30.4 U/g in control and CPBF8gs strain, respectively) (Figure 4B). The activity toward MUGS or MUG is known to attributable to β-hexosaminidase activity of a homodimer of α-subunit, or that of both a homodimer of β-subunit and a α/β-subunit heterodimer [43]. The β-hexosaminidase activity toward MUGS and MUG, secreted to the culture medium, also decreased by 37 and 43% in CPBF8gs strain, respectively. The lysozyme activities in the whole cell lysates of CPBF8gs strain appear to be slightly decreased (4.3%), while the amylase activity remained unchanged (Figures 4, C and D). One should know that the degree of lysozyme secretion was much higher than that of β-hexosaminidase. β-Hexosaminidase activity detected in the culture supernatant was almost negligible (Figure 4A and B), and may be attributable to lysed cells. In addition, lysozyme activity detected in the whole lysates and the culture supernatant appear to be attributable to proteins other than lysozyme 1 and 2 because the lysozyme activity in the isolated phagosomes and the amount of lysozyme 2 in the whole cells and phagosomes detected by specific antibody in immunoblot analysis greatly decreased (see below).
In order to further investigate whether CPBF8 is involved in trafficking of β-hexosaminidase α-subunit and lysozymes to phagosomes, we compared these activities in phagosomes isolated and purified, as previously described [12], from CPBF8gs and control strains (Figure 4E). We observed that β-hexosaminidase α-subunit and lysozyme activities in purified phagosomes decreased by 90 and 96%, respectively, in CPBF8gs, compared to the control strain, while the amylase activity in phagosomes remained unchanged. Immuno blot analysis also confirmed the results of the activity assays, and indicated that lysozyme 2 is not transported to phagosomes in CPBF8gs strain (Figure 4F).
To understand biological significance of CPBF8, we examined phagocytosis and degradation of a representative Gram-positive bacillus Clostridium perfringens in CPBF8gs strain. We microscopically monitored a course of degradation of ingested C. perfringens (Figure 5A). Intact and rod-shaped C. perfringens becomes rounded in phagosomes when it is permeabilized and degraded. After 4 h co-incubation of SYTO-59-prestained bacteria with the amoebae, both the rod-shaped and rounded bacteria were counted (Figure 5B). While the total number of bacteria ingested were comparable in the control and CPBF8gs strains (12.2±3.8 and 9.1±3.8 per amoeba, respectively), the number of rounded bacteria (0.3±0.4 per amoeba) dramatically decreased in CPBF8gs compared to the control (8.1±4.1 per amoeba), whereas that of rod-shaped bacteria increased by two fold (8.8±3.9 and 4.0±3.2, respectively). These results clearly indicate that degradation of C. perfringens was inhibited by the repression of CPBF8.
We investigated whether CPBF8 is involved in the cytopathic effects on monolayers of cultured mammalian cells. The monolayers of Chinese hamster ovary (CHO) cells were incubated with the control and CPBF8gs strains for 1–3 h, and destruction of CHO cells was measured. The cytopathic activity caused by CPBF8gs strain was lower by 23–29% at all time points compared to control strain (Figure 5C). The observed cytopathic effect was partially blocked by 200 µM of the cysteine protease inhibitor E-64 [44], [45]. The cytopathic effect by the control strain was reduced by 20–22%, whereas that by CPBF8gs was decreased by 41–45% (Figure 5D). These results support the hypothesis that the decrease in the cytopathic activity in CPBF8gs was due to the decrease in β-hexosaminidase α-subunit and lysozymes.
To confirm this hypothesis, we also created the strains where β-hexosaminidase α-subunit or lysozyme 1 genes was repressed (HexAgs and Lys1gs strains). These silenced strains showed reduced cytotoxity to CHO cells compared to the control mock transformant by 9–18% reduction, as measured at 60 mins of co-incubation (Supplemental information Figure S2). These data indicate that secreted (and maybe also intracellular) lysozymes are involved in CHO cytolysis, and that intracellular β-hexosaminidase α-subunit is also involved in pathogenesis against mammalian cells, though its mechanism remains undetermined. We also attempted to directly test cytotoxic activity of recombinant hexosaminidase and lysozymes produced by in vitro translation (up to 10 µg/ml final), but failed to demonstrate it.
CPBF family proteins show common structural organization: the signal peptide at the amino terminus, the transmembrane domain close to the carboxyl-terminal end, and the YxxL motif in the cytosolic tail located at the carboxyl terminus (Nakada-Tsukui K, et al., unpublished data). Besides, among 11 members, only 3 members, CPBF6, CPBF7, and CPBF8, have a stretch of serine-rich hydrophilic region prior to the transmembrane domain (Figure 6A and B). In order to investigate whether this region is involved in the binding of CPBF8 to the cargos and whether the region is involved in the phagosomal transport, we created a transformant that expressed HA-tagged CPBF8 lacking the 23-a.a.-long serine-rich region (CPBF8ΔSRR-HA). We immunoprecipitated CPBF8-HA and CPBF8ΔSRR-HA using anti-HA antibody from lysates of the corresponding strains. Both the silver-stained SDS-PAGE gel and immunoblot analysis with HA antibody showed that the size of CPBF8ΔSRR-HA (∼100 kDa) detected was ∼50 kDa smaller than that of CPBF8-HA (∼150 kDa), which was larger than predicted (Figure 6 E and F, see below “The nature of post-translational modifications of CPBF8”). The amount of the 75- and 25-kDa proteins, which correspond to β-hexosaminidase α-subunit and lysozymes, respectively, detected in the immunoprecipitated samples from the lysates of CPBF8ΔSRR-HA significantly decreased, compared to that from CPBF8-HA strain (Figure 6E and F). The identity of the precipitated proteins was confirmed by the immunoblots using anti-β-hexosaminidase α-subunit antibody and lysozyme 2 antibody (Figure 6F).
To further map the region and amino acids responsible for the cargo binding, we constructed the variant forms of CPBF8 in which one of two stretches of three serines in the SRR were replaced by alanines (CPBF8AAA1-HA and CPBF8AAA2-HA, respectively) (Figure 6C). CPBF8AAA1-HA showed reduced ability to bind β-hexosaminidase α-subunit and lysozyme 2, compared to CPBF8-HA and CPBF8AAA2-HA (Figure 6E and F). These data indicate that the region containing the first stretch of three serine residues is essential for the binding with the cargos. Furthermore, silver staining and immunoblots with anti-HA antibody of the lysate from CPBF8AAA1-HA showed a ∼20 kDa reduction in the apparent molecular size of CPBF8AAA1-HA, compared to CPBF8-HA. These data were also consistent with a premise that this portion is directly post-translationally modified or indirectly involved in post-translational modifications (see below). We also attempted to show direct evidence that β-hexosaminidase α-subunit and lysozymes bind to SRR by constructing a truncated form of CPBF8, in which the signal peptide, SRR, the transmembrane domain, and the cytosolic region of CPBF8 were included (designated as SRR-HA). However, neither β-hexosaminidase α-subunit nor lysozymes was detected by immunoprecipitation of SRR-HA (data not shown). This indicates that the SRR per se may not be sufficient for post-translational modifications required for cargo binding. Immunofluorescence assay showed that the localization of CPBF8ΔSRR-HA, CPBF8AAA1-HA, CPBF8 AAA2-HA, and SRR-HA such as phagosome recruitment (Figure 6D and Supplemental information Figure S3A–C) was indistinguishable from that of CPBF8-HA. Therefore, SRR does not appear to be essential to phagosome targeting.
The apparent molecular mass of CPBF8ΔSRR-HA detected with silver staining and immunoblots with anti-HA antibody was ∼50 kDa smaller than that of CPBF8-HA (Figure 6, E and F, Figure 7). The reduction of the apparent size was larger than the predicted decrease based on the deletion of the amino acids (23 a.a. corresponding to 2.4 kDa for CPBF8ΔSRR-HA). To better understand the nature of the post-translational modification of CPBF8, we treated the transformant with 10 µg/ml tunicamycin, which is an inhibitor of asparagine-linked glycan modification, for 24 h (40). However, tunicamycin treatment did not affect the apparent mobility of immunoprecipitated CPBF8-HA on SDS-PAGE (data not shown), despite the fact that a potential N-linked glycosylation site is present in CPBF8 (Asn383). It was previously shown that the major GPI-anchored surface antigen of E. histolytica trophozoites contains O-phosphodiester-linked sugars [46]. We examined if the post-translational modification of CPBF8 contains O-phosphodiester-linked sugars by the treatment of immunoprecipitated CPBF8 with trifluoroacetic acid (TFA). SDS-PAGE and immunoblot analyses showed that TFM treatment of immunoprecipitated CPBF8 reduced the apparent molecular size of CPBF8 to ∼120 kDa (Figure 7), which was similar to the size of CPBF8AAA1-HA (Figure 6, E and F), while that of CPBF8ΔSRR-HA remained unchanged by TFA treatment. Altogether, CPBF8 appears to possess O-phosphodiester linked carbohydrates via the first stretch of serines within SRR, and this region seems to be responsible for the binding with β-hexosaminidase α-subunit and lysozymes. It should also be noted that the apparent size of TFA-treated CPBF8-HA and CPBF8AAA1-HA was significantly (∼20 kDa) larger than that of CPBF8ΔSRR-HA. The difference between CPBF8AAA1-HA and CPBF8ΔSRR-HA was apparently larger than the predicted size of SRR (2.4 kDa), suggesting that other post-translational modification(s) may be present in other region(s) of SRR.
CPBF8 was first identified in phagosomes by our previous proteome study of the purified phagosomes [12]. We have recently rediscovered CPBF8 as a homolog of CPBF1. CPBF1 was isolated as a potential receptor/carrier of the major virulence factor of E. histolytica, CP5, by virtue of its binding activity to CP5 (Nakada-Tsukui K, et al., unpublished data). CPBF8 represents a novel hydrolase receptor for the following reasons. First, CPBF8 is the first receptor that binds to and transport β-hexosaminidase α-subunit and lysozymes to lysosomes/phagosomes, in the manner that is distinct from mannose-6-phosphate receptor- and sortilin-dependent pathway. Second, there is no CPBF8 homolog in other organisms, and showed no sequence similarity to mannose-6-phosphate receptors or sortilin at the primary sequence level. Mannose-6-phosphate receptor and sortilin 1 are the membrane receptor of cathepsin D/β-hexosaminidase [47] and prosaponin/sphingolipid activator protein [48], respectively. Third, CPBF8 is post-translationally modified at its unique serine-rich region, and the modification is essential for the cargo binding.
The two representative CPBF members, CPBF1 and CPBF8, are localized to distinct compartments in steady state. Immunofluorescence assay using two markers, LysoTracker and PNT, clearly showed distinct distribution of CPBF1 and CPBF8. CPBF8 was well colocalized with LysoTracker and PNT (Figure 1), whereas CPBF1 was seldom localized to lysosomes or colocalized with PNT (Nakada-Tsukui K, et al., unpublished data). The different localization of CPBF8 and CPBF1 may be attributable to the motif sequences at the carboxyl terminus. As mentioned above, the YxxL motif is located at the very end of the carboxyl terminus CPBF1, while CPBF8 ends with a stretch of YxxLA, suggesting a possibility that different accessory molecule(s) bind to CPBF 1 and CPBF8.
When CPBF1 and CPBF8 were recruited to phagosomes, the subdomain of the phagosomal membrane they first come in contact with, seems to be indistinguishable. CPBF8 (Figure 1A) and CPBF1 (Nakada-Tsukui K, et al., unpublished data) were recruited to the basolateral portion of the “phagocytic mouth”, similar to the domain where phosphatidylinositol-3-phosphate is localized [35]. The fact that CPBF8 and EhPNT colocalize in steady state and are simultaneously transported to phagosomes upon phagocytosis, suggests that similar trafficking pathway and mechanism may be used despite the apparent difference in the predicted strength of the membrane association (one versus 11–13 transmembrane regions, respectively). Further experiments are necessary to identify a key factor that determines cellular localization.
Affinity immunoprecipitation of CPBF8 led us to identify β-hexosaminidase α-subunit and lysozymes as the major cargos of CPBF8. The cargo specificity of CPBF8 was further supported by the dramatic reduction of their enzymatic activities in both the whole cell and phagosomes, by repression of CPBF8 gene. The observation is consistent with the previous finding on the mouse fibroblast, in which knockout of a specific receptor caused decrease in the intracellular activities of β-hexosaminidase, β-galactosidase, and β-glucuronidase [47].
Although deletion or repression of hydrolase receptors often causes missecretion of cargos (e.g., [47]), repression of CPBF8 gene did not result in missecretion of β-hexosaminidase α-subunit and lysozymes. This is also in good contrast to CPBF1, gene silencing of which caused missecretion of CP5 (Nakada-Tsukui K, et al., unpublished data). The outcome of deletion or repression of a hydrolase receptor largely varies. These data suggest that the trafficking, processing, activation, secretion, and degradation of cysteine proteases and β-hexosaminidase α-subunit/lysozymes largely differ despite they use the transport receptors that belong to the same protein family. It is likely that non- or mis-targeted β-hexosaminidaseα-subunit and lysozymes remain inactive or are swiftly degraded by proteasomes.
Lysozymes are the well-established anti-bacterial protein, which degrades the cell wall of Gram-positive bacteria. Thus, the reduction of destruction of C. perfringens caused by CPBF8gs can be directly attributable to the loss of lysozymes in phagosomes. It has also been reported that β-hexosaminidase in Drosophila melanogaster [49] and in murine macrophages is important to repress and control the growth of Mycobacterium marinum [49]. Therefore, the defect in the transport of β-hexosaminidase α-subunit to phagosomes may also be responsible for the decrease in degradation of C. perfringens in CPBF8gs strain.
We have shown that the cytopathic effects of trophozoites were decreased by repression of CPBF8, and the reduction of the cytopathy was not due to cysteine proteases, as the decrease in the cytopathy caused by CPBF8 gene silencing was not cancelled by the cysteine protease inhibitor. These results support the premise that the enzymatic activity that decreased in CPBF8gs strain, i.e., β-hexosaminidase α-subunit and/or lysozymes, is responsible for the cytopathic effect remaining after E-64 treatment. The present study is the first to show the causal link of β-hexosaminidase α-subunit and lysozyme with virulence in eukaryotic pathogens. In mammals, β-hexosaminidase is known to hydrolyze GM2 (sphingomyelin) [6]. Recently, it has also been shown that β-hexosaminidase is involved in fertilization in hamster [50]. In addition, β-hexosaminidase from the Asian corn borer Ostrinia furnacalis was shown to degrade chitin [51]. Phylogenetic analysis indicated that both α and β-subunit of β-hexosaminidase from E. histolytica belong to the same clade as insect counterparts [52]. This clade contains two different functional β-hexosaminisases from insects. One is involved in the alteration of the structure of N-glycans generated in the cell, while the other plays in the chitin degradation processes. Although it has not been demonstrated how β-hexosaminidase is involved in the cytopathy of E. histolytica, it is conceivable that E. histolytica β-hexosaminidase degrades glycoconjugates of the extracellular matrix components to pass basement membranes, as previously suggested [53]. It was reported that lysozyme gene was poorly expressed in E. histolytica Rahman strain, which apparently lost virulence, and non-virulent E. dispar, compared to E. histolytica HM-1:IMSS strain [54], [55]. Furthermore, lysozyme was also poorly expressed in E. histolytica trophozoites that were treated with 5-azacytidine (5-AzaC), a potent inhibitor of DNA methyltransferase, and showed reduced virulence [56]. Altogether, lysozymes are involved in in vitro cytotoxicity and in vivo virulence in E. histolytica.
The fact that CP was responsible for only 15–25% of the cytopathic effect on CHO cells in control transformant apparently disagreed to our previous finding [29], where 70–75% of the cytopathic effect in the HM-1 reference strain was attributable to CP. This is likely explained by the fact that the parental strain (G3) of the transformants for gene silencing has uncharacterized defects as well as lack of amoebapores.
As mentioned above, the most striking difference between non-lysosomal/phagosomal CPBF1 and lysosomal/phagosomal CPBF8, at the primary sequence level is the serine-rich domain in the luminal region, found exclusively in CPBF6, 7, and 8. While either the deletion of this region or point mutations of the serine stretch of the serine-rich region did not affect trafficking to phagosomes, the binding of CPBF8 to β-hexosaminidase α-subunit and lysozymes was significantly reduced. Although we cannot exclude the possibility that truncated CPBF8 was partially mis-folded and thus unable to bind to its cargos, characterization of the post-translational modifications of CPBF8 via the serine-rich region by chemical removal of its potential O-phosphodiester-linked glycans strongly indicates that the O- phosphodiester-linked glycan within this region appears to be involved in cargo binding.
In summary, we have discovered and characterized the novel membrane-associated receptor for β-hexosaminidase α-subunit and lysozymes, CPBF8, from E. histolytica. We have demonstrated that CPBF8 plays an important role in the degradation of ingested bacteria in phagosomes and the cysteine protease-independent cytopathy on mammalian cells.
Trophozoites of E. histolytica strain HM-1:IMSS Cl-6 and G3 [41] were cultured axenically at 35°C in 13×100 mm screw-capped Pyrex glass tubes or plastic culture flasks in BI-S-33 medium as previously described [57], [58]. CHO cells were grown in F-12 medium (Invitrogen, Grand Island, NY) supplemented with 10% fetal bovine serum on a 10-cm-diameter tissue culture dish (IWAKI, Tokyo, Japan) under 5% CO2 at 37°C. Clostridium perfringens was kindly given by Fumiya Kawahara, Nippon Institute for Biological Science Japan.
The protein coding region of CPBF8 gene was amplified by PCR from cDNA using sense and antisense oligonucleotides: 5′-GCGAGATCTATGTTGGCACTCTTCGCCATC-3′ and 5′-GCGAGATCTAGCTAAAGTAGCATATCCAGA-3′ (BglII restriction sites are underlined). The amplified PCR product was digested with BglII and ligated into BglII-digested pEhExHA [28], to produce pEhEx-CPBF8-HA. The plasmid to produce the mutant form of CPBF8 that lacks 23a.a.-long serine-rich region (pEhEx-CPBF8ΔSRR-HA) was constructed as follows. A DNA sequence was amplified by PCR from pEhEx-CPBF8-HA using sense and antisense oligonucleotides: 5′-CCAGTTGGATGG- ATTGTATTTGGTGTTCTT-3′ and 5′-GTCATCTGGTTGTGGATCTTC TTTAGCATC-3′. The PCR product was treated with BKL kit (Takara, Shiga, Japan), and self-ligated to produce pEhEx-CPBF8ΔSRR-HA. The resulting plasmid encodes a mutant CPBF8 protein lacking 843–865 a.a.-region (“SSSVPPTPSSSTDDKDDDSHHST”). Two variant forms of CPBF8 containing point mutations, S843A, S844A, and S845A, or S851A, S852A, and S853A, respectively (designated as CPBF8AAA1-HA and CPBF8AAA2-HA), are constructed as follows. DNA fragments corresponding to the portion of the protein coding region encompassing from the amino-terminal region to the mutated amino acids of CPBF8AAA1-HA and CPBF8AAA2-HA were amplified by PCR from pEhEx-CPBF8-HA, using sense and antisense oligonucleotides: 5′-ACAAACACATTAACAATGTTGGCACTCTTCGCC -3′ and 5′-TGGTACTGCAGCAGCGTCATCTGGTTGTGG -3′ (for CPBF8AAA1-HA) and 5′-ACAAACACATTAACAATGTTGGCACTCTTCGCC -3′ and 5′-ATCAGTTGCAGCAGCTGGTGTTGGTGGTAC -3′ (for CPBF8AAA2-HA), where the nucleotides corresponding to the mutated amino acids (serine to alanine) are double-underlined). DNA fragments corresponding the portion of the protein coding region encompassing the mutated amino acids of to the carboxyl-terminal region of CPBF8AAA1-HA and CPBF8AAA2-HA were amplified by PCR from pEhEx-CPBF8-HA using sense and antisense oligonucleotides: 5′- GCTGCTGCAGTACCACCAACACCATCTTC-3′ and 5′- ATCATATGGATACATAGCTAAAGTAGCATATCC -3′ (for CPBF8AAA1-HA), and 5′- GCTGCTGCAACTGATGATAAAGATGATG -3′ and 5′- ATCATATGGATACATAGCTAAAGTAGCATATCC-3′ (for CPBF8AAA2-HA), where the nucleotides corresponding to the mutated amino acids (serine to alanine) are double-underlined. The two amplified fragments were mixed and ligated to BglII-digested pEhExHA using GENEART Seamless ligation kit (Invitrogen, San Diego, CA, USA), to produce pEhEx-CPBF8AAA1-HA and pEhEx-CPBF8AAA2-HA. SRR-HA, which consisted of the signal peptide, SRR, the transmembrane domain, and the cytosolic region of CPBF8, was constructed as follow. A DNA fragment corresponding to the signal peptide and additional four amino acids was amplified by PCR using sense and antisense oligonucleotides: 5′- ACAAACACATTAACAATGTTGGCACTCTTCGCC -3′ and 5′- AGAACATGTGTACTGTCCATACGCAAC -3′. A DNA fragment corresponding to SRR, the transmembrane domain, and cytosolic region was amplified by PCR using sense and antisense oligonucleotides: 5′- CAGTACACATGTTCTGATGACTCTTCTTCAGTACC -3′ and 5′- ATCATATGGATACATAGCTAAAGTAGCATATCC-3′. The two amplified fragments were mixed and ligated to BglII-digested pEhExHA using GENEART Seamless ligation kit to produce pEhEx-SRR-HA. The transformants that expressed CPBF8-HA, CPBF8ΔSRR-HA, CPBF8AAA1-HA, CPBF8AAA2-HA, or SRR-HA were established by transfection of the wild-type HM1:IMSS Cl6 strain by liposome-mediated transfection as previously described [59].
Repression of gene expression was accomplished by gene silencing, which has recently demonstrated to be mediated by nuclear localized antisense small RNAs with 5′-polyphosphate termini [42]. Gene silencing has been seen only in G3 strain, in which amoebapore genes are repressed. Thus, whatever phenotypic changes are observed by gene silencing of additional gene of interest, need to be compared against the control G3 strain transformed by the mock gene silencing plasmid (pSAP2-Gunma). Furthermore, it should be evaluated, if possible, whether the phenotypic changes are not caused by synergistic effects with amoebapore silencing. For gene silencing of CPBF8, β-hexosaminidase α-subunit, and lysozyme 1 genes, the 420-bp-long 5′-end of the protein coding region was amplified by PCR from cDNA using sense and antisense oligonucleotides: 5′-CGCAGGCCTATGTTGGCACTCTTCGCCATC-3′ and 5′-GCAGAGCTCATTTTCTTCAACTAACTTAAC-3′ (CPBF8); 5′-CGCAGGCCTATGCCATATCCAAGCTCAG-3′ and 5′-CGCGAGCTCGTTTGATGAAATTCTAATT-3′ (β-hexosaminidase α-subunit); 5′-CGCAGGCCTATGTTCGCTCTCTTTTTGTG-3′ and 5′-CGCGAGCTCACCATGGACAATACCAATAGC-3′ (lysozyme 1) (StuI and SacI restriction sites are underlined). The PCR-amplified DNA fragment was digested with StuI and SacI, and ligated into StuI- and SacI-digested pSAP2-gunma [60], to produce pSAP2-CPBF8, pSAP2-HexA, and pSAP2-Lys1. The gene-silenced strains were established by the transfection of G3 strain with the corresponding plasmids as described above.
CPBF8 and lysozyme 2 antibodies were raised against recombinant histidine-tagged partial CPBF8 (a.a. 14–292) and full-length lysozyme 2, respectively. The method of production and purification of recombinant proteins were described before [40]. The expression plasmids for histidine-tagged CPBF8 (a.a.14–292) and lysozyme 2 were introduced into BL21(DE3) competent cells (Invitrogen). Expression of the recombinant protein was induced with 0.1 mM isopropyl-β-thiogalactoside at 37°C for 3 h. The histidine-tagged fusion protein was purified under denaturing condition using Ni-NTA agarose (QIAGEN, Hiden, Germany), according to the manufacturer's protocol. β-hexosaminidase α-subunit antibody was raised against mixture of the peptides LQQQTGLQDFKVSL (a.a. 77–90) and GWSKSKEYSDIQKF (a.a. 348–361).
Anti-HA 11MO mouse monoclonal antibody was purchased from Berkeley Antibody (Berkeley, CA). Alexa Fluor anti-mouse and antirabbit IgG and horseradish peroxidase (HRP)-conjugated goat anti-mouse were purchased from Invitrogen.
For the staining of lysosomes, amoebae were incubated in the BI-S-33 medium containing LysoTracker Red DND-99 (Invitrogen) (1∶500) at 35°C for 12 h. To visualize phagosomes, CHO cells were pre-stained with 10 βM of CellTracker Blue (Invitrogen) in F-12 medium supplemented with 10% fetal bovine serum at 37°C for 3 h. Labeled CHO cells were washed with phosphate-buffered saline (PBS), and added to 8-mm wells containing E. histolytica trophozoites on a slide glass (8 well 8 mm standard slide glass, Thermo Scientific, Rockford, IL) and further incubated at 35°C for 10–60 minutes. After the incubation, cells were fixed with 3.7% paraformaldehyde for 10 min, and permeabilized with 0.2% saponin/PBS for 10 min at ambient temperature. The cells were then reacted with anti-HA 11MO mouse monoclonal antibody (diluted at 1∶1000) and Alexa Fluor-488 antimouse secondary antibody (1∶1000). The samples were examined on a Carl-Zeiss LSM510 conforcal laser-scanning microscope (Thornwood, NY). Images were further analyzed using LSM510 software. We defined CPBF8-HA- and LysoTracker-positive and negative vacuoles/vesicles as follows: 1) we measured and averaged the signal intensity (per pixel) of the whole intracellular area of a cell, and also five randomly-chosen areas outside the cell to obtain a background fluorescence level; 2) for individual vacuoles/vesicles that had continuous fluorescent signal lining the membrane, two straight lines were drawn, which make a right (90-degree) angle, and a projected fluorescence histogram was obtained for each line; 3) if the peak intensity of the point on the membrane of the vacuole/vesicle was >2 fold of the background fluorescence level, the vesicle/vacuole was defined as “signal positive”. The LysoTracker-positive vacuoles/vesicle was defined similarly, except that 1) the fluorescence of not a point on the membrane, but a whole intravesicular/vacuolar area was measured and averaged; and 2) the threshold of the average peak intensity of the LysoTracker-positive area in the vacuole/vesicle is >5 fold of the background fluorescence level.
Approximately 3×106 cells of CPBF8-HA-expressing amoebae were lysed in 2 ml of lysis buffer [50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100 (Tokyo Kasei, Tokyo, Japan), 0.5 mg/ml E-64 (Sigma-Aldrich, St. Louis, MO)], and suitable amount of Complete mini mix (Roche, Barsel, Switzerland), and was incubated with protein G-Sepharose beads (50 µl of a 80% slurry) (Amersham Biosciences, Uppsala, Sweden) at 4°C for 90 min, centrifuged at 800× g at 4°C for 3 min to remove proteins that bind to the protein G-Sepharose beads non-specifically. The precleaned lysate was mixed with 90 µl of anti-HA-conjugated agarose (50% slurry-, Sigma-Aldrich), and incubated at 4°C for 3.5 h. The agarose beads were collected by centrifugation at 800× g at 4°C for 3 min, and washed four times using wash buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton-X 100). The agarose beads were then incubated with 180 µl of HA peptide (20 µg/ml) at 4°C for overnight to dissociate proteins from the beads. The eluate was applied to SDS-PAGE and silver staining as previously described [61].
Silver stained gels were excised, destained, and tryptic-digested using modified trypsin (Applied Biosystems Darmstadt, Germany). Briefly, excised gels were transferred to a siliconized tube, dehydrated in acetonitrile, rehydrated in 30 µl of 10 mM dithiothreitol in 0.1 M ammonium bicarbonate and reduced at room temperature for 30 min. The sample was then alkylated in 30 µl of 50 mM iodoacetamide in 0.1 M ammonomium bicarbonate at room temperature for 30 min. The reagent was removed and the sample was dehydrated in 100 µl acetonitrile, rehydrated in 100 µl of 0.1 M ammonium bicarbonate, and then dehydrated again in 100 µl acetonitrile and completely dried by vacuum centrifugation. Samples were then rehydrated in 20 ng/ml trypsin in 50 mM ammonium bicarbonate on ice for 10 min. Any excess trypsin solution was removed and 20 µl of 50 mM bicarbonate added. The samples were digested overnight at 37°C and resultant peptides were extracted in two 30 µl aliquots of 50% acetonitrile/5% formic acid. The tryptic peptides were eluted from the gel and then desalted by Ziptip. The resulting peptide mixture was separated by reverse phase chromatography (DiNa nano LC system; KYA Tech, Tokyo, Japan) using a 0.15 mm×50 mm ID HiQ sul C18W3 column (KYA Tech) and elution with 0.1% formic acid/2% CH3CN (solvent A) and 0.1% formic acid/80% CH3CN (solvent B) using a program 0% solvent B for 15 min, gradient at 4%/min for 2 min, gradient at 0.86%/min for 43 min, 11%/min for 5 min, 100% solvent B for 10 min with a total flow rate of 300 nl/min. The eluting peptides were ionized by electrospray ionization and analyzed by a 3200 Q TRAP LC/MS/MS System (Applied Biosystems). Peptide MS/MS spectra were acquired in an information-dependent manner using the Analyst QS software 2.0 acquisition features (Smart Exit, rolling collision energy, and dynamic exclusion).
Peptide sequence data obtained by mass spectrometry were analyzed against the E. histolytica genome database at The Institute for Genomic Research (TIGR) (http://www.tigr.org/tdb/e2k1/eha1/) using the Sequest algorithm. Sequencing data were also analyzed against the non-redundant database at the National Center of Biotechnology Information (NCBI). Individual predicted protein sequences were manually analyzed by BLAST search (http://www.ncbi.nlm.nih.gov/BLAST/) against the non-redundant database at NCBI. The identification of the protein was considered significant when at least two non-overlapping peptides of a protein were detected with the probability score >95%. The identified proteins were classified using the annotations provided in the TIGR and NCBI database and results of BLAST search.
Total RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. The synthesis of cDNA was performed using the SuperScript III First Strand Synthesis System (Invitrogen) according to the manufacturer's instructions. The cDNA synthesis was completed on a DNA Engine Peltier Thermal Cycle (Bio-Rad Laboratories, Inc., Hercules, CA, USA) and treated with deoxyribonuclease I (Invitrogen) to exclude genomic DNA. PCR was performed with the resulting cDNA as a template and specific oligonucleotide primers. Primers used were 5′ (-CAAGTGCTTCAAGTACTCAACCATC-3′and 5′-ACCATTGTTACTTCTCTTTTTACGA-3′ (CPBF6); 5′-ATTTTATATGAACTCTCCAGACGAT-3′ and 5′-TAATAGTAATAATCAGCACAATAAC-3′ (CPBF7). 5′-AGAAGGTTTTGTCGATGTTCAATTC-3′ and 5′-TAACACATCCAGCAAGAATACCAGC-3′ (CPBF8). PCR reaction mixture contains 0.2 U of high-fidelity DNA polymerase (Phusion, FINNZYME, Espoo, Finland), 5×Phusion buffer, 0.16 mmol/L of each dNTPs, 0.5 µM Primers. Parameters for PCR were: an initial step of denaturation at 98°C for 30 sec, 25 cycles of amplification (at 98°C for 5 sec, 53°C for 20 sec, and 72°C for 2 min), and a final extension at 72°C for 5 min.
Expression analysis was performed using a custom E. histolytica array from Affymetrix, Inc. (Santa Clara, CA, USA), as previously described [62]. Labeled cRNA for hybridization was prepared from 5 µg of total RNA according to published Affymetrix protocol. Hybridization and scanning were performed according to Affymetrix protocols.
A semi-confluent culture was harvested at 48–72 h after initiation of the culture and resuspended in modified Opti-MEM (Invitrogen), Opti-MEM supplemented with 1 mg/ml ascorbic acid and 5 mg/ml cysteine. Approximately 4×105 amoebae in 1 ml of the medium were seeded to wells of a 12-well plate. After the culture was incubated at 35.5°C for 2 h, the culture supernatant was centrifuged at 400× g for 5 min at 4°C to remove debris. The plates were chilled on ice for 5 min and detached trophozoites were collected.
Approximately 3–5×106 trophozoites (per flask) were cultured in 25-cm2 flasks for 48 h, and washed gently with warm modified Opti-MEM. Approximately 107 carboxylate-modified latex beads (Polyscience, Warrington, PA) were added to the flasks, and the flasks were centrifuged at 190× g for 5 min to bring the beads into contact with the trophozoites. After centrifugation, the flasks were placed on ice for 10 min. The trophozoites were washed three times with cold PBS containing 20% sucrose, followed by centrifugation at 190× g for 5 min to remove uningested beads. The cells containing latex beads were then resuspended in warm BI-S-33 medium, further incubated at 37°C, and harvested after 120 min. Bead-containing phagosomes were purified as previously described [12] with some modifications. Briefly, after harvesting, the amoebae that contained latex beads were resuspended in cold homogenization buffer (250 mM sucrose, pH 7.4, 3 mM imidazole, 10 mM cysteine protease inhibitor E-64, CompleteMini protease inhibitor cocktail) and homogenized with a Dounce homogenizer on ice. Phagosomes containing latex beads were then separated by flotation on a sucrose step gradient as described [12]. All sucrose solutions were made in 3 mM imidazole, pH 7.4 containing 10 mM cysteine protease inhibitor E-64. Sucrose was added to the homogenized lysate to 40%. 2 ml of the lysate containing 40% sucrose, 2 ml each of 35, 25 and 10% sucrose solutions were carefully overlaid in a 10 ml ultracentrifuge tube. The sample was centrifuged in a swinging bucket rotor MLS-50 (Beckman, Brea, CA) at 131,000× g at 4°C for 1 h. The phagosome fraction was collected from the interface of the 10 and 25% sucrose solutions. The collected fraction (1 ml) was mixed with 3 ml of 50% sucrose, and transferred to a new tube. To the sample, 4 ml of 25% and 2 ml of 10% sucrose solutions were overlaid, and the sample was centrifuged at 131,000× g at 4°C for 1 h. The separated phagosome sample, collected from the interface of the 10 and 25% sucrose solutions, was finally mixed with the same volume of 3 mM imidazole solution and centrifuged at 13,000× g at 4°C for 5 min. The pellet were suspended with 7% sucrose solution and stocked in −80°C.
β-hexosaminidase assay was performed as previously described [43] with some modifications. Briefly, the reaction mixture consisted of 25 mM citrate buffer, pH 4.0, 10 mM 4-methylumbelliferyl-6-sulfo-2-acetamido-2-deoxy-β-D-glucopyranoside (MUGS)(Merck, Darmstadt, Germany) or 4-methylumbelliferyl -2-acetamido-2-deoxy-β-D-glucopyranoside (MUG)(Sigma) as substrates, and amoeba cell lysates, culture supernatant or phagosome fraction. The reaction was initiated by addition of the substrates and stopped by addition of 0.2 M glycine/0.2 M sodium carbonate, pH 10.5). The fluorescence of the released 4-methylumbelliferone was measured at excitation and emission wavelengths of 360 and 440 nm, respectively. The lysozyme assay was performed by EnzChek Lysozyme Assay Kit (Invitrogen). Briefly, the reaction mixture contained amoeba lysates, culture supernatant or phagosomal fraction and 200 µg/ml of Bodipy-conjugated Micrococcus lysodeikticus cell wall, and the fluorescence was measured at excitation and emission wavelengths of 485 and 530 nm, respectively. The amylase assay was performed by EnzChek amylase Assay Kit (Invitrogen). Briefly, the reaction mixture contained amoeba lysates, culture supernatant or phagosomal fraction, and 200 µg/ml of substrate solution, and the fluorescence was measured at excitation and emission wavelengths of 485 and 530 nm.
Approximately 1.5×104 trophozoites of control or CPBF8gs strain and 1.5×106 C. perfringens were incubated in 150 µl of BIS medium with 10 µM SYTO-59 (Invitrogen) on glass bottom culture dish(Mattek, MA, USA) under anaerobic condition for 4 h. After incubation, the cell was washed with BIS and observed by microscopy under anaerobic condition on a Carl-Zeiss LSM510 conforcal laser-scanning microscope. The numbers of rod and round shape bacteria are counted. Images were further analyzed using LSM510 software.
The soluble region of lysozyme 1 and 2 were expressed by TN SP6 High-Yield Wheat Germ Protein Expression System (Promega, WI, USA). The HA-tagged lysozyme 1 and 2 were amplified by PCR from cDNA using sense and antisense oligonucleotides: 5′- GGGGCGATCGCATGTATCCATATGATGTTCCAGATTATGCTAAATTAGGTATTGATGTCTCTC -3′ and 5′- GGGGTTTAAACTTATGGTTTGTAGTTATAATC -3′ (for HA-lysozyme 1) and 5′- GGGGCGATCGCATGTATCCATATGATGTTCCAGATTATGCTGTAGATGTATCTCAACC -3′ and 5′- GGGGTTTAAACTTAAAAATTAAATAAAAAGAAATGAG -3′ (for HA-lysozyme 2), where PmeI and SgfI restriction sites are underlined and the sequence corresponding to the HA-tag are double underlined, respectively. The PCR-amplified DNA fragments were digested with PmeI and SgfI, and ligated into PmeI- and SgfI-digested pF3A WG (BYDV) Flexi vector, to produce pF3A-HA-lysozyme 1 and 2. Expression of HA-lysozyme 1 and 2 was performed according to the manufacturer's protocol. The reactions were done at 25°C for 2 h. The purification of HA-lysozyme 1 and 2 was performed as described above for immunoprecipitation.
CHO monolayer destruction was measured as described previously with minor modifications [63]. Briefly, CHO cells were grown with confluent in 24 well plate for over-night.at 37°C and 5% CO2.The medium was removed and the plates were washed with modified Opti-MEM medium. Approximately 5×104 trophozoites of control or CPBF8gs strains were resuspended in 0.5 ml of modified Opti-MEM medium and added to each well. The plates were incubated under anaerobic conditions at 35.5°C for up to 3 h. The plates were placed on ice for 10 min to detach trophozoites. The number of CHO cells remaining in the wells was measured by WST-1 reagent (Roche) as described previously [64]. The cytopathic activity of recombinant HA-lysozyme 1 and 2 toward mammalian cells was evaluated by incubating confluent CHO cells with the mixture of purified recombinant proteins and 10% FBS/F-12 (1∶99) on a 24 well plate. The plates were incubated at 35.5°C for 24 h. The number of CHO cells remaining in the wells was estimated by WST-1 reagent.
CPBF1, EHI_164800, XP_655218; CPBF2, EHI_087660, XP_653276; CPBF3, EHI_161650, XP_649180; CPBF4, EHI_012340, XP_655897; CPBF5, EHI_137940, XP_654065; CPBF6, EHI_178470, XP_653036; CPBF7, EHI_040440, XP_649361; CPBF8, EHI_059830, XP_652899; CPBF9, EHI_021220, XP_655360; CPBF10, EHI_191730, XP_649015; CPBF11, EHI_118120, XP_656044; β-hexosaminidase α-subunit, EHI_148130, XP_657529; Lysozyme 1, EHI_199110, XP_653294; Lysozyme 2, EHI_096570, XP_656933
|
10.1371/journal.pntd.0001187 | Health Services for Buruli Ulcer Control: Lessons from a Field Study in Ghana | Buruli ulcer (BU), caused by Mycobacterium ulcerans infection, is a debilitating disease of the skin and underlying tissue. The first phase of a BU prevention and treatment programme (BUPaT) was initiated from 2005–2008, in the Ga-West and Ga-South municipalities in Ghana to increase access to BU treatment and to improve early case detection and case management. This paper assesses achievements of the BUPaT programme and lessons learnt. It also considers the impact of the programme on broader interests of the health system.
A mixed-methods approach included patients' records review, review of programme reports, a stakeholder forum, key informant interviews, focus group discussions, clinic visits and observations.
Extensive collaboration existed across all levels, (national, municipality, and community), thus strengthening the health system. The programme enhanced capacities of all stakeholders in various aspects of health services delivery and demonstrated the importance of health education and community-based surveillance to create awareness and encourage early treatment. A patient database was also created using recommended World Health Organisation (WHO) forms which showed that 297 patients were treated from 2005–2008. The proportion of patients requiring only antibiotic treatment, introduced in the course of the programme, was highest in the last year (35.4% in the first, 23.5% in the second and 42.5% in the third year). Early antibiotic treatment prevented recurrences which was consistent with programme aims.
To improve early case management of BU, strengthening existing clinics to increase access to antibiotic therapy is critical. Intensifying health education and surveillance would ultimately increase early reporting and treatment for all cases. Further research is needed to explain the role of environmental factors for BU contagion. Programme strategies reported in our study: collaboration among stakeholders, health education, community surveillance and regular antibiotic treatment can be adopted for any BU-endemic area in Ghana.
| Buruli ulcer (BU), caused by Mycobacterium ulcerans infection, is a debilitating disease of the skin and underlying tissue which starts as a painless nodule, oedema or plaque and could develop into painful and massive ulcers if left untreated. Using a combination of quantitative and qualitative methods, the study assessed the effectiveness of the BUPaT programme to improve early detection and management of BU in an endemic area in Ghana. The results of the study showed extensive collaboration across all levels, (national, municipality and community), which contributed to strengthening the programme. Health staff were trained to manage all BU cases. School teachers, municipal environmental staff and community surveillance volunteers were trained to give the right health messages, screen for detection of early cases and refer for medical treatment. WHO-recommended antibiotics improved treatment and cure, particularly for early lesions, and prevented recurrences. Improving access to antibiotic treatment is critical for early case management. Health education is required to emphasise the effectiveness of treatment with antibiotics to reduce deformities and the importance of seeking medical treatment for all skin lesions. Further research is needed to explain the role of environmental factors in BU contagion.
| In the absence of a proven strategy for preventing infection, control of Buruli Ulcer (BU) relies on efficient health services to prevent progression of pre-ulcerative conditions and treat ulcers. According to the World Health Organisation (WHO), service delivery is the primary function of any health system and entails the provision of “effective, safe, good quality care to those that need it with minimal waste”, [1] and to address health care needs through promotion, prevention, treatment and rehabilitation. WHO defines a health system as “all organisations, people and actions whose primary intent is to promote or to restore health” [1].
Buruli ulcer, caused by Mycobacterium ulcerans infection is a debilitating disease of the skin and underlying tissue which starts as a painless nodule, oedema or plaque and could develop into painful and massive ulcers if left untreated [2]. It is the third most common mycobacterial pathogen of humans, after M. tuberculosis (tuberculosis) and M. leprae (leprosy), but the most poorly understood [2], [3]. Even though case fatality is low, morbidity is high for all age groups [3]–[5] and the socio-economic implications to the individual and cost of management to the health system are enormous [6], [7].
Surprisingly, estimates of Disability Adjusted Life Years (DALYs) for Buruli ulcer, like other neglected tropical diseases (NTDS) such as guinea worm, endemic syphilis and food-borne trematode infections are not explicitly stated [8]. BU has been reported in more than 33 tropical and sub-tropical climates particularly West African countries [2], [9], and Ghana reports an average of 1000 cases each year [9]. The first case of BU was reported in Ghana in 1972 in the Ga-district. [10]. A national case search in 1998 indicated a national prevalence of 20.7/100,000 and a prevalence of 87.7/100,000 for the former Ga-district (now the Ga-West and Ga-South municipalities), the fifth most endemic in the country, yet with the highest burden in terms of healed and active lesions [11].
The first phase of a BU prevention and treatment programme (BUPaT) was initiated from 2005–2008, in the Ga-West and Ga-South municipalities in the Greater-Accra region, Ghana, to increase access to BU treatment and improve early case detection. Before the inception of the BUPaT programme, surgery was the main treatment for all BU patients. There was limited accessibility to treatment since all surgeries had to be done at the Amasaman hospital (AH), the main treatment and referral hospital for all BU cases in the Ga-West municipality. Antibiotic treatment had not been introduced and health staff had limited expertise in surgical procedures and BU case management.
The BUPaT programme employed WHO-recommended strategies which are: Building capacity of nurses and other para-medical staff for effective case detection, and management at designated health centres; training of community-based surveillance volunteers (CBSVs), school teachers, other health workers and traditional healers (THs), to enhance BU knowledge for early detection; establishing a community-based surveillance system with the help of CBSVs; compiling a database; providing surgical and antibiotic therapy for all BU patients [12].
These strategies were undertaken by a health team that consisted of staff of the national Buruli ulcer control programme (NBUCP), the BUPaT programme from World Vision Ghana, the health directorates of the Ga-West and Ga-South municipalities, surgeons from the Korle-Bu teaching hospital in Accra, the municipal school health education programme (SHEP) coordinator, municipal environmental health officers (MEHOs), CBSVs, THs and community members. This paper assesses achievements of the BUPaT programme and lessons learnt for early case detection, case management and access to treatment in a BU-endemic rural area of Accra. It also considers the impact of the programme on broader interests of the health system.
This study was conducted from November 2008 to July 2009 in the Ga-West and Ga-South municipalities. The Ga-West municipality shares boundaries with the Ga-South municipality to the west. It has a population of 215,824 inhabitants of which 48.2% are males and 51.8% are females. About 60% of the municipality's landscape is rural with about 200 scattered communities; 40% is urban and peri-urban and is densely populated. The population of the Ga-South municipality is estimated at 210,727 distributed in 362 communities. Like the Ga-West municipality, 48.2 % inhabitants are males and 51.8% are females. The population is mainly concentrated along the peri-urban areas of the municipality.
At the time of conducting the survey, the Ga-West and Ga-South municipalities were known as the Ga-West District. The Ga-West district covered the same geographical area as these two municipalities (Ga-West and Ga-South). Through a government legislative instrument, the Ga-West district was divided into two separate municipalities in 2009 for easy governance and accessibility of health services.
Since 1999, BU continues to be a major cause of morbidity in both municipalities with increasing numbers of related disabilities. Health services are provided by 3 main government health centres, Weija hospital, Amasaman hospital (AH) and the Obom Health Centre (OHC), a few private clinics, and family planning and maternity homes.
The study employed a mixed methods approach using quantitative and qualitative methods to assess the effectiveness of the BUPaT programme in improving early detection and management of BU in the Ga-West and Ga-South municipalities. This approach provided the needed framework for obtaining, understanding, comparing and cross-validating contextual information from providers and beneficiaries of BU-related health service delivery strategies. The various methods were complementary; emerging and divergent issues arising during the course of one approach were clarified with another. Aside document reviews which was ongoing over the course of the study, all the other approaches followed sequentially.
A day's forum was held with thirty five (35) persons that included the programme manager of the NBUCP, the municipal chief executive (MCE) of the Ga-West municipality, some municipal health staff, officials and BUPaT staff of World Vision Ghana, doctors and nurses from the AH and the OHC, officials from the Ghana education service, CBSVs and MEHOs. This forum reviewed the BUPaT programme activities, explored issues regarding health services delivery, capacity of health staff to deliver BU-related services and the integration of programme activities in communities and schools. Health service delivery interventions such as the role of CBSVs in case detection, early reporting and strengthening existing clinics in the community to increase access to health services were discussed. Consideration was given to community participation, sustainability of the programme as well as the next steps for future strategies at BU control.
Quarterly and annual BUPaT programme reports were studied to provide background information and insights into programme objectives, strategies and challenges.
KIIs were held with the municipal health director (MHD) of the Ga-West municipality, the programme managers of the NBUCP and the World Vision Ghana, Ga-West municipality development programme. These persons were selected because of their pivotal role in the BUPaT programme. KIIs highlighted issues on access to care, successes and challenges of the programme and emphasised strategies requiring further strengthening.
Already analysed records of 297 patients from the AH were reviewed to indicate the statistical trend, demographic characteristics of patients, assess indicators of treatment procedures, effectiveness of treatment and outcomes.
Visits were made to the OHC and the Kojo Ashong clinic to assess the effectiveness of decentralising treatment and management of Buruli ulcer.
One FGD each was held in three randomly selected endemic communities (Kojo Ashong, Avornyokope, and Balagono). Each focus group was made up of 10 purposively selected persons, comprising treated and discharged adults, and care-takers of child patients. FGDs examined community perceptions about the programme, school-based strategies, and the effectiveness of medical treatment, particularly antibiotic treatment. FGDs also considered challenges and concerns that were raised at the SF and KIIs regarding low hospital/clinic attendance and late reporting.
Information from BUPaT programme reports were subjected to a thematic content analysis. Themes were derived from activities that formed health service delivery strategies. Thematic related activities, (community-based surveillance, community education, school-based education and antibiotic therapy) were examined for their contribution to awareness creation, access to timely treatment, care and management of BU, and how best they addressed the overall aim of the BUPaT programme. Consideration was also given to the extent of collaboration and coordination of activities among stakeholders. Documented successes and challenges of the programme as well as those mentioned at the SF and during KIIs and FGDs were noted.
Discussions and interviews from the SF and the KIIs were subjected to a thematic content analysis. Interviews were conducted in English and tape-recorded. During the interviews, elaborate notes were taken and themes that emerged during these discussions were noted. Subsequently, interviews were transcribed using Microsoft Word. Transcriptions were translated and edited, preserving the original style and context. The authors developed a coding framework based on themes pertinent to the main features and strategies of the BUPaT programme [13]. These themes included ‘collaboration’, ‘health services’, ‘health education’, ‘access and utilisation’, ‘coverage’, ‘adequacy of facilities’, ‘antibiotics’, ‘surgery’, ‘complications’, ‘recurrence’, ‘patients’, ‘feeding’, ‘transportation’, ‘community’, ‘traditional healers’ and ‘community-based surveillance volunteers’.
FGDs were conducted and recorded electronically in the local languages. Notes on content and context referred to recurring themes. FGDs were translated into English and transcribed using Microsoft Office Word. Similar to the procedure for analysing the SF and the KIIs, transcriptions were subjected to a thematic content analysis. A coding scheme was devised using themes that clarified perceptions of health service delivery strategies and medical treatment. These themes included ‘volunteers’, ‘treatment’, ‘late treatment’, ‘traditional healers’, ‘herbal treatment’, ‘medicines’ and ‘costs’.
Observations during clinical visits were recorded in a notebook. We paid attention to the type of treatment given to patients, number of patients who received antibiotic care and documentation of patient data. Subsequently, clinical registers were examined to ascertain the extent to which patients adhered to treatment.
Patient data captured on the WHO BU01 forms had already been extracted and analysed by health staff and therefore there was no need for any further analysis.
The study was approved by the ethical review committee of the Ministry of Health, Ghana, and the ethics commission of Basel (Ethikkommission beider Basel (EKBB)) in Switzerland. Verbal consent was preferred to written ones since it did not pose any psychological threat and reassured all interviewees of anonymity. Both ethical review boards approved of verbal consent as long as participation in the study was voluntary, participants had been informed of the study aims and had the opportunity to ask questions. Prior to the start of all interviews, interviewees were informed about: the study aims, their rights to withdraw participation from the study, the intended use of findings to improve BU related health services and, for publications in academic journals and reports. Informed verbal consent was witnessed by two members of the BUPaT team who were not members of the research team.
Programme documents indicated that the BUPaT programme was initiated by WVG and the MHD of the Ga-West Municipality. The Municipal Chief Executive (MCE) of the Ga-West Municipality and the NBUCP were engaged at the design stage. At the onset, a Memorandum of Understanding (MOU) was formalised with the MCE to ensure partnership with the local government authorities, and subsequently the municipal health staff and beneficiary communities. Table 1 shows a timeline of BU activities in the country and study municipalities.
Programme documents, the SF and KIIs indicated a strong partnership with the NBUCP which provided technical expertise and training of health staff. To create awareness and ensure the participation of civil society, programme documents revealed that the BUPaT programme was duly launched at a durbar in the capital of the municipality, Amasaman. THs, WVG staff, officials from the NBUCP, municipal executives, health staff, teachers, CBSVs, school children and community members were in attendance.
Programme documents indicated that the core management team of the programme was the WVG Ga-West municipality manager, the MHD and the Municipal SHEP Coordinator. Selection of members for this team was guided by the main activities of the programme which were community and school health education, screening, medical treatment, surgery and wound care, community surveillance, documentation and compilation of a patients' database. Some individuals from the Municipal Health Management Team (MHMT) served as focal persons for various aspects of the programme. WVG too had a focal person for the programme, officially known as the BUPaT programme coordinator. This person was responsible for financial issues, logistics, monitoring, collation and analysis of patients' records, and served as a liaison between WVG and the MHMT. The MHD and the MHMT coordinated health activities related to BU.
A coalition of stakeholders including health, environmental, educational professionals, CBSVs and traditional rulers was formed to ensure diversity of expertise as well as community participation. As a practice, stakeholder meetings were organised quarterly to report on the progress of the programme. Additionally, a monitoring team comprising selected individuals from the stakeholder group was constituted to evaluate programme goals and objectives and follow-up on treated and discharged patients.
According to programme documents, 120 CBSVs, 40 THs, 4 MEHOs and 113 teachers from 60 schools were trained to detect early cases of BU in communities and refer promptly to health facilities for treatment. BU information was included in the school curriculum. Documents and narratives from the SF revealed that officials from the NBUCP also trained 40 nurses in BU case-detection, surveillance, wound care and prevention of disabilities associated with BU. After training, these nurses were distributed among the municipal health facilities: AH, OHC and two newly opened health centres (one each at Dome Sampahman and Kojo Ashong communities). Programme documents, the SF and KIIs also revealed that refresher courses were held quarterly for nurses, CBSVs and MEHOs. The NBUCP arranged for two surgeons from the Korle-Bu teaching hospital to perform weekly surgical operations on patients.
Programme documents indicated that the BUPaT programme aimed to reduce BU-related suffering and disability through early detection and treatment of pre-ulcer cases. The programme therefore employed health education to create awareness, screening and surveillance to detect all forms of BU, particularly early cases to increase early reporting for medical care, antibiotic care, wound dressing and surgery.
According to programme documents, AH staff and the SHEP coordinator conducted BU education and screening in 80 schools. Health staff, BUPaT programme staff and CBSVs combined efforts to conduct health education in over 600 communities. Sometimes these education campaigns culminated in BU screening. MEHOs also organised night-time film shows on BU and followed up the next day for screening. CBSVs mounted intense surveillance in their localities and paid random home visits to screen and verify suspected cases of M. ulcerans infection.
Programme documents, the SF and KIIs revealed that the WHO-recommended antimicrobial (rifampicin and streptomycin) therapy was introduced at the beginning of the BUPaT programme in 2005, and administered to all patients. Health staff were trained in the appropriate protocols to be observed when administering these antibiotics. By policy, BU treatment is covered under the National Health Insurance Scheme (NHIS). Narratives from the SF and the KIIs indicated that these antibiotics which are anti-tuberculosis drugs were provided by the NBUCP. Medicines and dressings were provided by the Ministry of Health through the NBUCP and sometimes by World Vision Ghana when stocks were exhausted. The SF forum also mentioned that surgery was carried out at least once a week at the AH by a surgical team from the Korle-Bu teaching hospital. Documents highlighted the infrastructural limitations of the OHC and the Kojo Ashong clinic that made it impossible for surgical operations to be carried out there.
At the Kojo Ashong clinic, located 20 kilometres from the AH, in an endemic community, BU care was limited to antibiotic therapy. At the time of the research team's visit, 4 patients had been registered: 2 female adults and 2 male children. During the visit, the team observed treatment of the children and 1 adult. The children proceeded to school after treatment. In addition to antibiotic care, the OHC performs minor excisions; patients requiring major surgery are referred to the AH. At the time of the team's visit, 9 persons (6 children and 3 adults) had already received treatment, though clinic records indicated that 24 patients (15 children and 9 adults) had been registered. Patient records also showed that only those 9 registered patients had regular treatment and they lived close to the OHC. Although rehabilitation of patients with disabilities is an integral component of BU care, all key informants admitted that this did not feature on the programme's agenda for lack of capacity and infrastructure. One key informant explained:
The NBUCP trained all health staff on the appropriate use of the stipulated WHO BU01 forms to record patient information, disease outcomes, and clinical and surgical procedures. Analysed data from these forms indicate that 297 patients were treated from June 2005 to June 2008. Children below 15 years constituted nearly half (146; 49%) of all admissions over the 3-year period. Patients presenting with ulcers formed the majority of all clinical forms: 52 (52.5%) in the first, 62 (73%) in the second and 67 (59.3 %) in the third yearly periods. There were 14 (14%) patients with recurring lesions (June 2005-May 2006) and none during the latter yearly periods (Table 2).
Except for the last yearly period (June 2007-May 2008) where only 34.5 % of patients healed without deformities, more than 60 percent of patients healed without deformities for the first and second years (Table 3). The proportion of patients that reported early and therefore were given only antibiotic treatment over the programme period was encouraging, 35.4% in the first yearly period, 23.5% in the second yearly period and 42.5%, in the third. The programme recorded 4 BU-related deaths (Table 3). Utilisation of services for BU increased over the three-year period. Of the 297 BU patients treated during this period, 113 were treated in year 3 (38.0%) compared with 85 (28.6%) in year 2, and 99 (33.3%) in year 1 (Table 3). Irrespective of these achievements, a significant proportion of patients either absconded treatment, or were lost to follow-up (14.1% in the first yearly period 9.4% in the second yearly period and 14.2%, in the third) (Table 3).
WVG provided cash incentives to plastic surgeons to ensure continuity of surgical operations. It was apparent from programme documents that the BUPaT programme supported in-patients and in some cases relations or caregivers with two meals (breakfast and lunch). Other organisations and individuals within and outside the municipalities also contributed towards feeding of patients either through cash donations or food items. All transport costs of patients and accompanying CBSVs to the AH, OHC, and patients who were referred to Korle-Bu hospital for specialised care were reimbursed. Key informants remarked that although feeding and refund of transport costs was not considered in the original programme design, it had to be incorporated later taking into consideration the poverty of programme beneficiaries, and remarked that good nutrition enhanced the healing of wounds.
All 3 key informants and stakeholders highlighted the high costs of treatment which placed a huge strain on the limited health budgets of the municipalities. They perceived a major difficulty in sustaining the programme if World Vision Ghana withdrew its financial support especially in the absence of government budgetary funding.
Among the contributions of the BUPaT programme to BU control, the following achievements are notable: improved collaboration among stakeholders, early case detection and treatment, increased community awareness of the priority of BU and improved access to treatment. Promoting awareness and access to improved services has made it possible to minimise surgical interventions, which the earlier programme had relied on almost exclusively.
The priority of early detection and treatment highlighted in programme documents (quarterly and annual reports), was consistent with accounts in the SF, KIIs and FGDs. FGD participants commended the community and school health education programmes, use of media especially documentary films and the efforts of the CBSVs. Participants regarded these strategies as helpful for increasing their awareness, promoting disease surveillance and encouraging early presentation of affected persons for treatment. A participant at the SF summarised the achievements of the programme as follows:
Our three key informants asserted the primary success of the BUPaT programme in managing BU was best indicated by the increasing number of patients receiving treatment at the AH over the course of the programme period. Statistics from the Ga-West municipality showed that prior to establishing the programme there were 70 cases in 2001, 82 in 2002, 83 in 2003 and 71 in 2004 [14]. In 2005, when the BUPaT programme commenced, AH recorded 99 cases and the number increased to 113 in 2008 over the 3-year period of the programme.
Before the BUPaT programme, surgery and wound care had been the only available treatment interventions. Improved outcomes of antibiotic therapy have been highly valued by key informants and stakeholders, who regarded it as a breakthrough. Antibiotic treatment has been appreciated because it has minimised recurrence of lesions, which was not possible under the old treatment regime. FGD participants also valued the effect of antibiotic therapy in shrinking lesions and removing necrotic tissue (Figure 1). They made no mention of any negative side-effects of this treatment.
Despite the achievements of the programme, stakeholders and key informants mentioned some major challenges: the inadequacy of ward space to accommodate affected persons who required surgery, the lack of requisite infrastructure in other municipal health centres to perform surgery and the limited health budgets of municipalities. Another challenge was the delay of some affected persons in seeking medical treatment. One stakeholder commented on the challenge of the AH as the main referral and treatment facility for BU as follows:
FGD participants mentioned fears of amputation, loss of livelihoods and the inevitable long absence of the primary care-giver from the home (mostly the mother), when a child is on admission at the hospital, as reasons for delayed treatment. They also expressed concern about feeding (the programme provided two meals a day), transport costs (transport costs of care-givers paying repeated visits to children on admission were not refunded) and difficulties with the continuation of medical treatment if support for feeding and transport was withdrawn. A mother of a treated child explained:
However, other explanations for delayed medical treatment were linked to misinformation from THs on the likelihood of amputation with medical treatment. Some THs also tried to convince affected persons that herbal treatment was more effective than medical care. Stakeholders indicated that at the beginning of the programme, THs were trained to identify and refer promptly, all cases of BU that were brought to their attention, for appropriate treatment but they acted contrariwise.
FGD participants also expressed difficulties in early diagnosis of their conditions as BU, because of the various presentations of BU infection. For many, it was difficult to know whether cuts, stings, scratches and abrasions were uncomplicated injuries or the beginning of the BU disease. In most cases, these were either unnoticed or dismissed as trivial. As the condition progressed, an assortment of remedies including herbs, balms and hot compresses were applied until BU infection was established; in some cases, after the affected part opened up (revealing the necrotic tissue).
Stakeholders regarded collaboration, networking and the community-based surveillance system as vital components of the BUPaT programme that had to be sustained. Stakeholders and key informants also mentioned the need to equip existing clinics to serve as treatment centres for wound care and antibiotic treatment. This was considered important to improve access to treatment and reduce severity of reported cases and disabilities, thus reducing the cost burden to the health system.
FGD participants agreed that health education and community-based surveillance activities should continue to increase awareness, improve case detection and encourage early reporting. They also implored the programme to continue to defray transport costs to lessen the economic burden of the disease.
The primary goal of the BUPaT programme was to reduce BU-related suffering and disability through early detection and treatment of cases. Using a mixed method approach, study findings showed the contribution of the health system to BU control in an endemic area in Ghana. Extensive collaboration existed across all levels, (national, municipality and community), which contributed to strengthening the health system. The programme strengthened capacities of health staff in antibiotic treatment and wound care, and trained teachers, MEHOs and CBSVs in health education, screening, early detection and prompt referral for medical treatment. A patient database was also created using recommended WHO forms. WHO-recommended antibiotics improved treatment and cure, particularly for early lesions, thus preventing recurrences. Providing feeding and refund of transport costs proved a useful strategy in encouraging medical care. Irrespective of these achievements, there were still problems of access, accommodation (lack of sufficient ward space), use of traditional treatment, loss to follow-up and non-adherence to treatment.
The broader impact of the BUPaT programme on the health system could be seen in its effects on some of the six building blocks, or subsystems, of the health system, but not on others. With reference to the WHO framework [1], the programme mainly affected governance, human resources, medicines and technology, and health delivery; it had less impact on the financing and information systems. Collaboration and networking among stakeholders strengthened the governance sub-system and improved health delivery of the programme. Training different groups of stakeholders - namely, health staff, CBSVs, MEHOs, teachers and THs - enhanced the human resource sub-system. The administration of WHO-recommended antibiotics improved treatment outcomes and revolutionised the medicines and technologies sub-system. Each of these subsystems contributed to improved health delivery. Minimising expensive surgery by promoting alternative interventions reduced the strain on the limited resources of the finance sub-system. Although the BUPaT programme now routinely compiles patient data using WHO-recommended forms in an electronic database, community epidemiological data are needed for an integrated data system based on community surveillance.
Patient data showed that a significant proportion of admissions comprise children under 15 years-of-age (49 %), consistent with other study findings on the susceptibility of children to BU infection [15], [16]. Even though most cases of BU were not confirmed by laboratory tests, all cases were diagnosed by qualified health staff and surgeons on the basis of WHO clinical case definitions [17]. The BUPaT project aimed to improve early case detection, particularly for nodules, plaques and oedemas, though patient data showed the proportion of patients with pre-ulcer conditions remained less than for ulcer patients. Stakeholders argued that this was not a failure of the programme, however, because people with ulcers who would not previously have used the health system were now seeking medical care instead of remaining with THs.
Consequently, improved awareness has led to treatment of more patients with both pre-ulcerative conditions and ulcers. The reluctance of some people with BU to seek medical care is consistent with findings of other studies [18]–[20]. Studies suggest that the socio-economic impact of BU is a determining factor in the choice of treatment and adherence to medical treatment [6], [7]. Traditional therapy has been the first choice for treatment for some affected persons because of easy local access, compared with the burden of high transport costs, and loss of income due to absence from work while in medical treatment at a distant site [5], [19], [20].
Although increasing community awareness has been bringing more patients to medical treatment, FGDs also showed that various presentations (cuts, bites, stings and abrasions) were not identified as a possible indication of M. Ulcerans infection that would benefit from treatment. The effectiveness of antibiotics in preventing recurrences was documented in the patient data. Narratives from stakeholders and key informants referred to this, and they also indicated satisfaction with the minimal cost of antibiotic treatment compared with the high cost of surgery. These findings are consistent with other studies on drug effectiveness [2], [21], [22]. Even though there were no recurrent infections as observed previously when surgery was the only treatment procedure, a significant proportion of patients healed with deformities, most of these patients had ulcers. To minimise deformities, post-operative health care and physiotherapy is required and prosthesis would be needed for amputees. The cost of these services is indeed enormous for an already burdened and poorly resourced rural health service [2], [7]. WHO recommends the need for rehabilitation of patients [23], yet there is paucity of research on its success and integration in the health system.
Based on our study findings, we offer recommendations for effective BU control, particularly for poorly resourced rural health systems. These include health education and community surveillance, collaboration with research laboratories for confirmation of cases, improving access to antibiotic treatment and wound care, integrating BU care with the management of similar diseases and disease mapping:
Our findings show the tremendous impact of health education and community surveillance strategies in BU control. Though this is a laudable community-directed initiative, there is the need for more concerted efforts of the programme to intensify these strategies to reduce BU-related morbidity and increase timely access to medical treatment. All teachers should be trained to identify all forms of M. ulcerans infection and refer for medical treatment. School children and others in the community should be encouraged to identify and report suspected cases to teachers, school authorities and community-based surveillance volunteers for verification. Local political commitment is needed by involving chiefs, traditional and religious leaders to support these efforts.
Health education messages should not only focus on creating awareness. They should also emphasise the importance of early reporting and appropriate care to avoid disease sequalae. Messages should encourage affected persons to seek early medical treatment for cuts, abrasions, stings or suspicious swellings. They should correct local ideas about the cause of BU that may discourage appropriate help-seeking. In this regard, it is important that all suspicious pre-ulcerative lesions should be evaluated with laboratory tests. WHO recommends a polymerase chain reaction (PCR) test to confirm cases and diagnosis. Results of this test can be obtained in two days [9]. Given the absence of infrastructure and expertise to perform such analyses, the health system could benefit from collaboration with research laboratories and institutions.
The Ga-West municipality has opened health centres in a few localities to make chemotherapy accessible but these have proven woefully inadequate. There are quite a number of private clinics and maternity homes in both municipalities managed by qualified health personnel who have a large clientele. Integrating them in the health system could boost coverage and access to chemotherapy. The municipal health directorates should assume a supervisory and monitoring role to ensure compliance to case management and chemotherapy protocols.
The cost of managing BU like any other neglected tropical disease is enormous and places a huge strain on a limited rural health budget. Cost-effective interventions should aim at integrating diseases of similar characteristics. Since tuberculosis (TB) case management relies on the Directly Observed Treatment Strategy, all TB centres in the study municipalities could serve as referral treatment centres for identified cases of M. ulcerans infection.
Understanding the demographics, epidemiology and geographical distribution of areas that require interventions is critical for cost-effective BU control. The disease is known to be endemic in riverine communities and is attributed to a myriad of factors that include direct exposure to water and swampy areas [24], [25]. These features and documented cases could serve as indices for classifying communities into three categories: priority-endemic areas, requiring the most interventions, endemic and non-endemic, requiring further research to enhance understanding of the disease. First, basic demographic knowledge of all communities must be documented, updated periodically and entered into a central database that will enable mapping and tracking of cases. This is a task for which spatial analytic research is needed.
Findings demonstrate the role of extensive health education, community-based surveillance, capacity building and collaboration among stakeholders for BU disease control. Treatment with the administration of WHO-recommended antimicrobials has proven effective at least for early lesions. Threats to livelihoods and feeding and transport expenses influence delay to seeking medical care. Findings also indicate the need for an integrated health service delivery approach by incorporating diseases requiring similar antibiotic treatment regimes. A further step towards integration will be to include private health-care providers in the health system to increase access to antibiotic therapy in close proximity to the population. Health education is required in this regard to emphasise the effectiveness of treatment with antibiotics to reduce disease sequalae and the importance of seeking medical treatment for all skin lesions, whether big or small. Evidence from this study suggests that intensifying health education and surveillance would ultimately improve access to treatment for all cases. Further research is needed to explain the role of environmental factors for BU contagion. Health service delivery strategies reported in our study can be adopted for any BU-endemic area in Ghana.
|
10.1371/journal.ppat.1005591 | Neutrophil and Alveolar Macrophage-Mediated Innate Immune Control of Legionella pneumophila Lung Infection via TNF and ROS | Legionella pneumophila is a facultative intracellular bacterium that lives in aquatic environments where it parasitizes amoeba. However, upon inhalation of contaminated aerosols it can infect and replicate in human alveolar macrophages, which can result in Legionnaires’ disease, a severe form of pneumonia. Upon experimental airway infection of mice, L. pneumophila is rapidly controlled by innate immune mechanisms. Here we identified, on a cell-type specific level, the key innate effector functions responsible for rapid control of infection. In addition to the well-characterized NLRC4-NAIP5 flagellin recognition pathway, tumor necrosis factor (TNF) and reactive oxygen species (ROS) are also essential for effective innate immune control of L. pneumophila. While ROS are essential for the bactericidal activity of neutrophils, alveolar macrophages (AM) rely on neutrophil and monocyte-derived TNF signaling via TNFR1 to restrict bacterial replication. This TNF-mediated antibacterial mechanism depends on the acidification of lysosomes and their fusion with L. pneumophila containing vacuoles (LCVs), as well as caspases with a minor contribution from cysteine-type cathepsins or calpains, and is independent of NLRC4, caspase-1, caspase-11 and NOX2. This study highlights the differential utilization of innate effector pathways to curtail intracellular bacterial replication in specific host cells upon L. pneumophila airway infection.
| Legionella pneumophila is a motile gram-negative bacterium found mainly in fresh water environments where it replicates in amoeba. It uses a molecular syringe to inject effector molecules into these predatory host cells, reprograming them to support L. pneumophila growth. Upon inhalation of contaminated aerosols, L. pneumophila uses the same approach to replicate in human alveolar macrophages, which can result in a severe pneumonia known as Legionnaires’ disease. However, L. pneumophila is normally controlled by the innate immune system, and the key mechanisms and cells involved in this immune response remain unclear. Here we show that tumor necrosis factor (TNF) and reactive oxygen species (ROS) play a dominant role in the clearance of L. pneumophila from the lung. Neutrophils kill L. pneumophila using ROS, while alveolar macrophages are activated by TNF produced by neutrophils and monocytes that are recruited to the lung. TNF-activated alveolar macrophages kill L. pneumophila by recruiting lysosomes and acidifying L. pneumophila containing vacuoles. Caspases other than caspase-1 and 11 are involved in this mechanism, with a minor contribution from cysteine-type cathepsins or calpains. This study deepens our understanding of the mechanisms by which TNF contributes to the control of intracellular pathogens, and highlights the key elements of the innate immune response to L. pneumophila lung infection.
| L. pneumophila is a Gram-negative bacterium with global distribution in freshwater environments, where it replicates intracellularly mainly in amoebae [1–3]. L. pneumophila commonly causes community acquired and nosocomial pneumonia. Although it is normally controlled by the innate immune response, L. pneumophila has the potential to cause a severe pneumonia known as Legionnaires' disease with mortality rates of up to 30% if early bacterial replication is not controlled [4–6]. Infection occurs through inhalation of L. pneumophila contaminated aerosols, mostly generated by manmade technologies such as cooling towers, air conditioners or even car windshield wipers [7–9]. In the lung L. pneumophila initially exclusively infects alveolar macrophages (AM), using a type IV secretion system (T4SS) to inject over 300 effector proteins into the cytosol [7,10–12]. These effectors block phagosomal maturation and fusion with lysosomes, thus preventing L. pneumophila degradation, and promoting the establishment of a Legionella containing vacuole (LCV), the intracellular niche in which L. pneumophila replicates [13–16].
Though critical for L. pneumophila replication, the T4SS also potently induces the innate immune response by several mechanisms (reviewed in [17]). AM sense the action of the T4SS and respond by secreting IL-1α, inducing the secretion of chemokines by airway epithelial cells (AECs), resulting in the rapid recruitment of neutrophils and monocytes to the lung [10,18,19]. Neutrophils are known to be critical for the clearance of L. pneumophila lung infection, as evidenced by neutrophil depletion studies [18,20,21], in vivo blockade of CXCR2 [22] and studies examining the role of IL1R signaling [18,19,23]. However, the mechanisms by which neutrophils contribute to the resolution of L. pneumophila lung infection remain incompletely understood.
IL-1 is closely linked to the induction of TNF in a broad spectrum of unrelated models of inflammation, and these cytokines are known to have synergistic effects in vivo [24–26]. Indeed, anti-TNF therapy is a recognized risk factor for Legionnaire's disease, suggesting a role for TNF in the immune response to L. pneumophila [27–31]. Previous work has established that TNF is produced in response to L. pneumophila in a T4SS-dependent and flagellin-independent manner [32,33] and can limit replication in macrophages [34–36]. Furthermore, it was shown that TNF contributes to immune defense against L. pneumophila in vivo [37–39]. However, the mechanisms by which TNF contributes to innate immune control of L. pneumophila and the cells upon which it acts in vivo have yet to be elucidated.
Macrophages from C57BL/6 mice are not permissive for L. pneumophila replication due to the intracellular sensor NAIP5 which binds cytosolic flagellin and recruits NLRC4, resulting in inflammasome assembly and the activation of Caspase-1 [40,41]. Active caspase-1 can initiate a pro-inflammatory form of cell death known as pyroptosis, the secretion of IL-1β and IL-18, as well as activate Caspase-7, which induces the fusion of lysosomes with LCVs, resulting in bacterial degradation [42,43]. Murine macrophages missing key components in this pathway are permissive to L. pneumophila replication, including NAIP5-/-, NLRC4-/-, Caspase-1-/- and Caspase-7-/- macrophages [42]. NLRC4 also restricts L. pneumophila via caspase-1 independent mechanisms [44]. Similarly, it has been shown that human NAIP (hNAIP), the only NAIP protein identified in humans, can mediate inflammasome assembly and L. pneumophila restriction when overexpressed in murine macrophages, and that L. pneumophila replication is enhanced in human macrophages when hNAIP is silenced [45,46]. Furthermore, primary human macrophages sense L. pneumophila flagellin via hNAIP and activate caspase-1 [47,48].
Macrophages from A/J mice are permissive to L. pneumophila replication due to an allelic variation in the NAIP5 gene, resulting in 14 amino acid (aa) differences as compared to C57BL/6 mice [49,50]. A/J macrophages are able to activate Caspase-1 in response to L. pneumophila infection [51], but fail to activate caspase-7, suggesting that at least some of the 14 aa are involved in promoting caspase-1 and caspase-7 interactions [40,42]. Other mouse strains also display partial susceptibility to L. pneumophila infection and replication, including FvB/N, C3H/HeJ, BALB/cJ and 129S1 mice [49]. In this paper we make use of mice with the 129S1 NAIP5 allele (NAIP5129S1) that have a targeted TNF deletion in macrophages, monocytes and neutrophils (MN-TNF NAIP5129S1 mice) [52] to examine the role of TNF derived from macrophages, monocytes and neutrophils in L. pneumophila lung infection in the absence of strong NAIP5 signaling.
In the present study, we demonstrate that TNF and reactive oxygen species (ROS) are essential for the effective innate immune control of L. pneumophila, and that in vivo TNF can compensate for the well characterized NLRC4-NAIP5 flagellin pathway. While ROS are essential for the bactericidal activity of neutrophils, TNF produced by neutrophils and monocytes is required to enhance AM-mediated restriction of L. pneumophila via TNFR1 in vivo. This TNF-mediated antibacterial mechanism is independent of NLRC4, caspase-1 and 11, but involves other caspases with a minor contribution from cysteine-type cathepsins or calpains, and also the fusion of LCVs with lysosomes and their acidification. The striking susceptibility of MN-TNF NAIP5129S1 mice to L. pneumophila lung infection suggests that TNF is a key component of innate immunity to L. pneumophila lung infection, especially when NAIP5-NLRC4 mediated responses are dampened.
Many host immune factors have been shown to be involved in L. pneumophila control in vitro, whereas relatively few studies have assessed their impact in vivo. We therefore used an intranasal mouse infection model to identify crucial innate immune effector molecules and pathways that have been implicated in the clearance of L. pneumophila lung infection, by assessing their relative impact on bacterial burden in the lung 3–7 days p.i.. As has been previously demonstrated, we found that while IFNγR-/- and IFNAR-/- mice showed limited susceptibility to infection, double deficiency for IFNAR/IFNγR dramatically increased bacterial loads, in particular by day 7 post infection (Fig 1A and 1C, [53]). Similarly, by day 5 and 7 p.i., TNF deficiency resulted in severely increased bacterial burden, and deficiency in the phagocyte NADPH oxidase NOX2/gp91phox (CYBB-/- mice) resulted in potent impairment in bacterial control from day 3 through to day 7 (Fig 1A and 1C). In contrast, NLRC4, caspase-1/11, TLR5, IL-12, iNOS and IL17RA seem to play a less dominant role in controlling L. pneumophila lung infection (Fig 1A). These results show that TNF and ROS, as well as the combined action of Type I and II IFN signaling are crucial for the innate immune response to L. pneumophila lung infection.
To identify the receptor through which TNF exerts its protective effect, WT, TNF-/-, TNFR1-/-, TNFR2-/- and TNFR1/2-/- mice were infected intranasally with WT L. pneumophila and CFUs were compared in the BALF 5 days p.i.. Bacterial clearance was delayed to a similar extent in TNF-/-, TNFR1-/- and TNFR1/2-/- but not TNFR2-/- mice compared to WT mice, showing that TNF mediates its antibacterial effect via TNFR1 in vivo (Fig 1B).
A recent study using a T4SS-based reporter system has demonstrated that AM and neutrophils are the primary targets for L. pneumophila in vivo, with L. pneumophila replication having been demonstrated in AM [10]. We therefore examined the impact of TNF on AM and neutrophil-mediated killing of L. pneumophila in vivo. To circumvent the problem that TNFR1-/- mice have greater bacterial burdens in the lung than WT mice and allow for the direct comparison of AM and neutrophil bacterial loads in WT and TNFR1-/- cells within a single mouse, we used a mixed chimera approach. Mixed bone marrow (BM) chimeric mice were generated with a mix of 50% Ly5.1+ WT BM and either 50% Ly5.2+ WT or Ly5.2+ TNFR1-/- BM. After 8 weeks of reconstitution, WT:WT and WT:TNFR1-/- mice were inoculated intranasally with WT L. pneumophila, and 2 days p.i. Ly5.1+ and Ly5.2+ AM and neutrophils were sorted from the BALF, and cells were plated on CYE plates to quantify viable L. pneumophila. Significantly more CFU / AM were recovered from TNFR1-/- AM than from WT AM, indicating that TNF signaling via TNFR1 promotes the killing of L. pneumophila by AM in vivo (Fig 2A and 2B). In contrast, there was no difference in the number of viable L. pneumophila / neutrophil recovered from WT vs. TNFR1-/- neutrophils, indicating that TNF signaling does not contribute to neutrophil-mediated killing of L. pneumophila (Fig 2A). The killing of L. pneumophila lacking flagellin was also impaired in TNFR1-/- AM compared to WT AM, demonstrating that the antibacterial mechanism mediated in AM by TNF / TNFR1 is independent of the NAIP5-NLRC4 flagellin recognition pathway (Fig 2B). These results highlight that TNF / TNFR1 signaling mediates a non-redundant antibacterial mechanism that contributes to L. pneumophila killing in AM but not in neutrophils in vivo.
To analyze the impact of ROS on AM and neutrophil-mediated killing of L. pneumophila, we generated BM chimeric mice with a mix of 50% Ly5.1+ WT BM and either 50% WT Ly5.2+ or Ly5.2+ CYBB-/- BM. 2 days p.i. we observed that while sorted CYBB-/- AM did not contain more viable L. pneumophila / AM than WT AM, sorted CYBB-/- neutrophils contained more viable L. pneumophila / neutrophil than did WT neutrophils from the same mouse (Fig 2A). This indicates that in contrast to TNF, NOX2-derived ROS play a non-redundant role in neutrophil-mediated killing of L. pneumophila but not AM-mediated killing of L. pneumophila in vivo.
We performed similar experiments in which WT:WT and WT:CYBB-/- BM chimeric mice were inoculated with either L. pneumophila constitutively expressing GFP (Lpn-GFP), or with L. pneumophila containing a plasmid on which GFP expression can be induced by the addition of IPTG (Lpn-GFPind), thereby identifying metabolically active bacteria (Fig 2C). Neutrophils were analyzed by flow cytometry 38 hours p.i., and in the case of Lpn -GFPind infected mice, IPTG was administered intranasally at 35 hours p.i., resulting in the induction of GFP in all viable L. pneumophila. In line with the results of the BM chimera sort and plating experiments, there were more GFP+ CYBB-/- neutrophils than GFP+ WT neutrophils in WT:CYBB-/- BM chimeric mice, both with Lpn-GFP infection and with Lpn-GFPind infection (Fig 2C). In the case of Lpn-GFP infection this indicates that there were more NOX2-deficient neutrophils that contained dead or viable L. pneumophila than WT neutrophils, and in the case of Lpn-GFPind infection it indicates that there were more NOX2-deficient neutrophils that contained viable L. pneumophila than WT neutrophils in the same mouse. These data support the hypothesis that neutrophils require ROS to kill and degrade L. pneumophila in vivo.
Having established that NOX2-dependent mechanisms are involved in neutrophil-mediated killing of L. pneumophila, we sought to determine if neutrophils actively produce ROS in response to L. pneumophila. We infected WT and CYBB-/- mice with WT, T4SS deficient (ΔT) and ΔFlaA L. pneumophila and stained neutrophils and AM with a flow cytometry based ROS detection reagent (Dihydroethidium) 24 h p.i.. We observed that neutrophils but not AM produced ROS in response to WT and ΔFlaA L. pneumophila 24 h p.i., suggesting that ROS could have direct bactericidal effects in L. pneumophila containing neutrophils (Fig 3A). Since we did not observe neutrophil ROS production in response to ΔT L. pneumophila, our results suggest this ROS production is T4SS-dependent and flagellin independent (Fig 3A). Conversely, AM produced very little ROS in response to WT L. pneumophila, but more in response to ΔT L. pneumophila, in line with a publication suggesting that L. pneumophila actively inhibits ROS production in macrophages via T4SS-dependent effector molecules (Fig 3A, [54]).
The in vivo results presented in Fig 2A in combination with the observation that in vitro, TNFR1-/- and TNFR1/2-/- but not TNFR2-/- bone marrow derived macrophages (BMDM) were more permissive to L. pneumophila replication than WT BMDM, suggest that TNF directly inhibits L. pneumophila replication in macrophages via signaling through TNFR1 (Fig 4A). Furthermore, the addition of recombinant TNF (rTNF) to BMDM abrogated L. pneumophila growth in all of the genotypes with a functional TNFR1, including NLRC4-/- and CYBB-/- BMDM (Fig 4A). These data show that TNF-mediates an antibacterial mechanism in BMDM via TNFR1, which is independent of NOX2-derived ROS and the NAIP5-NLRC4 flagellin recognition pathway. The flagellin independence of this mechanism was further shown by the TNF-mediated abrogation of ΔFlaA L. pneumophila replication in WT BMDM but not TNFR1-/- BMDM (S3B Fig). Importantly, three day exposure to 100 ng/ml rTNF did not induce BMDM cell death (S1 Fig), suggesting an active antibacterial mechanism mediated by TNF rather than the induction of cell death. Membrane TNF knock-in (memTNF KI) BMDM, which are only able to make membrane bound but not secreted TNF, were also more susceptible than WT BMDM, suggesting that TNF signals as a soluble molecule on BMDM in vitro (Fig 4A). To consolidate this observation, we added a neutralizing anti-TNF antibody or TNFR1 fused to the Fc portion of human IgG1 (TNFR1-Fc) to WT BMDM infected with L. pneumophila, in order to neutralize soluble TNF secreted by the BMDM. This resulted in the sensitization of WT BMDM to L. pneumophila infection to a similar level as that observed for TNFR1-/- BMDM, suggesting that the difference in susceptibility between WT and TNFR1-/- BMDM is due to endogenously secreted TNF in response to L. pneumophila infection (Fig 4A and 4C). Also in line with the conclusion that lack of endogenous TNF results in moderate sensitivity of BMDM to L. pneumophila infection is the observation that MyD88-/- BMDM, which fail to secrete TNF in response to L. pneumophila infection (S2 Fig and [36]), also have a similar susceptibility to L. pneumophila as TNFR1-/- BMDM (Fig 4A). Taken together, these data suggest that TNF activates an antibacterial mechanism in macrophages via TNFR1 that is independent of NLRC4 and NOX2. Furthermore, TNF production by BMDM in response to L. pneumophila is downstream of MyD88.
In order to determine which cells produce TNF in vivo, we infected WT mice and TNF-/- mice with WT, ΔT and ΔFlaA L. pneumophila and stained BALF cells for TNF 30 hr p.i.. We found that neutrophils and monocytes produced TNF in response to L. pneumophila lung infection, suggesting that neutrophils and monocytes are the relevant TNF source (Fig 3B).
As has been shown in published results, we observed that NLRC4-/- mice were only moderately susceptible to infection, despite the well-recognized role of NLRC4 in inflammasome activation in response to L. pneumophila flagellin, and the high susceptibility of NLRC4-/- macrophages to L. pneumophila replication in vitro (Figs 1A and 4B, [55], [44]). The fact that NLRC4-/- BMDM are highly susceptible to L. pneumophila replication in vitro, but NLRC4-/- mice are only moderately susceptible in vivo, suggests that mechanisms that are only present in vivo are able to compensate for a lack of NLRC4. To determine if paracrine TNF compensates for a lack of NAIP5-NLRC4-mediated signaling in vivo, we infected MN-TNF NAIP5129S1 mice, which have a hypofunctional NAIP5 allele (NAIP5129S1) and are deficient in TNF in macrophages, monocytes and neutrophils, with WT L. pneumophila. We found that MN-TNF NAIP5129S1 mice were highly susceptible to L. pneumophila lung infection, with much greater bacterial burdens in the BALF 5 days p.i. compared to WT mice, and also compared to TNF-/- and NLRC4-/- mice (Figs 1A and 5B). Taken together, these results suggests that TNF produced by neutrophils and monocytes is essential for in vivo control of L. pneumophila lung infection. In addition, BMDM from MN-TNF NAIP5129S1 mice were almost as susceptible to L. pneumophila replication as NLRC4-/- BMDM, and this susceptibility could be abrogated by the addition of rTNF (Fig 4B). Taken together, these data suggest that neutrophil and monocyte derived TNF enhances AM-mediated L. pneumophila killing and partially compensates for a lack of NAIP5-NLRC4 signaling in AM in vivo.
To verify that TNF is indeed abrogated in MN-TNF NAIP5129S1 mice, we infected WT and MN-TNF NAIP5129S1 mice intranasally with WT L. pneumophila, and measured TNF in the BALF (Fig 5A). MN-TNF NAIP5129S1 mice had almost undetectable TNF in the BALF following intranasal L. pneumophila infection, confirming that in the context of intranasal L. pneumophila lung infection, neutrophils and monocytes are the primary source of TNF.
Since 129 mice have a documented mutation in caspase-11 [56], we sequenced this gene in MN-TNF NAIP5129S1 mice and found it to be WT. To rule out the possibility that further genes besides NAIP5 from the 129S1 genetic background influenced the phenotype of MN-TNF NAIP5129S1 mice in our experiments, we backcrossed them to C57BL/6 mice, and used the F2 offspring to conduct littermate controlled experiments. All the offspring we used for experiments were positive for MLys-Cre and the NAIP5 locus was sequenced for each individual mouse. We compared in vitro L. pneumophila replication in BMDM derived from the homozygous F2 offspring (TNF+/+/NAIP5B6, TNF+/+/NAIP5129S1, TNFfl/fl/NAIP5B6, TNFfl/fl/NAIP5129S1), C57BL/6 (WT), TNF-/- and MN-TNF NAIP5129S1 mice as well as in vivo bacterial loads in the BALF 5 days after intranasal L. pneumophila infection (Fig 5C and 5D). We observed that BMDM from F2 offspring that were TNF sufficient and carried the NAIP5B6 allele were as resistant to WT L. pneumophila infection as WT BMDM, indicating that these two genes were responsible for the enhanced susceptibility of MN-TNF NAIP5129S1 BMDM (Fig 5C). Furthermore, TNF+/+/NAIP5129S1 BMDM supported more L. pneumophila growth than did TNF+/+/NAIP5B6 BMDM, though this was not statistically significant, and TNFfl/fl/NAIP5129S1 BMDM were as susceptible as MN-TNF NAIP5129S1 BMDM (Fig 5C). These data suggest that MN-TNF NAIP5129S1 BMDM are more susceptible to L. pneumophila infection than WT BMDM due to defects in both NAIP5 signaling and TNF production.
In vivo, we found that TNF+/+ littermates with NAIP5129S1 were not more susceptible than TNF+/+ littermates with NAIP5B6, which is in line with our previous findings and published data indicating that reduced NAIP5-NLRC4 signaling has only a moderate impact on susceptibility to L. pneumophila lung infection in vivo (Fig 5D). In contrast, TNFfl/fl/NAIP5129S1 mice tended to have greater susceptibility to L. pneumophila lung infection, similar to MN-TNF NAIP5129S1 mice, while TNFfl/fl/NAIP5B6 mice had similar susceptibility to TNF-/- mice (Fig 5D). These data suggest that NAIP5129S1 and TNF deficiency in macrophages, monocytes and neutrophils are the genetic elements that mediate the enhanced susceptibility of MN-TNF NAIP5129S1 mice to L. pneumophila lung infection.
To gain further insight into the antibacterial mechanism mediated by TNF in macrophages, we infected MN-TNF NAIP5129S1 BMDM with Lpn-GFP, in the presence or absence of rTNF or rIFNγ as a positive control [57], and examined the fate of the LCV with respect to lysosomal fusion using confocal microscopy. By 3 hours p.i. neither 100 ng/ml rTNF nor 200 U/ml rIFNγ resulted in Lpn-GFP co-localization with lysosomal compartments as defined by lysotracker staining (Fig 6A). However, when MN-TNF NAIP5129S1 BMDM were pre-treated with rTNF or rIFNγ overnight, by 1 hour p.i. 50% of L. pneumophila in rTNF pre-treated MN-TNF NAIP5129S1 BMDM co-localized with lysosomal compartments, but not in rIFNγ pre-treated BMDM. By 3 hours p.i., Lpn-GFP co-localization with lysosomal compartments was observed in both rTNF and rIFNγ pre-treated MN-TNF NAIP5129S1 BMDM, suggesting that TNF induces the fusion of lysosomes with the LCV, but with different kinetics than IFNγ (Fig 6A). Furthermore, pre-treatment with rTNF was also shown to induce co-localization of ΔFlaA Lpn in WT BMDM but not TNFR1-/- BMDM after 1 hr, confirming TNF-mediated flagellin-independent induction of the fusion of lysosomes with LCVs in macrophages (Fig 6B).
Given that the fusion of LCVs with lysosomes has been shown to be induced by caspase-11, as well as caspase-1 in conjunction with caspase-7 [42,58], we wished to determine if the antibacterial effect mediated by TNF is dependent on Caspase-1 or 11. We therefore infected caspase-1/11-/- BMDM with L. pneumophila, with or without the addition of rTNF, and CFU were quantified 3 days p.i.. The addition of rTNF prevented bacterial replication in Caspase-1/11-/- BMDM, demonstrating that the TNF-mediated antibacterial mechanism in BMDM is independent of Caspase 1 and 11 (Fig 7A). Since our co-localization experiments suggested that TNF redirected L. pneumophila to lysosomal compartments, we sought to determine if lysosomal acidification was required for the TNF-mediated mechanism. To test this we infected Caspase-1/11-/- BMDM with L. pneumophila with or without rTNF and in the presence or absence of bafilomycin A1, a vacuolar H+-ATPase (v-ATPase) inhibitor that blocks lysosomal acidification. We found that bafilomycin A1 abrogates the TNF-mediated inhibition of L. pneumophila, suggesting that lysosomal acidification is required downstream of TNF (Fig 7A).
To determine if further caspases were involved, we tested if the TNF-mediated inhibition of L. pneumophila growth could be blocked using the third generation pan caspase inhibitor Q-VD-OPh, which is highly potent and specific for caspases [59]. We infected WT BMDM with ΔFlaA L. pneumophila, in the presence of increasing concentrations of Q-VD-OPh, with or without rTNF, and quantified CFU 3 days p.i.. Q-VD-OPh blocked TNF-mediated growth restriction in a dose dependent manner, suggesting that caspases other than caspase-1 and 11 are required for the TNF-mediated restriction of L. pneumophila growth in BMDM (Fig 7B). In summary, our data show that the antibacterial mechanism mediated by TNF in BMDM is dependent on at least one caspase and lysosomal acidification, but is independent of Caspase-1 and 11.
To gain further insight into the TNF-induced bactericidal mechanisms responsible for L. pneumophila degradation in acidic compartments, we investigated the involvement of lysosomal proteases in the cathepsin family. For this we titrated the inhibitor E-64d, which potently inhibits cysteine-type cathepsins and calpains but not caspases [60–64], in the presence or absence of rTNF on ΔFlaA L. pneumophila infected WT BMDM. We observed that E-64d modestly reduced TNF-mediated restriction of L. pneumophila growth only at high concentrations, suggesting that cathepsins or calpains are involved in the TNF-mediated restriction of L. pneumophila growth in BMDM, but play a minor role (Fig 8A).
Next, we attempted to identify individual cathepsins involved in the TNF-mediated mechanism. Cathepsin D inhibition using pepstatin A did not interfere with the TNF-mediated restriction of L. pneumophila replication, suggesting that this aspartic protease is not required (Fig 8B). In follow up of the experiments with cysteine protease inhibitors described above, we observed that the pan caspase inhibitor Z-VAD-FMK partially blocked the TNF-mediated restriction of L. pneumophila growth, both in WT and caspase-1/11-/- BMDM (S4 Fig). Since Z-VAD-FMK is also known to inhibit cathepsin B, H, L and S [59,60], we tested their involvement using BMDM from the corresponding cathepsin knockout mice. We found that rTNF suppressed L. pneumophila replication to a similar degree as in WT BMDM, showing that cathepsin B, H, L and S are not critical for TNF-mediated restriction of L. pneumophila growth in macrophages (Fig 8C). This, however, does not rule out that these cathepsins contribute redundantly to TNF-mediated inhibition of L. pneumophila growth in BMDMs.
To test if caspases and cathepsins / calpains have a synergistic role in TNF-mediated restriction of L. pneumophila replication, which could indicate that they are involved in separate converging pathways, we tested if the effect of E64d and Q-VD-OPh was additive. However, the concomitant addition of E-64d with Q-VD-OPh did not increase the ability of Q-VD-OPh to block the TNF-mediated effect (Fig 8D). These data suggest that caspases and cathepsins / calpains do not synergize to enhance TNF-mediated restriction of L. pneumophila growth, and could be involved interdependently in the same pathway. In conclusion, our data show that cathepsins or calpains contribute somewhat to TNF-mediated restriction of L. pneumophila replication in macrophages, but there is not a non-redundant requirement for cathepsin B, D, H, L or S. Therefore, among the proteases tested, the major cysteine proteases reducing the replication of L. pneumophila upon TNF treatment appear to be the caspases.
In this study, we identified cell-type specific key innate immune effector functions responsible for effective control of pulmonary L. pneumophila lung infection. Neutrophil-mediated mechanisms that lead to L. pneumophila clearance in vivo are twofold. On the one hand, neutrophils directly kill L. pneumophila via ROS-mediated mechanisms, and on the other hand, neutrophil and monocyte-derived TNF initiates microbicidal mechanisms in AM via TNFR1, which increase their capacity to inhibit L. pneumophila replication. The latter involves rerouting the bacteria to lysosomal compartments despite the presence of T4SS effectors, and requires at least one caspase other than caspase-1 or 11. The importance of TNF and NOX2-mediated mechanisms in the control of L. pneumophila infection are underscored by the marked susceptibility of TNF-/- and CYBB-/- mice to L. pneumophila infection.
The impact of TNF-mediated antimicrobial mechanisms directed against L. pneumophila cannot be fully appreciated by the study of macrophages in vitro. In accordance with other studies, we observed that TNFR1-/- BMDM only support moderate L. pneumophila growth in comparison to NAIP5-/- or NLRC4-/- BMDM which support several orders of magnitude more growth (Fig 4A and 4B, [36,55]). However, this difference is not observed when comparing the bacterial burden of TNFR1-/- and NLRC4-/- mice in vivo, where there is even a trend for TNF to play a more dominant role (Fig 1A). A possible explanation for these apparently incongruent results is that paracrine TNF produced in vivo by neutrophils and monocytes, rather than autocrine TNF produced by AM, mediates the increased resistance to L. pneumophila, and further that this TNF can compensate for a lack of NAIP5-NLRC4-mediated immune defense. Though we and others did observe modest endogenous TNF production by BMDM in response to L. pneumophila infection (S2 Fig, [36]), and found that this TNF accounted for the increased susceptibility of TNFR1-/- BMDMs (Fig 4C, [36]), this was not enough to compensate for lack of NAIP5-NLRC4 flagellin sensing (Figs 4B and S3B, [36]), arguing against a dominant role for autocrine TNF production by AM. In fact, NLRC4-/- BMDM were highly susceptible to infection despite secreting more TNF than WT BMDM in response to L. pneumophila infection, possibly due to increased bacterial burden or a failure to undergo pyroptosis (S2 Fig, [36]). However, we propose that in vivo, AM are exposed to much higher local concentrations of TNF than produced endogenously by BMDM. In vitro, 200–600 pg/ml TNF were observed in the supernatant of L. pneumophila infected WT BMDM (S2 Fig, [36]), in comparison to 1 ng/ml we observed in the BALF (Fig 5A) and up to 20 ng/ml reported in the BALF at peak concentration [65,66]. Making the conservative estimate of an epithelial lining fluid volume of 100 μl, or a 10–20 fold dilution in 1–2 ml BALF, the actual TNF concentration in the undiluted endothelial lining fluid would be 10–400 ng/ml. Indeed, the addition of 100 ng/ml rTNF markedly suppressed L. pneumophila replication in NLRC4-/- BMDM and increased cell viability (Figs 4B and S1). In addition, TNF has been shown to synergize with other cytokines such as IFNγ and type 1 interferons (IFN) in the restriction of L. pneumophila, which might also be present at higher concentrations in the epithelial lining fluid [34,36,53]. In line with this idea, bacterial burden is more severely impaired in TNF-/- and IFNAR/IFNγR-/- mice at later time points, which could reflect a shared mechanism of action (Fig 1B).
In order to verify the hypothesis that TNF might compensate for reduced NAIP5-NLRC4 mediated mechanisms, we made use of MN-TNF NAIP5129S1 mice, in which TNF is ablated in macrophages, monocytes and neutrophils and which carry the NAIP5129S1 allele. BMDM from MN-TNF NAIP5129S1 mice were almost as susceptible to L. pneumophila infection as BMDM from NLRC4-/- mice, as expected in the absence of strong NAIP5 signaling (Fig 4B). Strikingly, MN-TNF NAIP5129S1 mice were also much more susceptible to L. pneumophila infection in vivo compared to either NLRC4-/- or TNF-/- mice, which in combination with the intracellular staining results suggests that neutrophil and monocyte derived TNF compensates to a large degree for weak NAIP5-NLRC4 flagellin sensing in vivo (Figs 1A, 3B and 5B). Together with the observation that TNF is important for AM but not neutrophil-mediated killing, these experiments highlight the importance of TNF-mediated antibacterial mechanisms in AM in the context of L. pneumophila lung infection.
Our results indicating the functionally relevant production of TNF by neutrophils and monocytes are in agreement with a study by Copenhaver et al. [67]. However, there it was found that AM and DCs are also important for TNF production in response to L. pneumophila lung infection. In contrast, we do not observe significant TNF production by AM, which may reflect differences in the strains of bacteria used between the studies. Though we also observed TNF production by DCs, our results with MN-TNF NAIP5129S1 mice suggest that neutrophil / monocyte-derived TNF is physiologically more relevant for the innate immune response to L. pneumophila (Fig 5A, [52]).
In light of the finding that neutrophil and monocyte-derived TNF mediates an essential AM-driven immune response that can compensate for weak NAIP5-NLRC4-mediated immunity, it is interesting to note that ΔFlaA L. pneumophila is able to replicate in AM within the first 2 days p.i., after which bacteria are cleared [23]. These kinetics fit with the observations that TNF peaks in the BALF 2 days p.i. [39,65], that macrophages require pre-activation of around 20 hours with TNF before they become restrictive for L. pneumophila replication (Fig 6A, [36]), and that failure to recruit neutrophils to the lung from 12 hours up to around 2 days p.i. does not greatly impact bacterial burden, though these kinetics may vary with the size of the inoculum [19,23]. In addition, anti-TNF Ab treatment of A/J mice resulted in an increase in lung bacterial burden only as of around day 3 p.i. [39]. Also consistent with a need for neutrophil-derived TNF is the observation that clearance of ΔFlaA L. pneumophila is delayed to 72 hours p.i. in IL1R-/- mice, in which neutrophil recruitment is delayed, and that in MyD88-/- mice clearance is postponed to 6 days p.i., or even abrogated [23]. Since MyD88-/- BMDM fail to secrete TNF in response to L. pneumophila [36,68], and neutrophils secrete TNF in a flagellin-independent manner (Fig 3B, [10]), it seems highly likely that impaired TNF production by neutrophils and monocytes contributes to the striking susceptibility of MyD88-/- mice to L. pneumophila lung infection. The fact that AM do not produce much TNF in response to L. pneumophila infection but instead rely mostly on neutrophils and monocytes, which must first be recruited to the airways to produce TNF, likely reflects a mechanism which limits overzealous lung inflammation. Indeed, TNF is a very potent cytokine, and it's leakage from the airspace to the circulation can on its own strongly contribute to anaphylactic shock, as shown by systemic anti-TNF treatment in a rabbit model of Pseudomonas aeruginosa pneumonia [69]. Congruent with this idea, though neutrophils are essential for the resolution of L. pneumophila lung infection, they are also associated with lung pathology in Legionnaires' disease [70,71]. This may in part be due to their role in TNF secretion.
We also show that neutrophils kill L. pneumophila in the lung directly by NOX2-dependent mechanisms. Interestingly, AM do not produce ROS in response to WT L. pneumophila. This is in line with a study demonstrating that L. pneumophila actively represses ROS in AM by a T4SS-dependent mechanism [54] and our observation that AM produce ROS in response to ΔT but not much ROS in response to WT L. pneumophila (Fig 3B). Why this mechanism is not active in neutrophils remains unclear, given that both neutrophils and AM are targeted by the T4SS and harbor live L. pneumophila in vivo (Fig 2A, [10]). In fact, for neutrophils the opposite is true, as our results show that ROS induction in neutrophils is T4SS-dependent. On a similar note, a recent study has shown differential responses between macrophages and neutrophils to Salmonella flagellin, in that NAIP5-NLRC4 triggered pyroptosis in macrophages but not neutrophils [72]. How L. pneumophila adapts to these two different intracellular environments also remains unknown. The differential activation of neutrophils and AM by L. pneumophila will likely yield interesting insights into this host-pathogen interaction in future investigations.
In this study, we show that the TNF-mediated antibacterial mechanism in AM is dependent on the rerouting of L. pneumophila to lysosomal compartments, where they are degraded via processes that involve acidification. This acidification likely occurs early in the infection cycle, since fusion of LCVs and lysosomes can be observed within an hour of infection in BMDM pre-treated with TNF. Consistent with this view, a previous study found that L. pneumophila has at least one T4SS effector, SidK, which inhibits the v-ATPase [73]. SidK is highly induced when L. pneumophila begins a new growth cycle, presumably counteracting the early acidification of LCVs [73]. The observation that bafilomycin A1 alone reduced L. pneumophila replication in BMDM is expected, since L. pneumophila requires the acidification of the LCV in late stages of infection for proper LCV maturation [74].
Our data implicate the involvement of at least one caspase other than caspase-1 or 11 in the TNF-mediated growth-restriction of L. pneumophila in macrophages, since the mechanism is active in caspase-1/11-/- BMDM and can be partially blocked by Q-VD-OPh. Of the eight remaining caspases encoded in the mouse genome, namely caspase-2, 3, 6, 7, 8, 9, 12 and 14, a number have been shown to be involved in non-apoptotic functions related to host defense [75]. Caspase-7 has been shown to mediate the fusion of LCVs with lysosomes, though this was dependent on caspase-1 activity [42]. However, caspase-7 has also been demonstrated to protect cells from plasma membrane damage with the pore-forming toxin Listeriolysin O, and this was caspase-1 independent [76]. Caspase-8 has also been shown to mediate innate immune responses involving NFκB activation in response to dsRNA, as well as cell motility [77,78]. Caspases 7 and 8 might therefore be good candidates for involvement in the TNF-mediated mechanism.
Our results also implicate modest involvement of cathepsins or calpains in the TNF-mediated restriction of L. pneumophila replication in macrophages, as demonstrated by the partial inhibition of the TNF-mediated effect by E-64d. However, we did not find a requirement for cathepsin B, D, H, L or S, though a redundant requirement among these cathepsins cannot be excluded. Of note, the cathepsin B inhibitor CA-074-Me partially blocked the TNF-mediated restriction of L. pneumophila growth, however this was shown to be non-specific as the compound blocked the effect equally well in WT and CtsB-/- BMDM (S3C Fig).
Further, we find that caspase and cathepsin or calpain activity may be interdependent. This may not be a surprising result, as cathepsins have been documented to have an involvement upstream of caspase activation in other biological contexts and in vitro [79,80]. Similarly, calpains have been shown to impact the activation of caspase-8, 9 and 12 [81–83]. Furthermore, the intracellular pathogen Francisella tularensis was shown to exploit this relationship to manipulate caspases and promote its survival in neutrophils [82]. Further investigation of these mechanisms will surely yield a better understanding of TNF-mediated host defense mechanisms directed at intracellular pathogens.
This study was conducted in accordance to the guidelines of the animal experimentation law (SR 455.163; TVV) of the Swiss Federal Government. The protocol was approved by Cantonal Veterinary Office of the canton Zurich, Switzerland (Permit number 125/2012).
All mice used in this study were bred at the Swiss Federal Institute of Technology Zürich or purchased (Janvier Labs, Le Genest Saint Isle, France) and used at 6–20 weeks of age (age- and sex-matched within experiments). All mice were backcrossed >9 generations on the C57BL/6 background with the exception of MN-TNF NAIP5129S1 mice. MemTNF KI mice and MN-TNF NAIP5129S1 mice have been previously described [52,84]. Sequencing of the MN-TNF NAIP5129S1 mice revealed the same mutations in 129S1 NAIP5 (NAIP5129S1) as previously described [49], with the exception of two mutations in exon 15, which matched the C57BL/6 DNA sequence. Bone marrow chimeric mice were generated as described previously [18], reconstituting with a total of 5 x 106 bone marrow cells and allowing at least 8 weeks for reconstitution of lethally irradiated Ly5.1+ WT recipient mice. Neutrophil and AM chimerism was around 40:60 in WT:WT mice, 35:65 in WT:CYBB-/- mice and 33:67 in WT:TNFR1-/- mice.
The L. pneumophila strains used in this study were the wildtype strain JR32 (Philadelphia-1) [85], as well as modifications of JR32 including an aflagellated mutant (ΔFlaA) [86], JR32-GFP [87], JR32-GFPind (pGS-GFP-04) [88], a deletion mutant lacking a functional Icm/Dot T4SS (ΔT) [89], and ΔT-GFP [87]. L. pneumophila was grown for 3 days at 37°C on charcoal yeast extract (CYE) agar plates before use, with chloramphenicol (5 mg/ml) added for selection of strains containing GFP-encoding plasmids.
For intranasal (i.n.) infections mice were anesthetized with an i.p. injection of 5 mg xylazine/100 mg ketamine per gram body weight, and 5 x 106 CFU L. pneumophila (unless otherwise specified) resuspended in 20 μl PBS were directly applied to one nostril using a Gilson pipette. Bacterial titers in bronchoalveolar lavage fluid (BALF) were determined by plating serial dilutions in PBS on CYE plates. For quantification of CFU from sorted AM and neutrophils, cells were lysed to release viable L. pneumophila by vortexing 30 seconds in 1 ml PBS with 0.7% Tween 20 prior to plating serial dilutions in PBS on CYE plates.
Bone marrow-derived macrophages (BMDM) were generated by plating bone marrow in L929 conditioned medium containing M-CSF in 5 cm diameter non-cell culture treated Petri dishes as described previously [18]. On day 7, BMDM were harvested in ice cold PBS, 5% FBS, 2.5 mM EDTA by incubating 12 min in the fridge and resuspending by pipetting. The cells were then seeded at 1 x 105 cells/well in 96-well plates and rested overnight prior to infection. L. pneumophila used for infection was grown for 3 days at 37°C on CYE agar plates, then inoculated in ACES yeast extract medium at an OD600 of 0.1 and grown for 21 h at 37°C before use, with 5 mg/ml chloramphenicol added to maintain plasmids. BMDM were infected at MOI 0.1, synchronized by centrifugation, and incubated for 3 days at 37°C, 5% CO2. Intra- and extracellular CFU were quantified on day 3 by plating on CYE plates after a 10 min incubation in dH2O to lyse BMDM. Where indicated, 20 nM V-ATPase inhibitor bafilomycin A1 (Enzo Life Sciences, BML-CM110-0100), 25 μM cathepsin B inhibitor CA-074-Me (Enzo Life Sciences, BML-PI126-0001), 25 μM cathepsin D inhibitor pepstatin A (Enzo Life Sciences, ALX-260-085-M005), 2 μg/ml TNFR1-Fc (Adipogen, AG-40B-0074-C050), 25 μg/ml anti-IL1β (R&D, AB-401-NA), 25 μg/ml anti-TNF (Bioxcell, BE0058, clone XT3.11) or 100 ng/ml TNF (Peprotech, 315-01A) were added 15 min prior to infection.
BMDM were seeded in 24-well plates containing 0.01% polylysine solution (Sigma P4707) coated 12 mm cover glasses (Faust 6080181) at 2.5 x 105 cells/well and rested overnight. Where indicated 100 ng/ml TNF (Peprotech, 315-01A) or 200 U/ml IFNγ was added to pre-activate the BMDM. Cells were infected with Lpn-GFP as described above at MOI 5 for 1 or 3 hours at 37°C, 5% CO2, with the simultaneous addition of100 ng/ml TNF or 200 U/ml IFNγ where indicated. For the final 30 minutes of incubation 1μM lysotracker Red DND-99 (Life Technologies, L7528) and 0.5 μg/ml Cholera toxin B AF647 (CTB-AF647, Life Technologies, C34778) were added to the cells. Cells were then washed with 1 ml PBS, and cover glasses were then placed on parafilm, and fixed 5–10 min at RT with 200 μl 4% PFA in PBS. Cells were washed 3 times with 200 μl PBS, incubating 2 min after applying each wash. Cover glasses were dipped in dH2O, blotted on paper towel to remove excess water and mounted on glass slides with cells facing downwards with 6 μl Mowiol (VWR, 475904–100). Z-stack images were acquired on a spinning-disk confocal microscope (Visitron confocal system) using a 100x objective, and analyzed with volocity software (PerkinElmer, Waltham, MA). To assess co-localization of L. pneumophila and lysosomes, at least 100 bacteria were scored per coverslip.
BALF was recovered from mice at the specified timepoint in 1 ml sterile PBS containing 5 mM EDTA as previously described [90]. Cells were surface stained 30 min in cold FACS buffer (PBS with 2.5% FBS, 5 mM EDTA) with Siglec-F (clone E50-2440, Biolegend), CD11c (clone N418, Biolegend), Ly6G (clone 1A8, BD Biosciences), Ly6C (clone AL-21, BD Biosciences, Allschwil, Switzerland), CD11b (clone M1/70, Biolegend), CD45.1 (clone A20, BD Biosciences), CD45.2 (clone 104, BD Biosciences).
For intracellular staining of TNF (clone MP6-XT22, Biolegend), mice were injected i.p. with 50 μl of 5 mg/ml Brefeldin A in EtOH (diluted with 100 μl PBS) 3 hours prior to taking BALF. Lavage was performed with 1 ml PBS 5mM EDTA containing 5 μg/ml Brefeldin A, and was immediately placed on ice. After surface stain, cells were washed with FACS buffer and fixed, permeabilized and stained using the BD Biosciences Cytofix/Cytoperm Kit according to the manufacturer's instructions. Data were acquired on an LSRII (BD Biosciences) and analyzed with FlowJo software (TreeStar, Ashland, OR). An Aria III instrument (BD Biosciences) was used for cell sorting.
ROS was stained in BALF cells by collecting BALF as usual in 1 ml PBS 5 mM EDTA, washing with 2 ml RPMI 10% FBS at RT, and staining with 60 μM Dihydroethidium (Sigma, D7008) for 1 hour at 37°C, 5% CO2. For a positive control, cells were stimulated with PMA/ionomycin. Cells were then washed in 2 ml cold FACS buffer and stained as usual with fluorescence-labeled Abs. Data were acquired on an LSRII (BD Biosciences), Dihydroethidium was measured in the FITC channel.
Non-parametric tests, including the Kruskal-Wallis test with Dunn's post test, the Mann-Whitney test, or in the case of paired samples, the Wilcoxon test, were applied for statistical analysis using Prism GraphPad software (La Jolla, CA).
|
10.1371/journal.pcbi.1000425 | Dynamic Modeling of Vaccinating Behavior as a Function of Individual Beliefs | Individual perception of vaccine safety is an important factor in determining a person's adherence to a vaccination program and its consequences for disease control. This perception, or belief, about the safety of a given vaccine is not a static parameter but a variable subject to environmental influence. To complicate matters, perception of risk (or safety) does not correspond to actual risk. In this paper we propose a way to include the dynamics of such beliefs into a realistic epidemiological model, yielding a more complete depiction of the mechanisms underlying the unraveling of vaccination campaigns. The methodology proposed is based on Bayesian inference and can be extended to model more complex belief systems associated with decision models. We found the method is able to produce behaviors which approximate what has been observed in real vaccine and disease scare situations. The framework presented comprises a set of useful tools for an adequate quantitative representation of a common yet complex public-health issue. These tools include representation of beliefs as Bayesian probabilities, usage of logarithmic pooling to combine probability distributions representing opinions, and usage of natural conjugate priors to efficiently compute the Bayesian posterior. This approach allowed a comprehensive treatment of the uncertainty regarding vaccination behavior in a realistic epidemiological model.
| A frequently made assumption in population models is that individuals make decisions in a standard way, which tends to be fixed and set according to the modeler's view on what is the most likely way individuals should behave. In this paper we acknowledge the importance of modeling behavioral changes (in the form of beliefs/opinions) as a dynamic variable in the model. We also propose a way of mathematically modeling dynamic belief updates which is based on the very well established concept of a belief as a probability distribution and its temporal evolution as a direct application of the Bayes theorem. We also propose the use of logarithmic pooling as an optimal way of combining different opinions which must be considered when making a decision. To argue for the relevance of this issue, we present a model of vaccinating behaviour with dynamic belief updates, modeled after real scenarios of vaccine and disease scare recorded in the recent literature.
| Since early vaccination campaigns against smallpox, vaccination policies have been a matter of debate [1]: mass vaccination versus blocking strategies; compulsory versus voluntary, are some highly debated issues. Despite these early controversies - and consequent alternative policies implemented in different countries - high disease scare in the past has led to very high vaccine coverage and consequent successful eradication of smallpox, as well as very low incidence of measles, polio, tetanus, diphtheria, etc, resulting in over 98% mortality reduction by vaccine preventable diseases in developed countries [2].
In recent years, after complete or almost complete elimination of these diseases, the debate is shifting towards issues of vaccine safety. Increased perception of vaccine risks and lowered perception of disease risks has challenged previous willingness to vaccinate (fundamental for the success of any immunization program, either voluntary or compulsory) [3]. In this scenario, understanding and predicting individual's willingness to vaccinate is paramount for estimating vaccine coverage and compare strategies to achieve coverage goals.
Willingness to vaccinate is highly dependent on the perceived risk of acquiring a serious disease [4]. When (perceived) disease risk is low, however small risk of adverse events from the vaccine become relatively important and may lead to vaccine coverage lower than required to control transmission [4]. When (perceived) serious disease risk is too high, on the other hand, vaccine coverage may increase above that required to guarantee population protection [5]. We illustrate these behaviors with two examples:
In the UK, MMR vaccine uptake started to decline after a controversial study linking MMR vaccine to autism [6]. In a decade, vaccine coverage went well below the target herd immunity level of 95%. Despite the confidence of researchers and most health professionals on the vaccine safety, the confidence of the public was deeply affected. In an attempt to find ways to restore this confidence, several studies were carried out to identify factors associated with parent's unwillingness to vaccinate their children. They found that ‘Not receiving unbiased and adequate information from health professionals about vaccine safety’ and ‘media's adverse publicity’ were the most common reasons influencing uptake [7]. Other important factors were: ‘lack of belief in information from the government sources’; ‘fear of general practitioners promoting the vaccine for personal reasons’; and ‘media scare’. Note that during this period the risk of acquiring measles was very low due to previously high vaccination coverage.
Sylvatic yellow fever (SYF) is a zoonotic disease, endemic in the north and central regions of Brazil. Approximately 10% of infections with this flavivirus are severe and result in hemorrhagic fever, with case fatality of 50% [8]. Since the re-introduction of A. aegypti in Brazil (the urban vector of dengue and yellow fever), the potential reemergence of urban yellow fever is of concern [9]. In Brazil, it is estimated that approximately 95% of the population living in the yellow fever endemic regions have been vaccinated. In this area, small outbreaks occur periodically, especially during the rainy season, and larger ones are observed every 7 to 10 years [10], in response to increased viral activity within the environmental reservoir. In 2007, increased detection of dead monkeys in the endemic zone, led the government to implement vaccine campaigns targeting travellers to these areas and the small fraction of the resident population who were still not protected by the vaccine. The goal was to vaccinate 10–15% of the local population. Intense notification in the press regarding the death of monkeys near urban areas, and intense coverage of all subsequent suspected and confirmed human cases and death events led to an almost country-wide disease scare (Figure 1), incompatible with the real risks [5], which caused serious economic and health management problems, including waste of doses with already immunized people (60% of the population was vaccinated when only 10–15% would be sufficient), adverse events from over vaccination (individuals taking multiple doses to ‘guarantee’ protection), national vaccine shortage and international vaccine shortage, since Brazil stopped exporting YF vaccine to supply domestic vaccination rush (www.who.int/csr/don/2008_02_07/en/).
The importance of public perceptions and collective behavior for the outcome of immunization campaigns are starting to be acknowledged by theoreticians [9],[11],[12]. These factors have been examined in a game theoretical framework, where the influence of certain types of vaccinating behaviour on the stability and equilibria of epidemic models is analyzed.
In the present work, we propose a model for individual immunization behavior as an inference problem: Instead of working with fixed behaviors, we develop a dynamic model of belief update, which in turn determines individual behavior.
An individual's willingness to vaccinate is derived from his perception of disease risk and vaccine safety, which is updated in a Bayesian framework, according the epidemiological facts each individual is exposed to, in their daily life. We also explore the global effects of individual decisions on vaccination adherence at the population level.
In summary, we propose a framework to integrate dynamic modeling of learning (belief updating) with decision and population dynamics.
We ran the model as described above for 100 days with parameters given by Table 1, under various scenarios to reveal the interplay of belief and action under the proposed model. Figures 2 and 3 show a summary output of the model dynamics under contrasting conditions. In Figure 2, we have VAE (Vaccine adverse events) preceding the occurrence of severe disease events. As expected, VAE become the strongest influence on , keeping low with consequences to the attained vaccination coverage at the end of the simulation. We characterize this behavior as a ‘vaccine scare’ behavior.
In a different scenario, Figure 3, we observe the effect of severe disease events occurring in high frequency at the beginning of the epidemics. In this case, disease scare pushes willingness to vaccinate () to high levels. This is very clear in Figure 3 where there is a cluster of serious disease cases around the 30th day of simulation. right after the occurrence of this cluster, we see rise sharply above , meaning that willingness to vaccinate () in this week was mainly driven by disease scare instead of considerations about vaccine safety(). A similar effect can be observed in Figure 2, starting from day 45 or so. Only here the impact of a cluster of serious disease cases is diminished by the effects of VAEs, and the fact that there aren't many people left to make the decision of wether or not vaccinate.
The impact of individual beliefs on vaccine coverage is highly dependent on the visibility of the rare VAE. Figure 4 shows the impact of the media amplification factor on and vaccination coverage after ≈14 weeks, for a infectious disease with and . If no media amplification occurs, willingness to vaccinate and vaccine coverage are high, as severe disease events are common and severe adverse events are relatively rare. As vaccine adverse events are amplified by the media, individual's willingness to vaccinate at the end of the 14 weeks tend to decrease. Such belief change, however, has a low impact on the vaccine coverage. The explanation for this is that vaccine coverage is a cumulative measure and, when VAE appear, a relatively large fraction of the population had already been vaccinated. These results suggest that VAE should not strongly impact the outcome of an ongoing mass vaccination campaign, although it could affect the success of future campaigns.
Fixing amplification at and , we investigated how (at the end of the simulation) and vaccine coverage would be affected by increasing the rate of vaccine adverse events, (Figure 5). As increases above , willingness to vaccinate drops quickly, while vaccine coverage diminishes but slightly.
In the present world of mass media channels and rapid and inexpensive communications, the spread of information, independent of its quality, is very effective, leading to considerable uncertainty and heterogeneity in public opinions. The yellow fever scare in Brazil demonstrated clearly the impact of public opinion on the outcome of a vaccination campaign, and the difficulty in dealing with scare events. For example, no official press release was taken at face value, as it was always colored by political issues [5]. In multiple occasions, people reported to the press that they would do the exact opposite of what was being recommended by public health authorities due to their mistrust of such authorities. This example shows us the complexity of modeling and predicting the success of disease containment strategies.
The goal of this work was to integrate into a unified dynamical modeling framework, the opinion and decision components that underlie the public response to mass vaccination campaigns, specially when vaccine or disease scares have a chance to occur. The proposed analytical framework, although not intentionally parameterized to match any specific real scenario, qualitatively captured the temporal dynamics of vaccine uptake in Brasilia (Figure 1), a clear case of disease scare (compare with simulation results, presented on Figure 2).
After conducting large scale studies on the acceptance of the Influenza vaccine, Chapman et al. [13] conclude that perceived side-effects and effectiveness of vaccination are important factors in people's decision to vaccinate. Our model suggests that, if the perception of disease risk is high, it leads to a higher initial willingness to vaccinate, while adverse events of vaccination, even when widely publicized by the media, tend to have less impact on vaccination coverage. VAE are more effective when happening at the beginning of vaccination campaigns, when they can sway the opinions of a larger audience. Although disease scare can counteract, to a certain extent the undesired effects of VAE, public health officials must also be aware of the risks involved in overusing disease risk information, in vaccination campaign advertisements since this can lead to a rush towards immunization as seen in the 2008 Yellow Fever scare in Brazil.
Vaccinating behavior dynamics has been modelled in different ways in the recent literature, from behaviors that aim to maximize self-interest [12] to imitation behaviors [14]. In this paper we modeled these perceptions dynamically, and showed its relevance to decision-making dynamics and the consequences to the underlying epidemiological system and efficacy of vaccination campaigns. We highlight two aspects of our modeling approach that we think provide important contributions to the field.
First, the process through which people update beliefs which will direct their decisions, was modeled using a Bayesian framework. We trust this approach to be the most natural one as the Bayesian definition of probability is based on the concept of belief and Bayesian inference methodology was developed as a representation human learning behavior [15]. The learning process is achieved through an iterative incorporation of newly available information, which naturally fit into the standard Bayesian scheme. Among the advantages of this approach is its ability to handle the entire probability distributions of the parameters of interest instead of operating on their expected values which would be the cased in a classical frequentist framework. This is especially important where highly asymmetrical distributions are expected. The resulting set of probability distributions, provides more complete model-based hypotheses to be tested against data. The inferential framework has an added benefit of simplicity and computational efficiency due the use of conjugate priors, which gives us a closed-form expression for the Bayesian posterior without the need of complex posterior sampling algorithms such as MCMC.
The second contribution is the articulation between the belief and decision models through logarithmic pooling. Logarithmic pooling has been applied in many fields [16],[17] to derive consensus from multiple expert opinions described as probability distributions. Genest et al. [15], argue that Logarithmic pooling is the best way to combine probability distributions due to its property of “external Bayesianity”. This means that finding the consensus among distributions commutes with revising distributions using the Bayes formula, with the consequence that the results of this procedure can be interpreted as a single Bayesian probability update. Here, we apply logarithmic pooling to integrate the multiple sources of information (equation (1)) which go into the decision of whether or not to vaccinate. In this context, the property of external bayesianity, is important since it allows the operations of pooling and Bayesian update (of , equation (2)) to be combined in any order, depending only on the availability of data.
This framework can be easily used as a base to compose more complex models. Extended models might include multiple beliefs as a joint probability distribution, more layers of decision or multiple, independently evolving belief systems.
The contact strucure of the model was intentionally kept as simple as possible, since the goal of the model was to focus on the belief dynamics. Therefore, a reasonably simple epidemiological model, with a simple spatial structure (local and global spaces) was constructed to drive the belief dynamics without adding potentially confounding extra dynamics.
In this work we have played with various probability levels of VAEs and SDs in an attempt to cover the most common and likely more interesting portions of parameter space. However, to model specific scenarios, data regarding the actual probabilities of VAEs and SDs are a pre-requisite. Also important are data regarding the perception of vaccine safety and efficacy [18], obtainable through opinion surveys which could also include questions about factors driving changes in vaccination behavior. We therefore suggest that questions regarding these variables should be included in future surveys concerning vaccine-preventable diseases. This would improve our ability to predict of the outcome of vaccination campaigns.
We set the vaccination decision problem in the context of a population experiencing a vaccine preventable disease outbreak which leads to a mass vaccination campaign. Individuals receive information regarding vaccine and disease events from local and global sources. We assume that 'good' events (prompt recovery from infection or safe vaccine events) are visible locally only while severe cases of disease or potentially adverse events from the vaccine enjoy global visibility due to the natural preference of media channels for scary stories. In order to integrate behavioral and epidemiological dynamics, an individual based model was developed. Individual's behavior regarding vaccination is represented in a belief-decision model which describes the dynamics of belief updates in response to epidemiological events and the decision making based on the person's current beliefs. The epidemiological model determines the disease dynamics in a population with hierarchical contact structure, representing a large urban setting.
The belief model describes the temporal evolution of each individual's willingness to vaccinate, , in response to his evaluation of vaccine safety and disease risk. To account for the uncertainties regarding vaccinating behavior, is modeled as a random variable, whose distribution is updated weekly as the individual observes new events. The update process is based on logarithmically pooling with other random variables as described below. Logarithmic pooling is a standard way of combining probability distribution representing opinions, to form a consensus [15].
The belief update model takes the form:(1)
where must equal one as act as weights of the pooling operation. We attributed equal weights to and (), with remaining taking values according to the following conditions:where is the number of serious disease cases witnessed by the individual, and and are random variables describing individual's belief regarding vaccine safety and disease risk, respectively. The values for and are set to 1/2 since either or are to be pooled against the combination of and : . This choice of weights corresponds to the most unassuming scenario regarding the relative importance of each information source, different weights may be chosen for different scenarios. Every individual starts off with a very low expected value for the Beta-distributed .
The last term in (1), , is a reduction force which causes to move towards the minimum value of . This term is important since without it, the psychological effects of witnessing serious disease events would continue to influence the individual's decisions for and indetermined period of time. Thus, allows us to include the memory of such events in the model. By setting appropriately, we can model events that leave no memory as well as ones that are retained indefinetly.
We model disease spread in a hypothetical city represented by a multilevel metapopulation individual-based model where individuals belong to groups that in turn belong to groups of groups, and so on (Figure 9), forming a hierarchy of scales [20]. In this hypothetical city, individuals live in households with exactly 4 members each; neighborhoods are composed by 100 households and sets of 10 neighborhoods form the city's zones. During the simulation, individuals commute between home and a randomly chosen neighborhood anywhere in the population graph. Each individual has a probability 0.25 of leaving home daily.
This same hierarchical structure is used to define local and global events. Locally visible events can only be witnessed by people living in the same neighborhood while globally visible events are visible to the entire population regardless of place of residence.
The epidemiological model describes a population being invaded by a new pathogen. This pathogen causes an acute infection, lasting 11 days (incubation period of 6 days and an infectious period of 5 days). Once in the infectious period, individuals have a fixed probability, of becoming seriously ill. After recovery, individuals become fully immune. The proportion of the population in each immunological state at time is labeled as and , which stands for susceptibles, exposed, infectious and recovered states.
At the same time the disease is introduced in the population, a vaccination campaign is started, making available doses per week to the entire population, meaning that individuals may have to compete for a dose if many decide to vaccinate at the same time.
Once an individual is vaccinated, if he/she has not been exposed yet, he/she moves directly to the recovered class, with full immunity (thus, a perfect vaccine is assumed). If the individual is in the incubation period of the disease, disease progression is unaffected by vaccination. Vaccination carries with it a fixed chance of causing adverse effects.
Transmission dynamics is modelled as follows: at each discrete time step, , each individual contacts others in two groups: in his residence and in the public space. The probability of getting infected at home is given by where is the probability of transmission per household contact and is the number of infected members in the house. In the public space, that is, in the neighborhood chosen as destination for the daily commutations, each infected person contacts persons at random, and if the contact is with a susceptible, infection is transmitted with probability .
|
10.1371/journal.ppat.1004200 | Ubiquitin-Mediated Response to Microsporidia and Virus Infection in C. elegans | Microsporidia comprise a phylum of over 1400 species of obligate intracellular pathogens that can infect almost all animals, but little is known about the host response to these parasites. Here we use the whole-animal host C. elegans to show an in vivo role for ubiquitin-mediated response to the microsporidian species Nematocida parisii, as well to the Orsay virus, another natural intracellular pathogen of C. elegans. We analyze gene expression of C. elegans in response to N. parisii, and find that it is similar to response to viral infection. Notably, we find an upregulation of SCF ubiquitin ligase components, such as the cullin ortholog cul-6, which we show is important for ubiquitin targeting of N. parisii cells in the intestine. We show that ubiquitylation components, the proteasome, and the autophagy pathway are all important for defense against N. parisii infection. We also find that SCF ligase components like cul-6 promote defense against viral infection, where they have a more robust role than against N. parisii infection. This difference may be due to suppression of the host ubiquitylation system by N. parisii: when N. parisii is crippled by anti-microsporidia drugs, the host can more effectively target pathogen cells for ubiquitylation. Intriguingly, inhibition of the ubiquitin-proteasome system (UPS) increases expression of infection-upregulated SCF ligase components, indicating that a trigger for transcriptional response to intracellular infection by N. parisii and virus may be perturbation of the UPS. Altogether, our results demonstrate an in vivo role for ubiquitin-mediated defense against microsporidian and viral infections in C. elegans.
| Microbial pathogens have two distinct lifestyles: some pathogens live outside of host cells, and others live inside of host cells and are called intracellular pathogens. Microsporidia are fungal-related intracellular pathogens that can infect all animals, but are poorly understood. We used the roundworm C. elegans as a host to show that ubiquitin pathways provide defense against both a natural microsporidian infection of C. elegans, as well as a natural viral infection. Our study shows that ubiquitin, the proteasome and autophagy components are all important to control intracellular infection in C. elegans, although microsporidia seem to partially evade this defense. We also show that SCF ubiquitin ligases help control both microsporidia and virus infection. Furthermore, we find that C. elegans upregulates expression of SCF ligases when ubiquitin-related degradation machinery is inhibited, indicating that C. elegans monitors the functioning of this core cellular process and upregulates ligase expression when it is perturbed. Altogether, our findings describe ubiquitin-mediated pathways that are involved in host response and defense against intracellular pathogens, and how this machinery is regulated by infection to increase defense against intracellular pathogens such as microsporidia and viruses.
| The Microsporidia phylum contains over 1400 species of obligate intracellular pathogens most closely related to fungi [1]. These pathogens can infect a wide variety of animal hosts including humans, where they can cause significant disease. Infections in humans can cause lethal diarrhea in immunocompromised people such as AIDS patients, and microsporidia are considered priority pathogens at the National Institutes of Health [2], [3]. Microsporidia can also plague agriculturally significant animals such as fish and honeybees [4], [5], [6]. Treatment options for microsporidia infections are limited and often ineffective [7], [8]. In mammals, studies have shown that T cells and dendritic cells provide protection against infection, but little is known about the innate and/or intracellular responses to these pathogens [9], [10], [11].
Previously, we described Nematocida parisii, a microsporidian species isolated from a wild-caught C. elegans near Paris, which causes a lethal intestinal infection in its host [12], [13]. N. parisii infection of the simple nematode C. elegans provides a convenient system in which to investigate host responses and defense against microsporidia infection. Interestingly, canonical C. elegans defense pathways, such as the conserved PMK-1 p38 MAPK pathway that provides defense against bacterial and fungal infections, are not important for defense against N. parisii [12], [14]. Thus, distinct immunity mechanisms may be involved in the C. elegans response to microsporidia. In addition to microsporidia, another natural intracellular infection has recently been described in C. elegans: wild-caught animals from Orsay, France, were shown to harbor a viral infection [15]. The Orsay virus is a positive strand RNA virus of the family Nodaviridae, and like N. parisii it appears to undergo its entire replicative cycle inside C. elegans intestinal cells. The RNAi pathway has been shown to provide defense against viral infections in C. elegans [15], [16], [17], [18], [19], but little else is known about host defense against this natural intracellular pathogen of C. elegans.
Defense against intracellular pathogens in diverse animal hosts is increasingly appreciated to involve ubiquitin-mediated degradation pathways [20], [21], [22], [23]. Ubiquitylation is the process by which an E3 ubiquitin ligase catalyzes the conjugation of a ubiquitin tag onto substrates, which can be further ubiquitylated to generate poly-ubiquitin chains [24]. Ubiquitylated substrates have a number of different fates, two of which involve degradation. The most well characterized fate is degradation by the proteasome, but larger substrates can be targeted for degradation by the process of autophagy, which is termed 'xenophagy' when it involves degradation of intracellular microbes [25], [26]. Recently, ubiquitin ligases that mediate ubiquitin targeting to human bacterial pathogens Salmonella enterica [21] and Mycobacterium tuberculosis [22] have been identified, and they, together with the autophagy pathway, are important for controlling levels of these intracellular pathogens [23], [27], [28], [29]. However, while several ubiquitin-mediated defense components and mechanisms have been defined, there are many unanswered questions about which host ubiquitin ligases are involved in targeting ubiquitin to different pathogens, how these systems are regulated, and their overall importance for defense in vivo.
One major class of E3 ubiquitin ligases includes the Skp1−Cul1−F-box protein (SCF) multi-subunit RING-finger type, which is a modular complex found throughout eukaryotes [30]. SCF ligases are usually composed of three core components (a cullin protein, Skp1, and a RING-containing subunit) and a variable F-box protein component, which enables recognition of different substrates depending on which F-box protein is associated with the complex [31]. Interestingly, the C. elegans genome has a greatly expanded and diversified family of F-box proteins (∼520 genes compared to 69 genes in humans), as well as other SCF components (21 Skp1-related genes compared to 1 in humans), suggesting they use SCF ligases to recognize an extremely diverse array of substrates [32], [33]. In particular, it has been proposed that C. elegans uses these SCF ligases to target toxins and intracellular pathogen proteins for degradation, and that the expanded C. elegans SCF ligase repertoire is the manifestation of a host/pathogen arms race between nematodes and their natural intracellular pathogens [32]. At the time this intriguing idea was proposed however, there were no known intracellular pathogens of C. elegans to test the role of ubiquitin-mediated responses in defense.
Here we describe the C. elegans host response to the natural intracellular pathogens N. parisii and the Orsay virus, and find a role for ubiquitin-mediated defense against both infections. We perform gene expression analyses of the transcriptional response to microsporidia infection and find that the response is strikingly similar to the response to viral infection, but not to extracellular pathogens. We see upregulation of SCF ligase components, which help to restrict microsporidia growth, and find that defense against microsporidia appears to rely on the proteasome, as well as the autophagy pathway. We find a subset of parasite cells targeted by host-derived ubiquitin, which relies partly on the SCF cullin component CUL-6. Notably, this ubiquitin targeting, as well as the role for ubiquitin-mediated defense, increases upon inhibition of microsporidia growth by anti-microsporidia drugs. These results suggest that N. parisii may suppress or evade ubiquitin-mediated host defenses. Interestingly, expression of specific infection-upregulated SCF ligase components is also upregulated by genetic or pharmacological inhibition of UPS function, suggesting that stress placed upon the UPS may be a hallmark of intracellular infection, and that hosts monitor UPS function to upregulate appropriate defenses during intracellular infection. Finally, we show that SCF ligase components, in particular CUL-6, promote defense against viral infection in C. elegans. Altogether, these studies show the involvement of ubiquitin-mediated defense and xenophagy against natural intracellular pathogens in a whole animal host, and provide insight into their regulation in response to infection in vivo.
We examined the C. elegans transcriptional response over the course of an infection with N. parisii using strand-specific deep sequencing of RNA (RNA-seq). Like other microsporidia, the life cycle of N. parisii is complex and its growth and replication takes place entirely inside the host cell (Figure 1A). Microsporidian spores initiate an intracellular infection by firing an infection apparatus called a polar tube, which pierces the host cell membrane and then injects into the host cell a nucleus and sporoplasm, which replicates as a stage called a meront. In the case of N. parisii, meronts become very large, multi-nucleate cells that replicate in direct contact with the cytoplasm. Meronts will eventually differentiate into spores and these spores then exit from infected cells to infect new hosts. We collected and sequenced cDNA from age-matched uninfected controls and infected animals at 8, 16, 30, 40 and 64 hours post inoculation (hpi) (Figure 1A, B), which are timepoints that correspond to specific stages of N. parisii infection as described in our previous study [34] (Table S1). A large number of C. elegans genes had significantly altered expression during N. parisii infection (edgeR, FDR<0.05, Table S2). The overall number of upregulated genes was relatively stable throughout infection, while the number of downregulated genes increased markedly with time (Figure 1C). To validate our RNA-seq studies, we also performed Affymetrix microarrays, which had substantial agreement in the genes found to be regulated by N. parisii infection (see Supplemental Text S1, Table S3). Notably, we found that a significant number of genes upregulated by infection were associated with the intestine, which is the site of N. parisii infection (Figure 1D).
Next, we compared genes regulated by N. parisii (Table S4) to gene sets regulated by infection with other pathogenic microbes, by treatment with non-biotic stressors, and by known immunity and stress-response pathways in C. elegans [17], [35], [36], [37], [38], [39], [40], [41] (Figure 1E, Table S5, Table S6). Here, we used a well-established analytical method called Gene Set Enrichment Analysis (GSEA), which analyzes gene expression data at the level of gene sets instead of individual genes (see Materials and Methods) [42]. We found limited but significant correlations with gene sets upregulated by heat shock treatments, the pore-forming toxin Crystal protein-5B (Cry5B), and Drechmeria coniospora fungal infection, predominantly at the 30 hpi timepoint (Figure 1E). The heat shock pathway has been shown to play a role in resistance to bacterial pathogens as well as other stresses [43], [44]. However, despite the overlap between genes induced by heat shock and microsporidia, we found that N. parisii infection upregulated only two canonical heat shock protein-encoding genes, hsp-17 at 30 hpi and hsp-16.1/hsp-16.11 (which have identical sequence and are indistinguishable in RNA-seq data) at 64 hpi (Table S7, Figure S1A). Notably, there was almost no correlation between C. elegans genes upregulated in response to N. parisii infection compared to infections with the extracellular bacterial pathogens Pseudomonas aeruginosa and Staphylococcus aureus, the fungal pathogen Harposporium, or to genes affected by known C. elegans immunity regulators (Figure 1E). However, there was extensive correlation between genes downregulated by N. parisii and genes downregulated by other pathogens - for further discussion of this correlation, and other comparisons to previously published gene expression analyses see Supplemental Text S1 and Figure S2. Strikingly, we found a very strong correlation between genes most strongly upregulated by N. parisii, (e.g. genes of unknown function C17H1.6 and F26F2.1) and genes upregulated by viral infection (Figure 1E, Figure S1B). Thus, N. parisii induces robust gene expression changes that are largely distinct from changes induced by extracellular pathogens, but share similarity to changes induced by the Orsay virus, which is another natural intracellular pathogen of C. elegans.
To understand the nature of the C. elegans response to microsporidia infection, we analyzed the enrichment of gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) terms for the significantly induced and repressed genes [45], [46] (Table 1 and S8). Early during infection upregulated genes were enriched for GO terms associated with regulation of growth, while at later timepoints they were enriched for GO terms associated with the nucleosome, defense response, and structural components. At 30 hpi, upregulated genes were enriched for association with the ubiquitin-mediated proteolysis KEGG pathway (Table 1). To extend our analysis, we also identified specific enrichment of Pfam protein domains among N. parisii regulated genes (Table 1 and S8). At early times following infection these included two Caenorhabditis domains of unknown function, DUF713 and DUF684. Notably, genes upregulated at 8, 16 and 30 hpi were also enriched for the F-box, FTH (fog-2-homology), and MATH (meprin and Traf homology) protein-protein interaction domains, which are domains associated with ubiquitin-mediated proteolysis. For more details on regulated proteins containing these domains, analysis of gene enrichment at later time points, and analysis of downregulated genes, see Supplemental Text S1.
Previously it had been hypothesized that F-box and MATH domain-containing proteins could function in C. elegans to target foreign pathogen proteins for proteasomal degradation, as part of SCF multi-subunit E3 ubiquitin ligases [32]. Indeed, we found that C. elegans SCF ligase components, Skp1-related (skr) genes skr-4 and skr-5, were significantly upregulated at 30 hpi with N. parisii (Table S2), while skr-3 and the cullin gene cul-6 were also upregulated at 30 hpi over 6.5- and 5.5-fold respectively, although the difference was not significant (Table S4). While these SCF ligase components were not reported to be significantly upregulated in a published dataset of the wild-type C. elegans response to viral infection [17], we found that in the virus-susceptible rde-1 strain of C. elegans, the SCF ligase components cul-6, skr-3, skr-4, and skr-5 were upregulated in response to viral infection (data not shown). Overall, this increased expression of genes encoding SCF ligase components (see Table S9 for list of significantly upregulated ubiquitylation-associated genes) is consistent with ubiquitylation being upregulated in virus and microsporidia-infected animals.
To examine a functional role for genes induced by N. parisii infection we used RNAi to knock-down expression of specific genes, then infected these animals with N. parisii and measured pathogen load at 24 hpi by quantifying N. parisii rRNA FISH signal (Figure 2A, S3). We tested several genes highly induced by infection, as well as genes that belong to gene classes identified through our GO term and Pfam domain analysis. Knock-down of most genes showed little to no effect on pathogen load (see Supplemental Text S1 and Figure S4). When we examined whether the upregulated SCF ligase components have a functional role in defense against N. parisii infection we found a more substantial role. Because there are a large number of F-box proteins in the C. elegans genome (∼520 proteins), we focused on the core SCF ligase components that belong to smaller families, namely the Skp1-related skr family (21 proteins) and the cullin family (6 proteins), which likely have less functional redundancy than F-box proteins. In particular, we knocked down expression of cul-6, skr-3, skr-4 and skr-5 because these were upregulated upon N. parisii infection. Here, we found a modest but significant increase in pathogen load in cul-6, skr-3 and skr-5 RNAi-treated animals (Figure 2B), suggesting that these SCF ligase components limit the growth of N. parisii during infection.
After substrates have been ubiquitylated by ubiquitin ligases, they can either be degraded by the proteasome or by the autophagy pathway. First, we examined whether components of the proteasome may be acting downstream of SCF ligase components in limiting growth of N. parisii. We reduced expression of ubiquitin itself with RNAi against ubq-2, as well as two components of the proteasome: pas-5 and rpn-2. In order for animals to develop properly, we introduced the RNAi in a diluted form at a late larval stage, infected animals and measured pathogen load. We found that reducing expression of any of these three genes led to an increase in pathogen load, suggesting that the UPS is important for defense against N. parisii (Figure 2C).
Because the effect of ubiquitin knock-down on pathogen load was modest, we hypothesized that, like other intracellular pathogens, N. parisii may suppress this defense system or subvert some aspects of the UPS to promote its replication. To test this hypothesis, we treated animals with drugs that block N. parisii growth but have minimal effects on adult C. elegans (see Supplementary Text S1) [47], [48]. First, we treated animals with a low dose of the anti-microsporidia drug fumagillin [49], [50], [51], [52], which limits N. parisii growth (Figure 2D and data not shown). After fumagillin treatment we found that ubq-2 RNAi had a more robust effect on pathogen load (150% increase) than in the absence of this drug (50% increase) (Figure 2D). Similarly, ubq-2 RNAi had a stronger effect on pathogen load when N. parisii growth was repressed with a DNA synthesis inhibitor, FUdR, (320% increase) than in the absence of this drug (70% increase) (Figure 2E). Taken together, these results suggest that the host UPS plays a greater role in controlling infection when pathogen growth is inhibited.
We next investigated a role for the autophagy pathway in response to N. parisii infection. We used RNAi to knock-down expression of different autophagy components, infected these animals with N. parisii, and quantified pathogen load. Similar to the effects of knocking down components of the UPS, we found a modest but significant increase in pathogen load when expression of several key autophagy components was reduced (Figure 2F). Furthermore, RNAi of the C. elegans nutrient sensor TOR (Target Of Rapamycin) ortholog let-363, which activates autophagy in C. elegans [53], caused a dramatic 70% decrease in pathogen load (Figure 2G). To determine whether autophagy machinery was directed toward N. parisii cells, we examined localization of GFP-tagged LGG-1 (homolog of Atg8/LC3 in yeast/mammals) [54], a protein whose distribution is often used to assess autophagy [55]. We found that early during infection only 7% of parasite cells (25/360 parasite cells, n = 6 animals) were targeted by GFP::LGG-1 (Figure 3A–C). When animals were treated with let-363 RNAi, we found that there was a greater than 2-fold increase in parasite cells targeted by GFP::LGG-1 (Figure 3D), consistent with this treatment causing an upregulation of the autophagy machinery directed toward N. parisii cells. Thus, the autophagy machinery appears to be targeted to N. parisii cells, and promotes resistance against infection.
One potential caveat to the results described above is that specific RNAi treatments might affect the feeding rates of nematodes, which could then result in changes in pathogen load simply due to differences in the initial dose of N. parisii spores ingested by these animals. To address this concern, we measured the accumulation of fluorescent beads in the intestinal lumen of animals fed dsRNA against the genes described above (Figure S5A–C). Importantly, RNAi against let-363/TOR, which causes decreased pathogen load, did not cause a decrease in the accumulation of fluorescent beads. In addition, RNAi against most autophagy genes that caused increased pathogen load did not cause an increase in accumulation of fluorescent beads. Furthermore, UPS RNAi, which increases pathogen load, did not increase fluorescent bead accumulation, and to the contrary, knock-down of ubq-2 or pas-5 marginally inhibited accumulation. Finally, feeding rates as measured by pharyngeal pumping were not affected by RNAi treatments, with the exception of ubq-2 RNAi, which caused a decrease in feeding (Figure S5D–F). For further details on these controls, see Supplemental Text S1. Altogether, our data support the model that defense against N. parisii infection involves ubiquitylation components, the proteasome, and the autophagy pathway, although microsporidia appears to partially evade or suppress this ubiquitin-mediated response.
To examine whether N. parisii itself is targeted by ubiquitin, we stained infected animals with the FK2 antibody, which recognizes ubiquitin that is conjugated to a substrate, and with a FISH probe against N. parisii rRNA to label the pathogen. Because N. parisii is a eukaryote, it contains its own ubiquitin, which is recognized by the FK2 antibody. However, distinct from this staining, we observed very strong accumulation of conjugated ubiquitin surrounding a subset of N. parisii meronts, with signal far above background of the microsporidia-derived ubiquitin (Figure 4A). To confirm that this ubiquitin was host-derived, we created a transgenic C. elegans strain that expresses a GFP::ubiquitin fusion protein under the control of an intestinal-specific promoter. Using these transgenic animals, we observed targeting of GFP::ubiquitin to parasite cells (Figure 4B). In contrast, we did not observe significant targeting to parasite cells by a conjugation defective GFP::ubiquitinΔGG fusion protein (Figure 4C, Figure S6). Altogether these experiments demonstrate that host ubiquitin is specifically targeted to N. parisii cells, where it is conjugated to a substrate.
The percentage of N. parisii cells specifically targeted by ubiquitin was relatively low: using the FK2 antibody we found only about 5% of pathogen cells were targeted by ubiquitin at 12 hpi (Figure 4D). Similarly, we found only about 7% of pathogen cells were targeted by GFP::ubiquitin (Figure 4E). Therefore, we examined whether N. parisii is suppressing or evading ubiquitin targeting by the host. If so, inhibiting the growth/vigor of the pathogen should cause an increased level of ubiquitin targeting. Indeed, we found increased targeting of ubiquitin to parasite cells after fumagillin treatment, with 16–18% of cells targeted (Figure 4D, E). This effect was dose-dependent, and was apparent both with the FK2 antibody, as well as the GFP::ubiquitin fusion protein. These results support the hypothesis that N. parisii is actively suppressing or evading ubiquitin targeting by C. elegans, and that after inhibition of N. parisii growth with an anti-microsporidia drug, the host is better able to target pathogen cells with ubiquitin.
Because the SCF ubiquitin ligase components cul-6, skr-3 and skr-5 serve to limit N. parisii growth (Figure 2B) we hypothesized that they could be responsible for ubiquitin targeting of parasite cells. Thus, we examined ubiquitin targeting to N. parisii cells in animals that had been treated with cul-6 RNAi compared to the RNAi control (Figure 4F). Indeed, we found that cul-6 RNAi had significantly reduced targeting of ubiquitin to N. parisii cells (two-tailed unpaired t-test, p<0.05). Thus, cul-6 is important for efficient ubiquitylation of parasite-associated proteins, suggesting that cul-6-containing SCF ligases may mediate recognition of N. parisii infection by the host.
The ubiquitin targeting of parasite cells described above was only observed at early timepoints of infection, when pathogen cells were small and mono-nucleate. When the pathogen cells grew bigger and became multi-nucleate meronts, we observed virtually no parasite cells targeted by ubiquitin or by autophagy (data not shown). Similarly, once meronts have differentiated into spores at later stages of infection, we found exceedingly few spores targeted by ubiquitin (Figure 5A). Although there was virtually no specific ubiquitin targeting to the parasite at these later stages of infection, we did observe an increased number of clusters of ubiquitylated proteins (Figure 5B–F). These clusters were dispersed throughout the infected intestinal cells, but in some cases were closely associated with N. parisii, although not encircling the parasite cells (Figure 5D). In addition, we found that infection caused increased clustering of the autophagy marker GFP::LGG-1 in regions distinct from the pathogen cells (Figure S7A–C) and found that GFP::LGG-1 partially colocalized with ubiquitylated protein clusters (Figure S7E–F). In order to determine whether this is a specific response, we examined GFP::LGG-1 upon infection with the extracellular bacterial pathogen P. aeruginosa, and did not find a significant increase in clustering (Figure S7D). Thus, as infection proceeds, an increased amount of conjugated ubiquitin and GFP::LGG-1 clusters accumulate in the host cytosol, and these markers are almost never seen specifically surrounding the pathogen cells.
Recent studies have indicated that host cells monitor the functioning of core processes that are commonly perturbed by pathogen infection and that disruption of these processes can trigger defense-related gene expression by the host [56], [57], [58], [59], [60]. Because intracellular infection by N. parisii leads to an increase in ubiquitylated protein clusters, which may reflect an increase in demand on the UPS, we investigated whether perturbation of the UPS might be responsible for inducing gene expression changes upon N. parisii infection. To conveniently monitor gene expression in vivo and to examine where genes are induced upon N. parisii infection, we made promoter-GFP fusions for C17H1.6 and F26F2.1, two genes of unknown function that are among the most highly upregulated genes at all infection timepoints (eg. at 8 hpi, C17H1.6 and F26F2.1 are upregulated 1.2×1011- and 1441-fold, respectively) (Table S2, S3). Expression of GFP driven by promoters of these genes was strongly induced in intestinal cells of infected animals by 8 hpi and even more robustly by 24 hpi (Figure 6A). These GFP reporters indicated that N. parisii infection drives expression of genes in intestinal cells of infected animals and provided convenient tools for monitoring expression of infection response genes.
To disrupt UPS function, we first performed RNAi knock-down of ubiquitin, pas-5 and rpn-2 in C17H1.6p::gfp and F26F2.1p::gfp transgenic animals. Strikingly, RNAi against the UPS components dramatically induced GFP expression in the intestine in both of these strains (Figure 6B). To confirm these results we performed qRT-PCR and saw levels of endogenous C17H1.6 and F26F2.1 mRNA transcripts also increased by UPS RNAi (Figure 6C). To perturb UPS function pharmacologically, we used the proteasome inhibitor MG-132 and similarly saw that this led to dramatic increase in C17H1.6 and F26F2.1 expression (Figure S8A–C). Because C17H1.6 and F26F2.1 are genes of unknown function, we extended these analyses to genes upregulated by intracellular infection that have predicted function, namely the genes that encode the SCF ubiquitin ligase components skr-3, skr-4, skr-5 and cul-6 (Figure 6C, Figure S8C). Similar to other infection response genes, we found that these genes were also induced by RNAi against the UPS, while another SCF component, skr-1, whose expression was not altered during microsporidia infection, was not affected (Figure 6C, Figure S8C). Thus, C. elegans appears to monitor efficacy of the UPS, and when this core process is disrupted it can trigger expression of a number of specific genes, including SCF components such as cul-6 that are used by C. elegans to limit intracellular infection.
The C. elegans gene expression response to N. parisii was most similar to its response to viral infection, including the upregulation of SCF ligase components (Figure 1E, Table S5, Table S6). Because of this similarity, we investigated whether the SCF ligases implicated in response to N. parisii also played a role in response to viral infection. Indeed, we found that cul-6 RNAi caused a 13-fold increase in viral load, and skr-3 and skr-4 RNAi caused 5- and 4-fold increases in viral load respectively (Figure 7A), indicating that these SCF ligase components promote anti-viral defense. However, contrary to N. parisii infection, global inhibition of the UPS by RNAi-mediated knockdown of UPS components drastically reduced viral replication (Figure 7B). Many viruses exploit host UPS in order to replicate, for example to degrade host RNAi and immune signaling machinery, or to control function and stability of viral proteins [61], [62], [63], [64], [65], and thus the Orsay virus may likewise be hijacking this host pathway. Importantly, this result also suggests that the increased susceptibility to N. parisii infection of UPS-compromised nematodes (Figure 2C) is not likely just a result of general 'sickness' in these animals. Thus, the UPS appears to play two different roles in response to the Orsay virus, involving an unknown ligase(s) that promotes susceptibility to viral infection, and the cul-6, skr-3 and skr-4 SCF ubiquitin ligases promoting anti-viral defense.
Because N. parisii infection caused increased clustering of ubiquitylated proteins in C. elegans intestine, and robust gene expression changes in response to infection appeared to be a reflection of increased demand on the UPS, we investigated whether similar host responses occurred upon viral infection. Indeed, we found that infection with Orsay virus caused clustering of ubiquitylated proteins (Figure 7C), and clustering of GFP::LGG-1 (Figure S7A–D). Thus, infection with the Orsay virus induces similar cell biological changes as N. parisii infection. Furthermore, we found that viral infection induced the GFP reporter F26F2.1p::gfp (Figure 7C), which is also induced when the UPS is perturbed. Thus, it appears that the C. elegans transcriptional response to viral infection, like the response to N. parisii infection, involves surveillance pathways that detect perturbation of the UPS caused by infection, to upregulate defense gene expression.
Based on our results we propose a model for the C. elegans intestinal response to intracellular infection (Figure 8), which highlights an important role for ubiquitin-mediated defense. In response to N. parisii infection, C. elegans upregulates expression of SCF ligase components, which restrict growth of the microsporidian pathogen N. parisii, as well as the Orsay virus. Restriction of N. parisii growth appears to also depend on the proteasome, as well as the autophagy pathway. While SCF ligase components such as CUL-6 have a substantial role in restricting growth of the virus, their more modest role in defense against N. parisii may be due to functional redundancy and/or the relatively inefficient targeting of ubiquitin to this pathogen. Inefficient targeting may be a result of suppression or evasion of host defenses by the parasite, as we find increased ubiquitin targeting of pathogen cells and a greater role for ubiquitin-mediated defense after treatment with drugs that inhibit N. parisii growth. Furthermore, we observe an increase in autophagy machinery targeting to N. parisii cells after activation of autophagy by inhibition of the TOR pathway. Interestingly, the increased demand on the UPS caused by intracellular pathogens like N. parisii and the Orsay virus may induce gene expression in response to infection, because genetic or pharmacological perturbation of the UPS upregulates expression of SCF ligase components and other genes that are induced by these intracellular infections.
SCF ligases comprise one of the major classes of E3 ubiquitin ligases that catalyze transfer of ubiquitin onto substrates. These ligases have very well characterized roles in controlling levels of endogenous proteins that regulate the cell cycle and development. Intriguingly, the expanded and diversified repertoire in C. elegans and plants of ubiquitin ligase adaptors such as F-box and BTB-MATH domain proteins, as well as other SCF components, has led to the hypothesis that these ligases may also be involved in recognition of foreign substrates. Our study with microsporidia and virus infection provides the first experimental support for this hypothesis. In particular, we see that the C. elegans SCF ligase components cul-6 and skr-3, skr-4, and skr-5, mediate a defense response of C. elegans to N. parisii and virus infection. Previous reports indicated that CUL-6 and SKR-3 interact physically in a yeast two-hybrid assay, indicating these components could assemble in vivo to produce a functioning SCF ligase [33]. Moreover, we see targeting of ubiquitin to N. parisii cells that depends on the cul-6 cullin component of the SCF ligase, which could conjugate host ubiquitin onto pathogen proteins or to host proteins that are associated with the pathogen cell. SCF ligases may also be involved in processing of proteins distinct from the pathogen cells, such as virulence factors that are secreted out of the pathogen cell into the cytosol. Another intriguing possibility is that SCF ligases are important for degrading inhibitory host proteins to trigger host innate immunity, analogous to ubiquitin-mediated degradation of IκB in NFκB signaling in mammals. However, the actual signaling proteins in C. elegans would be different because the NFκB transcription factor has been lost in this lineage [66]. Processing of host signaling proteins could occur in the clusters of conjugated ubiquitin we see later during infection, which are not associated with pathogen cells. Indeed, all of these possibilities are not mutually exclusive, and there are likely many roles for the SCF ligases and ubiquitin-mediated responses to intracellular infection in C. elegans.
Our analysis of the gene expression response to N. parisii infection indicated that C. elegans has a very distinct response to this pathogen compared to previously described extracellular pathogens. Responses to extracellular pathogens like S. aureus and P. aeruginosa are marked by upregulation of secreted anti-microbials and detoxifying enzymes [37], [41], [67], [68], [69], which did not comprise a substantial part of the gene sets upregulated by N. parisii. Instead we found enrichment for genes associated with ubiquitylation (Table 1, S9), and that the response to N. parisii shared greatest similarities with the response to Orsay virus infection. The commonality of transcriptional response to these two very distinct pathogens (N. parisii is a eukaryotic organism with 2661 genes and the Orsay virus has only 3 genes) is quite striking, and our data indicate that some genes induced by infection such as SCF ligases can also be induced by perturbation of UPS function. Indeed, inhibition of the proteasome has been shown to induce stress response genes in other C. elegans studies as well [58], [59]. These results fit with the growing theme that C. elegans epithelial defense relies on monitoring of core host processes as an important cue to indicate the presence of pathogen attack [56], [57], [70], [71]. Such surveillance pathways are increasingly appreciated in mammalian defense as well, and may constitute a major mode by which hosts discriminate pathogens from other microbes [72], [73]. It is possible that surveillance of UPS function is responsible for controlling the transcriptional response to intracellular infection, although it is possible that UPS perturbation and infection are distinct triggers that converge to upregulate the same response genes.
Intracellular infection as well as perturbation of the UPS would be expected to cause substantial stress on the protein homeostasis (proteostasis) network of intestinal cells [74], [75], [76]. Intracellular infection by both N. parisii and virus should introduce a suite of foreign proteins into the host cell, may also cause damage to host proteins, and lead to activation of inducible immune responses. Any and all of these physiological changes may cause stress on the protein degradation and/or chaperone/folding systems of the host. This stress could explain the partial overlap we saw between the transcriptional response to intracellular infection and prolonged heat shock, a condition known to disrupt cellular proteostasis, although we saw an upregulation of only two hsp chaperones in response to infection (Table S7). In particular, hsp-16.1, which was significantly upregulated at 64 hpi when animal intestines are filled with parasite spores and large vacuoles (Figure 1A), has been shown to act in the Golgi where it helps to maintain cellular Ca2+ balance and protects cells against necrotic cell death triggered by heat as well as insults unrelated to thermal stress [43]. Further comparison between the responses to UPS stress and intracellular infection will likely shed light on mechanisms of cytosolic quality control and how they regulate defense against intracellular infection.
While ubiquitin-mediated defense does play a role in limiting N. parisii growth, it appears to be only a minor one. There are several reasons that could account for this small effect. First, because UPS components are essential for animal development and overall health we relied on partial knockdown of UPS components to compromise UPS function. Second, in analyses of genes that are not essential, such as SCF ligase components, there may be redundancy in the proteins involved in defense. Third, we anticipate that like other intracellular pathogens [77], [78] (for example the Orsay virus in this study), N. parisii may subvert host ubiquitylation machinery to promote its own growth. In this case, compromised host UPS would negatively impact both the replication of N. parisii as well as the ability of C. elegans to clear infection, yielding a small net change in pathogen load. Lastly, it is possible that N. parisii suppresses or evades the host ubiquitin-mediated defense. Consistent with this idea, C. elegans is better able to target ubiquitin to pathogens and induce their degradation when N. parisii is treated with drugs that slow its growth. Additionally, if N. parisii were suppressing C. elegans ubiquitin-mediated defenses, then genetic inhibition of these processes in the context of infection would only have a minor effect on pathogen resistance, while genetic activation could have a greater effect. Indeed, we found that activating autophagy through RNAi against let-363/TOR led to improved targeting and clearance of N. parisii cells, with a greater effect on resistance than autophagy inhibition. However, it is important to note that let-363/TOR is upstream of several other processes, including protein synthesis [79], which may also account for the increased resistance of this strain.
Other pathogens have been shown to actively suppress ubiquitin-mediated defenses of other eukaryotic hosts [21], [78], [80], [81], [82]. For example, in human cells, the bacterial pathogen Salmonella enterica suppresses ubiquitin-mediated host defenses with the GogB effector, which inhibits a human SCF ligase by interacting with Skp1 and the human F-box only 22 (FBXO22) protein, an interaction that impedes NFκB signaling and limits inflammation in infected cells [83]. Similarly, N. parisii might deploy effectors that block ubiquitylation of meronts, which are in direct contact with the cell cytosol of C. elegans intestinal cells and should be accessible to host ubiquitylation machinery. N. parisii might also evade ubiquitylation by the host by masking or simply lacking host-recognizable cues present during other intracellular pathogen infections. In particular, because N. parisii is itself a eukaryote, it may possess fewer pathogen-associated molecular patterns (e.g. bacterial peptidoglycan or lipopolysaccharide), which can be used by eukaryotes to recognize pathogens.
Microsporidia are increasingly recognized as natural pathogens of nematodes [84], [85], and Nematocida strains in particular have been isolated from multiple wild-caught Caenorhabditis nematodes [12]. It will be interesting to examine the interaction between other Nematocida pathogens and Caenorhabditis hosts to determine whether ubiquitin-mediated defenses have a greater or lesser role in those encounters, as part of the ever-shifting landscape of the host/pathogen arms race. Because microsporidia are obligate intracellular pathogens (which by definition cannot grow outside of host cells), it is imperative that they evade or suppress host defense pathways such as ubiquitylation to propagate the species. Thus suppression or evasion of host defense, together with extremely rapid intracellular replication [34], may be at the heart of why the Microsporidia have grown to be such a large and successful phylum able to infect virtually all animal hosts.
All C. elegans strains were maintained on nematode growth media (NGM) and fed with E. coli strain OP50-1, as described [86]. N. parisii spores were prepared as previously described [87]. Briefly, N. parisii was cultured by infecting large-scale cultures of C. elegans, followed by mechanical disruption of worms and then filtering to isolate spores away from worm debris. The temperature-sensitive sterile strain CF512 fer-15(b26);fem-1(hc17) was used for RNA-seq and other experiments to prevent internal hatching of progeny at later infection time points. This strain was maintained using standard laboratory techniques at the permissive temperature of 15°C and shifted to 25°C for pathogen infection experiments [39]. The DA2123 adIs2122[lgg-1p::gfp::lgg-1] strain was a kind gift from Dr. Malene Hansen [88], [89].
Promoter-GFP fusions for the N. parisii induced genes C17H1.6 and F26F2.1 were made using overlap PCR. Briefly, genomic DNA upstream of the predicted start for these genes was amplified (1273 bp for C17H1.6 and 796 bp for F26F2.1) with PCR and then fused in frame to GFP amplified from pPD95.75. These promoter-GFP fusions were co-injected with the myo-2p::mCherry marker that labels pharyngeal muscle. Several independent transgenic lines carrying extrachromosomal arrays for these fusions were isolated and these lines induced GFP upon infection with N. parisii. One line for each fusion was integrated using psoralen/UV-irradiation to generate the integrated transgenic strains ERT54 jyIs8[C17H1.6p::gfp; myo-2p::mCherry] × and ERT72 jyIs15[F26F2.1p::gfp; myo-2::mCherry].
A GFP-tagged ubiquitin construct pET341 was generated using three-part Gateway recombination by fusing the intestinal-specific vha-6 promoter to GFP at the N-terminus of ubiquitin (amplified from the C. elegans ubq-1 gene), with a unc-54 3'UTR, introduced into destination vector pCFJ150 that encodes for a wild-type copy of C. briggsae unc-119 gene under the control of the unc-119 promoter. This construct was injected into EG6699 ttTi5605 II; unc-119(ed9) III mutant animals and transgenic progeny were recovered, to generate a multi-copy array strain ERT261 jyEx128[vha-6p::gfp::ubiquitin cb-unc-119(+)];ttTi5605 II; unc-119(ed9). Likewise, construct pET346 was generated, which contains a mutant version of ubiquitin without its last two C-terminal glycines. This construct was injected into EG6699 to generate multi-copy array strain ERT264 jyEx131[vha-6p::gfp::ubiquitinΔGG cb-unc-119(+)]ttTi5605 II; unc-119(ed9).
C. elegans infections, RNA isolation, and library construction are previously described [34]. Briefly, synchronized fer-15(b26);fem-1(hc17) L1s were grown for 24 hours at 25°C on 10-cm NGM plates seeded with OP50-1 E. coli and then infected with N. parisii ERTm1 spores. Infected and control C. elegans were harvested at appropriate times and total RNA was extracted using TriReagent (Molecular Research Center, Inc.). RT-qPCR and the Bioanalyzer assessed quality of RNA samples. Strand-specific libraries were constructed using the dUTP second strand marking method [90], [91].
Reads were aligned using Bowtie[92] and transcript abundance estimated using RSEM [93]. Differentially expressed transcripts were identified using the edgeR Bioconductor package (Empirical analysis of digital gene expression data in R, v 3.0.8) [94]. FDR [95] cutoff was set to <0.05, which yielded lists of genes with >4-fold difference in expression. C. elegans reads comprised the majority of the infected sample reads, ranging from over 99% early during infection (8 and 16 hpi) to 71.6% at 40 hpi (Table S1). The progressive reduction in the fraction of C. elegans reads corresponded to replication of microsporidia in the C. elegans intestine resulting in increased contribution of parasite RNA to total RNA of each infected sample [34]. The number of expressed C. elegans genes in all samples ranged from 55.4% (64 hpi) to 62.1% (16 hpi) of the total genome (Table S1). Despite the growing input of parasite RNA, global C. elegans gene expression remained comparable between infected samples and uninfected controls, with the greatest absolute difference (3.61%) in total number of expressed genes, which occurred at 64 hpi (infected vs uninfected control).
Based on previous studies, genes were classified as either intestinal-associated (as determined by fluorescence-activated nuclei sorting) [96], germline-associated (as determined by SAGE) [97], or neither. Very few germline specific/enriched genes were among the differentially expressed genes (Table S2) and therefore we used all genes expressed in germ lines detected by SAGE as the germline-associated class. We then compared the number of differentially expressed genes from each category to the number expected from the classification using the chi-squared test.
Gene Set Enrichment Analysis (GSEA) v2.0 [42] was used to compare gene sets from relevant C. elegans expression studies to our RNA-seq data. The RNA-seq expression dataset file used to generate ranked gene lists (from most upregulated to most downregulated) based on changes in expression between infected and uninfected conditions is summarized in Table S4 while the compiled gene sets used for analysis are described in detail in Table S5. Genes from other studies were converted where necessary to WBGeneIDs according to Wormbase version WS235. Five independent analyses were performed, one for each infection timepoint, with 1000 permutations for each analysis. Results for gene sets with FDR<0.25 and nominal p-value<0.05 were compiled into a graphical representation based on their NES-values, and for gene sets where the NES was not considered significant a value of zero was assigned (Table S6).
Experiments were performed at 25°C and for each condition two biological replicates were included. About 200 synchronized fer-15(b26);fem-1(hc17) L1s were grown on 6-cm plates for two days, feeding on a lawn of E. coli RNAi clones from the Ahringer library or the skr-4 RNAi clone generated through amplification of C. elegans skr-4 genomic sequence (using primers 5′ CCGAATTCGTCTCACGAAAAGTGATC - and 5′- CCGAATTCGGCGTTATACATTTATTCAA) and cloned into the L4440 RNAi vector using EcoRI restriction sites. Animals were then infected with 2 million spores, fixed in 4% paraformaldehyde (PFA) 24 hpi, and stained with MicroB FISH probe against N. parisii rRNA as previously described [12], [34]. Stained animals were mounted on glass slides in Vectashield with DAPI (Vector Laboratories) and imaged using a Zeiss AxioImager microscope with a 10× objective. Exposure times were kept the same for all samples within a single experiment. For all experiments except for ones in Figures 2B, 2G and S4, where a custom fully automatic method for estimating pathogen load written in Matlab was used (see Figure S2 and Supplemental Methods in Text S1), images were analyzed semi-manually using ImageJ software, where the nematode body area, and the area of pathogen contained within were determined using two different thresholds of the MicroB FISH signal (a relaxed threshold to recognize the background staining of the animal body, and a stringent threshold to specifically recognize the pathogen). Due to developmental defects caused by knockdown of UPS components, for experiments targeting the UPS, animals were first grown for one day on E. coli strain OP50-1, and then transferred to plates seeded with UPS RNAi clones diluted with the L4440 RNAi vector control (1∶10 for ubq-2, 1∶5 for pas-5, and 1∶20 for rpn-2). C. elegans has two genes encoding for ubiquitin, ubq-1 and ubq-2. The ubq-2 RNAi clone was chosen for majority of experiments because it had less pronounced developmental defects then animals fed with RNAi against ubq-1 (data not shown). After one day on RNAi, animals were infected and processed as described above. For fumagillin and FUdR experiments, animals were grown, infected, and processed as described above, except at 8 hpi, 0 or 25 µM of fumagillin (Medivet Pharmaceuticals Ltd.) or 0 or 2.6 µg/µL of FUdR (Acros Organics) in 250 µL of M9 with 0.1% Triton-X was spread onto plates containing the animals for a final concentration of 0 to 0.26 µg/mL (fumagillin) and 0 to 59 µg/mL (FUdR) present for the remainder of the experiment (an additional 16 hours).
To quantify ubiquitin colocalization with microsporidia, about 200 synchronized fer-15(b26);fem-1(hc17) L1s were grown on 6-cm plates for 2 days at 25°C, and then were infected with 5 million N. parisii spores. At 8 hpi, the infected animals were treated with 250 µL of 0 µM, 25 µM, or 150 µM of fumagillin in M9 with 0.1% Triton-X (fumagillin final plate concentrations of 0 µg/mL, 0.26 µg/mL, or 1.56 µg/mL). At 12 hpi, animals were anesthetized with 10 mM levamisole, their intestines dissected out, and fixed for 15–30 min in 4% PFA. The intestines were stained with MicroB FISH probe against N. parisii rRNA, followed by staining with FK2 antibody (Millipore), and secondary antibody staining with FITC goat anti-mouse IgG (Jackson ImmunoResearch). Stained intestines were mounted in Vectashield with DAPI (Vector Laboratories) and imaged. For each condition, z-stacks spanning the width of twelve intestines were taken, and colocalization between each imaged parasite cell and the FK2 antibody was determined. All images, unless specified otherwise, were captured using a laser scanning confocal microscope with a 40× oil immersion objective (Zeiss LSM 700, equipped with an AxioCam digital camera and Zen 2010 acquisition software). Images were imported into Adobe Photoshop and assembled using Adobe Illustrator.
For ubiquitin immunofluorescence at different stages of infection, animals were infected with N. parisii as described for RNA-seq. After 30 or 40 hpi, animals were anesthetized with 10 mM levamisole, their intestines dissected out, and fixed for 30 min in 4% PFA. The intestines from the 30 hpi infected and uninfected control samples were stained as described above. Intestines from the 40 hpi infected and control samples were stained directly with antibodies without FISH staining. Stained intestines were mounted in Vectashield with DAPI (Vector Laboratories) and imaged.
To quantify GFP::ubiquitin colocalization with microsporidia, about 200 synchronized ERT261 or ERT264 L1s were grown on 6-cm plates, seeded either with OP50-1 E. coli or control L4440 and cul-6 RNAi clone, for 36 hours at 20°C and then infected with 5 million N. parisii spores. At 10 hpi, the infected animals were treated with 250 µL of 0 µM, 25 µM, or 150 µM of fumagillin in M9 with 0.1% Triton-X, and at 15 hpi animals were fixed in 4% PFA, stained with MicroB FISH probe against N. parisii rRNA, mounted in Vectashield with DAPI, and imaged as described above. For each condition and experiment, z-stacks spanning the width of twenty to eleven ERT261 and seven to ten ERT264 intestines were taken, and colocalization between each imaged parasite cell and GFP was determined. For RNAi experiments, eight to ten ERT261 animals were imaged for each condition and experiment.
For imaging of GFP::ubiquitin in live animals, synchronized ERT261 animals expressing the intestinal GFP::ubiquitin construct were grown and infected at 20°C to minimize ubiquitin aggregate formation in uninfected controls. Synchronized animals were grown for 24 hours on 6-cm plates prior to inoculation with 2 million N. parisii spores and 48 hpi were mounted on agarose pads, anesthetized with 1 mM levamisole and imaged. For quantification of GFP::ubiquitin aggregates, synchronized ERT261 or ERT264 animals were grown at 20°C for 31 hours on 6-cm plates prior to inoculation with 1 million spores. At 10, 30 and 45 hpi, animals were fixed with PFA and stained with MicroB FISH probe as described above. Stained animals were mounted in Vectashield with DAPI and viewed directly with a laser scanning confocal microscope with a 40× oil immersion objective (Zeiss LSM 700).
To image promoter-GFP reporter strains, synchronized ERT54 and ERT72 L1s were grown for 24 hours at 25°C and infected with 10 million N. parisii spores on 6-cm plates. Infected and control worms were anesthetized with 1 mM levamisole, mounted on agar pads, and imaged at 8 and 24 hpi using a Zeiss AxioImager microscope. For RNAi experiments, synchronized ERT54 and ERT72 L1s were grown for 48 hours at 20°C on plates seeded with RNAi clones and imaged as described above. For MG-132 experiments, synchronized ERT54 and ERT72 L1s were grown for 24 hours at 20°C, incubated on a nutator at room temperature for six hours in M9 with 0.1% Triton-X and 0 µM, 500 µM, or 1mM MG-132, and then imaged as described above.
To measure endogenous mRNA expression changes due to UPS component knockdown, synchronized fer-15(b26);fem-1(hc17) L1s were grown at 20°C for 48 hours on RNAi bacteria, and then collected in TriReagent (Molecular Research Center, Inc.) for RNA extraction. To measure endogenous mRNA expression changes due to pharmacological proteasome inhibition, synchronized fer-15(b26);fem-1(hc17) L1s were grown 24 h at 20°C, incubated on a nutator at room temperature for six hours in M9 with 0.1% Triton-X and 0 µM or 500 µM MG-132, and then collected in TriReagent for RNA extraction. RNA extraction, reverse transcription, and qRT-PCR were performed as previously described [41]. qRT-PCR primer sequences are available upon request. Each biological replicate was measured in duplicate and normalized to the snb-1 control gene, which did not change upon conditions tested. The Pffafl method was used for quantifying data [98].
Virus stock for infections was prepared as described previously [5], with minor modifications. Briefly, the virus-susceptible rde-1(ne219) nematodes were grown in large-scale cultures until just starved, mechanically disrupted, and filtered through a 0.2 µm filter to separate the virus away from nematode debris. When spread on a 6-cm plate in a 250 µL volume, the 1∶50 dilution of this filtrate was the maximum dilution tested that turned on the F26F2.1p::gfp reporter in all animals 24 hpi at 25°C (data not shown). These conditions were used for all viral infections. To measure changes in viral load upon RNAi-mediated knockdown of C. elegans genes of interest, the viral RNA1 levels were measured using primers GW195 and GW194 [5] and compared to those found in L4440 controls. For these experiments, fer-15(b26);fem-1(hc17) animals were grown and treated with RNAi the same as for the N. parisii pathogen load experiments, except about 300 synchronized L1 animals were used per 6-cm plate and following 24 hours of infection with the virus, animals were collected for RNA extraction and qRT-PCR. Intestine dissections from ERT72 animals and immunofluorescence with FK2 anti-conjugated-ubiquitin antibody were performed as described above, except the secondary antibody used was the Cy3 goat anti-mouse IgG (Jackson ImmunoResearch).
C. elegans genes analyzed:
cul-6, skr-3, skr-4, skr-5, ubq-1, ubq-2, pas-5, rpn-2, lgg-1, lgg-2, atg-18, sqst-1, let-363, C17H1.6, F26F2.1, skr-1, C17H1.14, F26F2.4, Y39G8B.5, sdz-6, T08E11.1, W08A12.4, ZC196.3, Y94H6A.2, his-10, his-16
RNA-seq data are part of NCBI BioProject #PRJNA163569.
|
10.1371/journal.pntd.0001135 | Chagas Disease among the Latin American Adult Population Attending in
a Primary Care Center in Barcelona, Spain | The epidemiology of Chagas disease, until recently confined to areas of
continental Latin America, has undergone considerable changes in recent
decades due to migration to other parts of the world, including Spain. We
studied the prevalence of Chagas disease in Latin American patients treated
at a health center in Barcelona and evaluated its clinical phase. We make
some recommendations for screening for the disease.
We performed an observational, cross-sectional prevalence study by means of
an immunochromatographic test screening of all continental Latin American
patients over the age of 14 years visiting the health centre from October
2007 to October 2009. The diagnosis was confirmed by serological methods:
conventional in-house ELISA (cELISA), a commercial kit (rELISA) and ELISA
using T cruzi lysate (Ortho-Clinical Diagnostics) (oELISA).
Of 766 patients studied, 22 were diagnosed with T. cruzi
infection, showing a prevalence of 2.87% (95% CI,
1.6–4.12%). Of the infected patients, 45.45% men and
54.55% women, 21 were from Bolivia, showing a prevalence in the
Bolivian subgroup (n = 127) of 16.53%
(95% CI, 9.6–23.39%).
All the infected patients were in a chronic phase of Chagas disease:
81% with the indeterminate form, 9.5% with the cardiac form
and 9.5% with the cardiodigestive form. All patients infected with
T. cruzi had heard of Chagas disease in their country
of origin, 82% knew someone affected, and 77% had a
significant history of living in adobe houses in rural areas.
We found a high prevalence of T. cruzi infection in
immigrants from Bolivia. Detection of T.
cruzi–infected persons by screening programs in non-endemic
countries would control non-vectorial transmission and would benefit the
persons affected, public health and national health systems.
| Chagas disease is a parasitic infection caused by the protozoan
Trypanosoma cruzi, and is becoming an emerging health
problem in non-endemic areas because of growing population movements. The
clinical manifestations of chronic T. cruzi infection include
the latent form (the indeterminate chronic form), the cardiac form, the
digestive or cardiodigestive form, and sudden death. Therefore, many diagnoses
of Chagas disease are based on epidemiological suspicion rather than on clinical
signs and symptoms. This study showed that the prevalence of Chagas disease in
Latin American patients attending at a health center in Barcelona is
2,87% and the highest prevalence was found among Bolivian patients
(16,53%). All the infected patients were in a chronic phase of Chagas
disease. Detection of T. cruzi–infected persons by
screening programs in non-endemic countries would control non-vectorial
transmission and would benefit the persons affected, public health and national
health systems. The data obtained in this study and the experiences described
elsewhere suggest that it is advisable to perform Chagas disease screening in
non-endemic countries on all patients from continental Latin America who:
(1)have a suggestive epidemiologic history, (2)are pregnant, (3)are
immunosuppressed, (4)have symptoms suggestive of Chagas disease, or (5)request
screening.
| Trypanosoma cruzi (T. cruzi) is a flagellate
protozoan that causes Chagas disease (CD). It is traditionally linked to rural areas
of continental Latin America, where it is transmitted by a variety of bug vectors.
In recent decades, the epidemiological pattern of this disease has undergone
considerable changes [1]. In the endemic countries of Latin America, the regional
Chagas programs are working to interrupt vector-borne and transfusional
transmission, to control congenital Chagas disease and to support initiatives aimed
at improving diagnosis, management and surveillance of the disease [2]. In non-endemic
countries that receive immigrants from Latin America or send tourists to endemic
areas, CD is an emerging disease and has become a public health problem because it
can be transmitted by non-vectorial mechanisms [3], [4].
Spain is a major European host country for people from Latin America. According to
the Spanish National Institute of Statistics, in 2009 more than 1.8 million
immigrants from Latin America were registered, accounting for 3.85% of the
total population [5]. In recent years several studies of CD in non-endemic
countries [6]–[9] have focused in particular on non-vectorial transmission
mechanisms such as pregnancy and childbirth [10], [11], blood transfusion [12], [13] and organ
transplantation [14],
[15].
However, when reviewing the literature we found little information on imported CD in
non-endemic countries at the primary care level [16], which is ideal for
screening the general population [17].
The clinical manifestations of chronic T. cruzi infection include
the latent form (the indeterminate chronic form), which occurs in 60% of
cases [18], the
cardiac form [19],
the digestive or cardiodigestive form, and sudden death [20]. Therefore, many diagnoses of CD
are based on epidemiological suspicion rather than clinical signs and symptoms.
The objectives of the present study were (1) to assess the prevalence of
Trypanosoma cruzi infection in the adult Latin American
population treated at a health center in Barcelona, Spain; (2) to analyze the
clinical phase of the disease; and (3) to determine whether screening for imported
CD in primary care should be recommended.
We performed an observational, cross-sectional prevalence study at the health
center of the Clot district, Barcelona. This center serves a population of 25442
people, with a total foreign population of 13.5% and a Latin American
population of 6.3% (according to the 2008 census of the Barcelona City
Council) [21].
The staffs participating in the study were 14 general practitioners, 13 nurses,
1 gynecologist and 1 midwife.
The study protocol was approved by the Ethical Committee of the Jordi Gol
Institute for Research in Primary Care of the Institut Català de
la Salut (Catalan Health Institute).
Written informed consent was requested from all participants. When participants
were children, their parents/guardians provided informed consent.
During the period October 2007 to October 2009, all patients from continental
Latin America under 14 years of age who presented at the health center for any
health reason were invited to participate in the study. After obtaining informed
consent, we collected clinical and epidemiological data.
We ascertained the reasons why the patients visited their doctor/nurse by
reviewing the electronic patient charts. On a patient's first visit to a
primary care centre the Preventive Activities and Health Promotion Program
(PAPPS) is initiated [22], and at the Clot health center this program included
CD screening of all persons originating from continental Latin America.
Serological screening was performed with an immunochromatographic test (ICT) that
uses recombinant antigens of T. cruzi (TcD, TcE, PEP-2 and
SAPA) on whole blood collected by finger prick.
If the screening was positive, a venous blood sample was collected to confirm the
diagnosis at the Parasitology Laboratory of the Faculty of Pharmacy, University
of Barcelona. We used 2 enzyme-linked immunosorbent assay (ELISA) methods: a
conventional, in-house ELISA (cELISA) with whole T. cruzi
epimastigote antigens [23] and a commercial kit with the recombinant antigens
TcD, TcE, PEP-2 and TcLo1.2 (rELISA).
In accordance with international criteria established by the World Health
Organization, sera that were reactive in two serological methods were considered
positive [24].
Positive results were confirmed by a third ELISA using T cruzi
lysate (Ortho-Clinical Diagnostics) (oELISA). In a subsample of 101 patients we
performed the ELISA serologies (cELISA and rELISA) regardless of the result of
the ICT, in order to test the usefulness of this test in screening for CD in
primary care [25].
All patients infected with T. cruzi were referred to the
Tropical Medicine Unit of Hospital Clínic de Barcelona and clinically
evaluated by a complete review of the epidemiologic history and consistent
symptoms/signs, a general physical examination and an electrocardiogram. If the
electrocardiogram was pathologic, it was assessed with an echocardiogram or
24-hour Holter according to the disease detected. If symptoms consistent with
gastrointestinal involvement were detected [26], an esophageal, gastric and
duodenal transit assessment or a barium enema was performed. Benznidazole (5
mg/kg/day for 60 days) was offered to all patients aged 18–50 years
without advanced Chagas cardiomyopathy and no other contraindication for start
benznidazole (pregnancy, severe renal or hepatic insufficiency [27]. None of them
refused to start it.
The sample size was calculated for an alpha level of 0.05 and a precision of
±0.05% in a bilateral comparison, assuming maximum uncertainty
(50% prevalence); for a population of 1516 subjects [28] a random sample of 758 was
necessary.
The programs SPSS version 17.0 and Epidat version 3.1 were used for the
statistical analysis. The χ2 test was used to compare hypotheses of
independence between two categorical variables and the Student
t test for continuous variables. The confidence interval
for all hypothesis comparisons was 95% and the tests were 2-tailed.
A total of 766 persons from continental Latin America were included in the study. The
epidemiological data are presented in Table 1 and the countries of origin in Table 2.
Of the 766 patients analyzed, 27 were reactive to the ICT and 20 of these were
reactive in cELISA, rELISA and oELISA. Also, 2 patients of the 101 tested by ICT,
cELISA, and rELISa regardless of the result of the first were reactive in cELISA and
rELISA. Both were also positive in oELISA, so they were considered positive.
A total of 22 patients were diagnosed with CD, corresponding to a prevalence of
2.87% (95% CI, 1.6–4.12%) in the sample studied. Of
these, 21 were from Bolivia; the prevalence of CD in the subgroup of Bolivian
patients studied (n = 127) was 16.53% (95% CI,
9.6–23.39%). The remaining patient was from Paraguay.
All the patients infected by T. cruzi were in the chronic phase of
CD. The clinical form and the reasons why they visited the health center are
presented in Table 3. Four
patients (18.2%) had been previously diagnosed in the country of origin, but
none of them mentioned it in the primary care visit because they thought it was a
health problem proper to their country that would be unknown to the Spanish health
staff (this information was obtained when they were asked for informed consent to
participate in the study). None of them were aware of their clinical phase and 2
patients had received incomplete treatment.
The prevalence of T. cruzi infection in the sample studied was
2.87% and in the subgroup of Bolivian patients it was 16.53%. In the
medical literature we found few studies of similar characteristics to ours
(involving screening of the adult Latin American population in primary care) and
their results varied [29], [12] due to the heterogeneity of the populations analyzed and
the distribution of CD in Latin America.
The laboratory confirmation of a clinical suspicion of CD is based on consistent
results of at least 2 different immunological tests [24]. ICTs are attractive in primary
care because they are easy to use in routine clinical practice and do not require
sophisticated facilities or specialized staff. In the substudy that we performed in
101 patients [25], for the ICT used we found a sensitivity of 92.5%
and a specificity of 96.8%. Other studies have evaluated the sensitivity and
specificity of ICTs [30], [31] with similar results. The current sensitivity of ICTs
must be increased so that they can be used as effective screening tests. Meanwhile,
they should be combined with other methods that offer greater sensitivity [25], [31].
The highest prevalence was found among Bolivian patients, in agreement with other
studies performed in Spain [9], [16] and other non-endemic countries [12], [32]. No cases were diagnosed among
the Peruvian or Ecuadorian patients, who formed 45% of the sample, probably
due to the heterogeneous distribution of CD in endemic countries and the lower
seroprevalence of T. cruzi estimated in Peru (0.69%) and
Ecuador (1.74%) [33].
An epidemiologic history of having lived in rural areas and/or adobe houses showed a
significant relationship with T. cruzi infection, consistent with
the dominant vector-borne transmission mechanism in the countries of origin. All
patients with T. cruzi infection had heard of CD in their countries
of origin and approximately 82% knew someone who was affected. These data
should be taken into account for establishing CD screening criteria in immigrants
from endemic zones, because mere knowledge of the disease may be considered as an
indirect indicator of its presence in the region of origin.
In non-endemic countries CD screening programs have been aimed at particularly
susceptible groups: in blood banks (France [34], the USA [35] and Spain [36]), and in
pregnant Latin American women and their neonates (the Spanish autonomous communities
of Catalonia [37] and Valencia [38]). In our study only
9.5% of the patients with T. cruzi infection had visited the
health center due to clinical symptoms suggestive of CD. As it is a silent disease
that has recently appeared in non-endemic countries, we stress the importance of
establishing in these countries health screening programs based on compatible
epidemiologic history among the general immigrant population from endemic areas.
These programs should be multidisciplinary [3], supported by the best
scientific evidence possible, and promoted by the health authorities.
In non-endemic countries, detecting persons infected by T. cruzi is important in
order to control the transmission (vertical, by transfusion, or by organ
transplant), reduce reactivations in immunodepressed persons, and delay the onset of
the chronic cardiac form through antiparasite treatment [39], all of which have a great
impact on the persons affected, on public health, and on health systems.
Nevertheless, the best solution for CD is a combination of treatment and prevention
in endemic countries [40], [41], where many programs and initiatives are underway [2], [42].
The data obtained in this study and the experiences described elsewhere [4], [12], [14], [16], [20], [27] suggest that it
is advisable to perform CD screening in non-endemic countries on all patients from
continental Latin America who: (1) have a suggestive epidemiologic history (having
lived in a rural area, in adobe houses or having knowledge of CD in the country of
origin), (2) are pregnant, (3) are immunosuppressed, (4) have symptoms suggestive of
CD, or (5) request screening.
|
10.1371/journal.pntd.0006679 | Comparative proteomics of the two T. brucei PABPs suggests that PABP2 controls bulk mRNA | Poly(A)-binding proteins (PABPs) regulate mRNA fate by controlling stability and translation through interactions with both the poly(A) tail and eIF4F complex. Many organisms have several paralogs of PABPs and eIF4F complex components and it is likely that different eIF4F/PABP complex combinations regulate distinct sets of mRNAs. Trypanosomes have five eIF4G paralogs, six of eIF4E and two PABPs, PABP1 and PABP2. Under starvation, polysomes dissociate and the majority of mRNAs, most translation initiation factors and PABP2 reversibly localise to starvation stress granules. To understand this more broadly we identified a protein interaction cohort for both T. brucei PABPs by cryo-mill/affinity purification-mass spectrometry. PABP1 very specifically interacts with the previously identified interactors eIF4E4 and eIF4G3 and few others. In contrast PABP2 is promiscuous, with a larger set of interactors including most translation initiation factors and most prominently eIF4G1, with its two partners TbG1-IP and TbG1-IP2. Only RBP23 was specific to PABP1, whilst 14 RNA-binding proteins were exclusively immunoprecipitated with PABP2. Significantly, PABP1 and associated proteins are largely excluded from starvation stress granules, but PABP2 and most interactors translocate to granules on starvation. We suggest that PABP1 regulates a small subpopulation of mainly small-sized mRNAs, as it interacts with a small and distinct set of proteins unable to enter the dominant pathway into starvation stress granules and localises preferentially to a subfraction of small polysomes. By contrast PABP2 likely regulates bulk mRNA translation, as it interacts with a wide range of proteins, enters stress granules and distributes over the full range of polysomes.
| Poly(A)-binding proteins (PABPs) bind to the poly(A) tails of eukaryotic mRNAs and function in regulating mRNA fate. Many eukaryotes have several PABP paralogs and the current view is that each PABP binds a specific subset of mRNAs. Trypanosoma brucei has two PABPs, and to understand the differential functionality of these paralogs we identified interacting proteins for each. We found unique interactors for both PABPs, and significant differences between the two interaction cohorts. Our data indicate that the two PABP paralogs of trypanosomes have very distinct roles in mediating mRNA fate.
| Gene expression is regulated by multiple transcriptional and post-transcriptional mechanisms. At the post-transcriptional level, regulation of protein synthesis by modulation of translation initiation is a major contributor. The first step in mRNA cap-dependent translation initiation is assembly of the eIF4F complex at the m7G cap of the mRNA 5’ end [1]. The eIF4F complex consists of a large (~180kDa) scaffold protein, eIF4G, bound to the cap-binding protein eIF4E and an RNA helicase, eIF4A. The latter is involved in secondary structure unwinding of the target mRNA, facilitating 40S subunit scanning, together with a further factor eIF4B. Significantly, eIF4G and eIF4B directly interact with the poly(A)-binding protein (PABP) associated with the poly(A) tail at the 3’ end of the target mRNA, to increase translation efficiency by mRNA circularisation and ribosome recycling.
Most higher eukaryotes have several paralogs of eIF4F complex subunits [2] and PABP [3]; increasing evidence suggests that these different paralogs can assemble into distinct eIF4F complexes, facilitating modulation of translation to distinct environmental and developmental conditions [4]. For example, in metazoa there is one eIF4F complex specialised to mediate cap-dependent translation under low oxygen conditions [5,6], and specific eIF4F complexes select distinct sets of mRNAs during development in C. elegans germ cells [7]. The specific functions of distinct eIF4F complexes are mediated by the properties of the individual subunits, for example H. sapiens eIF4E paralogs differ in their ability to localise to P-bodies and stress granules [8], ribonucleoprotein granules (RNA granules) with important functions in mRNA storage, regulation and quality control [9].
The presence of multiple PABP paralogs further increases the combinatorial complexity of this system. Arabidopsis thaliana has eight PABP paralogs [10] that differ in domain structure and expression patterns, with both overlapping and distinct functions [10–16]. Xenopus laevis has three paralogs that are all independently essential [17]. Many protozoa also possess several paralogs of each of the eIF4F complex subunits, but these are the product of lineage-specific expansions and hence unrelated to the paralogs found in higher eukaryotes. Very little is known about their specific functions [18,19].
Kinetoplastids, including the animal and human pathogens Leishmania, Trypanosoma cruzi and T. brucei, rely almost completely on post-transcriptional gene regulation [20]. mRNAs are transcribed poly-cistronically and processed by trans-splicing of a miniexon to the 5’ end, a process coupled to polyadenylation of the upstream transcript [21–26]. Furthermore, the mRNA cap structure is a highly unusual type four, with ribose 2’-O methylations at the first four transcribed nucleotides (AACU) and additional base methylations at the first (m26A) and fourth (m3U) positions [27,28]. This cap requires a kinetoplastid-specific decapping enzyme for degradation [29]. Hence, translational control is a major contributor to gene regulation [30]. As a possible consequence of this kinetoplastids possess a large number of translation initiation factor paralogs [31]: six for eIF4E (eIF4E1-6), five for eIF4G (eIF4G1-5) and two for eIF4A (eIF4A1-2), of which only one, eIF4A1, is known to be involved in translation [32]. Trypanosomes have two PABP paralogs (PABP1, PABP2), while Leishmania has an additional paralog (PABP3).
Multiple studies have addressed the composition of kinetoplastid translation initiation complexes, and whilst data are equivocal in some cases, several distinct eIF4F complexes were described (recently reviewed in [31]. The best characterised complex comprises eIF4E4, eIF4G3, eIF4A1 and PABP1 in both Leishmania and trypanosomes [33–37]. Evidence of a direct physical interaction between eIF4E4 and eIF4G3 was obtained in L. major using yeast two hybrid [37], but direct binding between LmPABP1 and eIF4G3 was not observed [35,37]. Instead, LmPABP1 interacted directly with eIF4E4, mediated by the non-conserved N-terminal extension of eIF4E4 [35], an interaction critical for the function of eIF4E4 [38]. The current assumption that eIF4E4/eIF4G3/PABP1 is the major translation initiation complex is predicated on the following: i) all proteins are of high abundance, ii) PABP1 has greater specificity for poly(A) than PABP2 [36,39], iii) eIF4E4 binds the type 4 cap with the highest affinity of all eIF4E4 paralogs [40–42] and iv) silencing of eIF4E4, eIF4G3 and PABP1 in at least some T. brucei life cycle stages is lethal [33,34,36] and eIF4E4 cannot be deleted in L. infantum [38].
At least three additional translation initiation complexes are known. The first consists of eIF4E5 bound to either eIF4G1 or eIF4G2 [41]. Two further proteins specifically interact with the eIF4G1 version of this complex: TbG1-IP (Tb927.11.6720) and TbG1-IP2 (Tb927.11.350). TbG1-IP is an mRNA cap guanine-N7 methyltransferase, suggesting involvement in nuclear mRNA capping [43], but such a function is unlikely, as the protein is cytoplasmic and localises to starvation stress RNA granules [44], and nuclear cap methylation is known to be performed by TbCGM1 [45,46]. TbG1-IP2 is an RNA binding protein with unknown function. The second complex consists of eIF4G5, which specifically interacts with eIF4E6 and one further protein TbG5-IP (Tb927.11.14590) [42]. Interestingly, similarly to TbG1-IP1, this protein contains a nucleoside triphosphate hydrolase and a guanylyltransferase domain in common with enzymes involved in cap formation. The third complex consists of eIF4G4, eIF4E3 and eIF4A1 [33]. However, neither PABP was identified in any of these complexes.
Several studies have directly addressed function, substrate specificity and localisation of kinetoplastid poly(A)-binding proteins. PABP1 and PABP2 are highly abundant and in excess of the total number of mRNA molecules, at least in the procyclic life cycle stage of T. brucei [36]. RNAi in T. brucei revealed that both isoforms are essential [36] and both isoforms stimulate translation when tethered to the 3’ end of a reporter mRNA [47,48]. Both PABPs are cytoplasmic in untreated cells, but differentially localise under stress conditions: PABP2, but not PABP1, localises to the nucleus under certain conditions [36,49] and only PABP2 localises to starvation stress granules [49,50], while PABP1 and its interacting partners eIF4E4 and eIF4G3 do not [49]. Both PABPs localise to polysomes [49,51], but PABP1 is mainly located in small polysomes while PABP2 is more equally distributed across all polysomes [49]. There is some evidence that PABP2 may have a function unrelated to poly(A) binding. PABP2 binds poly(A) with lower specificity (in comparison to PABP1) in Leishmania [36,39] and binds to the CAUAGAAG element present in cell-cycle regulated mRNAs of Crithidia fasiculata [52] and to the U-rich RNA binding protein UBP1 [53], which mediates instability of the T. cruzi SMUG mucin mRNA [54].
To probe for distinct roles of PABPs we examined their protein interactomes in T. brucei procyclic forms. PABP1 co-precipitates eIF4E4 and eIF4G3 and RNA-binding protein RBP23, but few additional proteins. In contrast, PABP2 co-precipitated a large number of RNA binding proteins, including all proteins that co-precipitated with PABP1 except RBP23. Most eIF4F paralogs co-precipitated with PABP2, most significantly the eIF4G1/eIF4E5 complex and its two interacting partners TbG1-IP and TbG1-IP2. These data, together with analysis of the localisations of PABP1 and PABP2 complex components challenge the current paradigm that PABP1 is the major poly(A)-binding protein in trypanosomes and an alternative model is discussed.
To isolate PABP complexes we used two previously published cell lines expressing C-terminal eYFP fusions of each PABP paralog from their endogenous locus; the second allele remained unaltered [49]. PABP1-eYFP is fully functional as deletion of the wild type allele has no phenotype, while RNAi that targets both alleles is lethal. In the cell line expressing PABP2-eYFP, the second allele could not be deleted, but cells have normal growth rates and localisation of the protein to various types of RNA granules was indistinguishable from that determined with specific antiserum against PABP2 [49]. This indicates that most functions of PABP2-eYFP are essentially identical to the wild type protein. Protein expression and localisation to the cytoplasm was demonstrated by fluorescence microscopy (Fig 1A). Wild type cells served as negative controls.
Cultures of each cell line were snap-frozen and subjected to cryomilling to generate a powder [55]. Aliquots of this powder were used to systematically optimise conditions for isolation of PABP complexes. Protein complexes were captured with polyclonal anti-GFP antibodies covalently coupled to magnetic Dynabeads and analysed by SDS-PAGE. In the optimised protocol, the cell powder was solubilised using CHAPS detergent with two different buffers: a low salt buffer and a high salt buffer, the latter contained 150 mM KCl but was otherwise identical to the low salt buffer. Coomassie-stained gels revealed clearly visible PABP bait proteins and several protein bands specific to one or both PABPs, but absent from the control pull down (Fig 1B).
For each cell line, protein complexes were isolated in two independent experiments for each buffer condition and the captured proteins analysed by liquid chromatography tandem mass spectrometry (LC-MS2) and subjected to label free quantification using MaxQuant [56]. 1901 distinct protein groups (peptides assigned to a specific coding sequence, but where these cannot be assigned to a single gene in the case of close paralogs) were identified (S1 Table); this list was reduced to 1224 after removing all protein groups with less than three unique peptides (S1B Table). For each protein group from each experiment we determined the enrichment ratio in relation to the wild type control cell line, based on quantification by unique peptides only. To avoid division by zero, a constant (0.001) was added to each LFQ value; such ‘infinite ratios’ are clearly distinguishable from genuine ratios by being significantly larger, smaller or exactly 1.0 (S1B Table).
For PABP1, we identified 25 proteins at least two-fold enriched in each of the two low salt replicates (S1C Table) and 66 proteins at least two-fold enriched in both high salt replicates (S1D Table). For PABP2, 77 and 170 proteins were enriched in both replicates under low salt and high salt conditions, respectively (S1E and S1F Table). Ribosomal proteins were exclusively co-precipitated under high salt conditions and not detected under low salt, consistent with intact ribosomes requiring physiological potassium concentrations and dissociating upon potassium depletion [57,58]. Interestingly, the number of co-precipitated ribosomal proteins differed between the PABP1 and PABP2 pull-downs: 43 proteins were co-purified with PABP2 (25% of all precipitated proteins), but only 7 ribosomal proteins with PABP1 (11% of all precipitated proteins). This could reflect differences in polysomal association between the two isoforms: PABP2 associates with heavier sucrose fractions than PABP1 on polysome fractionation gradients [49]. Alternatively, these differences could be explained by the RNA-binding ability of PABP2 being less specific to poly(A) tails in comparison to PABP1, as has been previously found for Leishmania orthologues [36]: unspecific binding of PABP2 to ribosomal RNA could cause co-precipitation of intact ribosomes under high salt conditions, resulting in the presence of ribosomal proteins in the proteomics data. Evidence for the second hypothesis is provided by the large number of nucleolus-localised proteins in the PABP2 pull-down with high salt buffer: 20 of the 127 non-ribosomal proteins purified with PABP2 are known to entirely or predominantly localise to the nucleolus, in comparison to only 4 of 59 non-ribosomal proteins purified with PABP1 [59]. PABP2 does not localise to the nucleolus, at least not to detectable levels, thus, these interactions are likely non-physiological.
All PABP interacting proteins were judged for their possible function in mRNA metabolism. A protein was classified as having a known role in mRNA metabolism (indicated as ‘yes’ in S1C–S1F Table), if it possesses an RNA-binding domain, or if there is direct experimental evidence for involvement in RNA metabolism (for example validated localisation to RNA granules). A protein was classified as having a predicted role in mRNA metabolism (indicated as ‘(yes)’ in S1C–S1F Table) if it was identified in one out of three large scale experiment that screened for posttranscriptional activators, repressors and RNA-binding proteins [47,48], without further experimental validation. The low salt precipitations contained mostly proteins with a known or predicted function in mRNA metabolism for both PABP1 (19/25 proteins) and PABP2 (62/77 proteins) and few obvious contaminants. High salt precipitations were still enriched in mRNA metabolism proteins (PABP1 28/66 and PABP2 54/170) but also contained a large fraction of likely or obvious contaminants, including mitochondrial, nucleolar and ribosomal proteins.
To obtain a high confidence list, we filtered for proteins that were at least two-fold enriched in all four experiments. In a second step, all protein groups with more than one infinite ratio were removed, and three further proteins were manually removed because they were obvious contaminations; two mitochondrial RNA-binding proteins (Tb927.7.2570, Tb927.2.3800) and one glycosomal protein (Tb927.10.5620). Average enrichment ratios were calculated for each protein, excluding ‘infinite ratios’ (S1G Table, Fig 2A) together with a PAPB1/PABP2 enrichment ratio, to determine the specificity of each interaction (S1G Table, Fig 2B).
All 27 PABP-interacting proteins have a known or predicted function in mRNA metabolism. 12/27 proteins were more than 2-fold enriched in both pull-downs. For 6 of the 27 proteins the interaction with PABP(s) had been independently validated in at least one of the Kinetoplastids: ALBA1-3 co-precipitate both T. brucei PABPs [60]. Both T. brucei PABPs were found in a large scale yeast 2-hybrid screen to interact with PBP1 [61]. Several studies have identified PABP1 as part of the eIF4G3/eIF4E4 complex in Leishmania [34–36], with an unusual direct interaction between PABP1 and eIF4E4 [35]. For Leishmania eIF4E4, additional interaction with PABP2 was shown [35]. Moreover, while this manuscript was in revision, a PABP1 interactome for Leishmania infantum was published [62] and is in agreement with our data: seven proteins consistently co-precipitated with Leishmania PABP1, of which six correspond to the six most enriched proteins in the T. brucei PABP1 pulldown (eIF4E4, eIF4G3, PABP1, RBP23, Tb927.7.7460, ZC3H41) and only one protein (Tb927.10.13800) was not identified with our conditions. As a further control, we performed reverse pull-downs of the proteins mostly enriched in either the PABP1 pull-down (eIF4E4) or the PABP2 pull-down (G1-IP2) (Fig 2C). For this, eIF4E4 and G1-IP2 were expressed as eYFP fusion proteins from their endogenous loci, in cell lines also expressing PABP1 or PABP2 C-terminally fused to a tandem of four Ty1 epitopes. Precipitations of eIF4E-eYFP and G1-IP2-eYFP were performed as above, using low salt buffer conditions. Co-precipitated PABP-Ty1 proteins were detected by western blot probed with anti-Ty1. Both PABP proteins were enriched in these pull-downs in comparison to the negative control. However, in agreement with the mass spectrometry, PABP1 had a much higher enrichment ratio than PABP2 in the eIF4E4 pull-down (64-fold/3-fold on average for PABP1/PABP2, n = 2) while the opposite was found for the G1-IP2 pull-down (4-fold/19-fold on average for PABP1/PABP2, n = 2).
Two proteins were particularly enriched in the PABP1 pull-down, with a more than 100-fold enrichment and more than 20-fold enrichment against PABP2. These are the two known PABP1 interactors eIF4E4 and eIF4G3, confirming the specificity of the pull-down. Only three further proteins had average enrichment ratios of >10 in comparison to the negative control, namely the RNA binding protein RBP23, a hypothetical protein Tb927.7.7460 and the CCCH type zinc finger protein ZC3H41; all experimentally uncharacterised. RBP23 was the only protein that was solely precipitated with PABP1: all other PABP1 interacting proteins also interact with PABP2, albeit in most cases with lower enrichment ratios.
In contrast, PABP2 co-precipitated a larger number of proteins than PABP1, but with much lower enrichment ratios, possibly reflecting greater promiscuity and interactions with a larger number of heterogenic target mRNAs and hence likely representing isolation of multiple PABP2 complexes. Of the seven proteins most specific to the PABP2 pull-down, three were members of the previously characterised eIF4G1/eIF4E4 complex [41], namely eIF4G1, the RNA-binding protein Tb927.11.350 (G1-IP2) and Tb927.11.6720, an mRNA cap guanine-N7 methyltransferase. One of the specific PABP2 targets with high enrichment ratio, CBP110, is localised to the nucleoplasm [63] (Fig 3B); this could be a true interacting protein given that PABP2 shuttles between the nucleus and the cytoplasm [36,49]. Among the PABP2 interacting proteins were 14 proteins that had enrichment rates of <2 in the PABP1 pull-down and thus appeared specific to PABP2 (Fig 2B).
As we observed major differences between the two PABPs in localisation to RNA granules, we analysed the localisation to RNA granules for all 27 proteins that interact with either or both PABPs (Fig 2B, S1G Table). We used published data that used either DHH1, PABP2 or poly(A) as stress granule markers [44,49,60,64] or co-expressed several proteins as eYFP fusions with a mChFP fusion of the stress granule marker protein PABP2 (Fig 3 and S1 Fig). In addition, we obtained information from the genome-tagging project TrypTag [59] (http://www.tryptag.org, with permission). For the TrypTag project, cells are washed in amino acid free buffer prior to imaging and starvation stress granules are therefore visible. The majority of proteins (20/27) localised to RNA granules, for one protein the localisation remains unknown as tagging failed, and only six proteins did not localise to RNA granules. At least four of the five proteins mostly enriched in the PABP1 pull-down were excluded from granules; these include the unique PABP1-interacting protein RBP23 (Fig 3A and S1D Fig). In contrast, for the majority of the PABP2-interactors there is evidence or proof for stress granule localisation. Only two proteins of the PABP2 interacting proteins are excluded from granules, one is the nuclear protein CBP110, which is not expected to localise to RNA granules and the other the zinc finger protein ZC3H28. Thus, the PABP1 complex appears largely excluded from granules, while most of the PABP2 interacting proteins localise to granules, similar to the majority of mRNAs [44].
The data above confirm the strong association of PABP1 with eIF4G3 and eIF4E4. PABP2 in contrast interacts with eIF4E4, eIF4G3, eIF4G1 and two further proteins of the eIF4G1/eIF4E4 complex indicating multiple binding abilities to different paralogs of the eIF4F complex. For a more comprehensive picture, we analysed the enrichment ratios for all members of the translation initiation complex of the four individual experiments (Fig 4). PABP1 shows strong interactions with eIF4G3, eIF4E4 in all four experiments and, in particular in low salt conditions, also interaction with eIF4A1. Interactions with other translation initiation factors and with the two proteins known to interact with eIF4G1 and eIF4E5 [41] are absent or have very small enrichment factors. In contrast, PABP2 co-precipitated all five isoforms of eIF4G under low salt conditions and eIF4G1 and eIF4G3 also under high salt conditions. Similarly, all eIF4E subunits co-precipitated with PABP2 at least under low salt conditions, with the exception of eIF4E2 that has very low abundance [33]. Interestingly, both PABPs clearly co-precipitate each other, indicating that there may be complexes containing both PABPs on the same mRNA protein complex. For Leishmania PABPs such an interdependency has not been observed [36].
Our data contribute towards better understanding of translation initiation control mechanisms in trypanosomes. Demonstration of highly distinct interactomes for the two paralogs of PABP in African trypanosomes indicates discrete functions.
Specifically, PABP1 has a small interactome, comprising eIF4E4 and eIF4G3, and the hypothetical RNA-binding proteins RBP23 and Tb927.7.7460. PABP1, eIF4E4, eIF4G3 and RBP23 are largely absent from stress granules; the localisation of Tb927.7.7460 remains unknown. In contrast, PABP2 has a rather more extensive interactome that includes all proteins precipitated with PABP1, except RBP23, and most subunits of the eIF4F complex. PABP2 and the majority of its interaction partners localise to starvation stress granules.
The impairment in stress granule localisation of the entire PABP1 complex challenges the previous assumption that this complex is the major translation initiation complex involved in bulk mRNA translation. At starvation, polysomes largely dissociate and most mRNAs and proteins involved in mRNA metabolism localise to starvation stress granules [44]. Localisation to granules is the default pathway, and impairment in stress granule localisation is the exception. As the PABP1/eIF4E4/eIF4G3 complex does not locate to stress granules, it is unlikely to regulate translation of bulk mRNAs. Instead, we propose that the PABP1/eIF4E4/eIF4G3 complex is specialised for the regulation of a small subgroup of mRNAs. It is tempting to speculate that this group of mRNAs could be those encoding ribosomal proteins, because these mRNAs are the only group of mRNAs that were found to be excluded from starvation stress granules [44] and these are of small size, consistent with a localisation of PABP1 to lower molecular weight polysomes [49]. The interaction of PABP2 with most eIF4F subunits and many mRNA metabolism proteins indicates a wider substrate specificity for this PABP subunit. The data are consistent with PABP2 being distributed over a range of different translation initiation complexes and mRNAs and thus being responsible for bulk mRNA translation. The eIF4F complex that was identified with highest confidence to bind to PABP2 is eIF4G1/eIF4E5 with its previously identified interactors G1-IP and G1-IP-2. The fact that both PABPs co-precipitate each other indicates that a separation of the two PABPs to a distinct group of mRNA targets is potentially not strict. A model of the PABP target mRNAs, consistent with the data, is shown in Fig 5. One limitation of this study is that only one life cycle stage, the procyclic stage, was examined and we can not exclude that the PABP interactomes and their localisations are different in other life cycle stages, for example in blood stream forms.
Notably, the functions of many eIF4F complex subunits (for example eIF4E1, 2, 6), and their association with the PABPs remains unsolved. The reason could be that all studies to date, including this one, focus only on the proliferating life cycle stages of the parasites. Translational control may, however, be particular important in G1-arrested stages or during differentiation processes and an analysis of these stages, albeit experimentally challenging, may be highly informative for a more comprehensive picture of the eIF4F/PABP complexes.
T. brucei procyclic Lister 427 cells were cultured in SDM79 medium (containing fetal bovine serum from Sigma). The generation of transgenic trypanosomes was done using standard methods [65]. For starvation, parasites were washed once in one volume PBS and stored in PBS for two hours; the starvation time started at the first contact with PBS.
Cell lines expressing PABP1-eYFP or PABP2-eYFP from endogenous loci were previously described [49]. Proteins were expressed as C-terminal (RBP23) or N-terminal (all others) eYFP fusion proteins by transfecting trypanosomes with PCR products obtained with the template plasmid pPOTv7-blast-blast-eYFP (RBP23) with oligonucleotides designed as described [66]. All transfected cell-lines co-expressed PABP2-mChFP from the endogenous locus [49] as a marker for starvation stress granules. The plasmid for the expression of a C-terminal 4Ty1 fusion protein was previously described for PABP1 [49]) and made accordingly for PABP2 [67].
Cells were washed with serum-free SDM79, fixed with 2.4% paraformaldehyde overnight, washed once in PBS and stained with 4′,6-diamidino-2-phenylindole (DAPI). Z-stacks (100 images, 100-nm spacing) were recorded with a custom-built TILL Photonics iMIC microscope equipped with a 100×, 1.4 numerical aperture objective (Olympus, Tokyo, Japan) and a sensicam qe CCD camera (PCO, Kehlheim, Germany) using exposure times of 500 ms for fluorescent proteins and 50 ms for DAPI. Images were deconvolved using Huygens Essential software (SVI, Hilversum, The Netherlands) and are presented as Z-projections (method sum slices) produced by ImageJ [68].
Procyclic trypanosomes were grown to a density of 5–8 x 106 cells/ml. Four litre cultures were harvested in a F14S-6x 250 Y rotor at 1500g at room temperature in four subsequent centrifugations and washed once with 250 ml serum free SDM-79. Finally, the cells were sedimented by centrifugation (1500*g) into a capped 20 ml syringe placed in a 50 ml Falcon tube. After discarding all supernatant, inserting the plunger and removing the cap the cells were passed slowly into liquid nitrogen in order to form small pellets suitable for subsequent cryomilling. Frozen cells were processed by cryomilling into a fine powder in a planetary ball mill (Retsch) [55]. For precipitation, aliquots of approximately 50 mg powder (corresponding to ~2 x 108 cells) were mixed with 1 ml ice-cold buffer (low salt buffer: 20 mM HEPES pH 7.4, 50 mM NaCl, 1 mM MgCl2, 100 μM CaCl2, 0.1% CHAPS; high salt buffer: 20 mM HEPES pH 7.4, 50 mM NaCl, 1 mM MgCl2, 100 μM CaCl2, 150 mM KCl, 0.1% CHAPS) complemented with protease inhibitors (Complete Protease Inhibitor Cocktail Tablet, EDTA-free, Roche). After sonication with a microtip sonicator (Misonix Utrasonic Processor XL) at setting 4 (~20 W output) for 2 x 1 second, insoluble material was removed by centrifugation (20,000 g, 10 min, 4°C). The clear lysate was incubated with 3 μl polyclonal anti-GFP llama antibodies covalently coupled to surface-activated Epoxy magnetic beads (Dynabeads M270 Epoxy, ThermoFisher) for two hours on a rotator. Beads were washed three times in the respective buffer (low salt or high salt buffer) and finally incubated in 15 μl 4 x NuPAGE LDS sample buffer (ThermoFisher), supplemented with 2 mM dithiothreitol, at 72°C for 15 minutes to elute the proteins. The precipitates were analysed on an SDS-PAGE gel stained with Coomassie. For subsequent proteomics analysis six pullout samples were pooled after the final washing step and eluted in 30 μl 4 x NuPAGE LDS Sample buffer, then run 1.5 cm into a NuPAGE Bis-Tris 4–12% gradient polyacrylamide gel (ThermoFisher) under reducing conditions. The respective gel region was sliced out and subjected to tryptic digest and reductive alkylation.
For the precipitation of eIF4E4 and G1-IP2, essentially the same protocol was used starting from 2 L cultures at a density 8 x 106 cells/ml. The immunoprecipitation was carried out in low salt buffer using 5 ul recombinant, monoclonal dimeric fusion anti-GFP nanobody LaG16-LaG2 [69] coupled to magnetic beads. The same beads, where the antibody coupling step was omitted were used as a control. Eluates were run on a NuPAGE Bis-Tris 4–12% gradient polyacrylamide gel (ThermoFisher) under reducing conditions, then subjected to western blotting using standard procedures. 4Ty1 tagged fusion proteins were decorated with monoclonal anti-Ty1 antibody clone BB2 (Sigma) at 1:10,000 dilution. Quantitation was performed on raw images gathered under nonsaturating conditions using ImageJ [68] and enrichment ratios calculated comparing against uncoupled control beads.
Liquid chromatography tandem mass spectrometry (LC-MS2) was performed on a Dionex UltiMate 3000 RSLCnano System (Thermo Scientific, Waltham, MA, USA) coupled to an Orbitrap VelosPro mass spectrometer (Thermo Scientific) at the University of Dundee FingerPrints Proteomics facility and mass spectra analysed using MaxQuant version 1.5 [56] searching the T. brucei brucei 927 annotated protein database (release 8.1) from TriTrypDB [70]. Minimum peptide length was set at six amino acids, isoleucine and leucine were considered indistinguishable and false discovery rates (FDR) of 0.01 were calculated at the levels of peptides, proteins and modification sites based on the number of hits against the reversed sequence database. Ratios were calculated from label-free quantification intensities using only peptides that could be uniquely mapped to a given protein. If the identified peptide sequence set of one protein contained the peptide set of another protein, these two proteins were assigned to the same protein group. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [71]ƒ partner repository with the dataset identifier PXD008839.
|
10.1371/journal.ppat.1003561 | Replication Vesicles are Load- and Choke-Points in the Hepatitis C Virus Lifecycle | Hepatitis C virus (HCV) infection develops into chronicity in 80% of all patients, characterized by persistent low-level replication. To understand how the virus establishes its tightly controlled intracellular RNA replication cycle, we developed the first detailed mathematical model of the initial dynamic phase of the intracellular HCV RNA replication. We therefore quantitatively measured viral RNA and protein translation upon synchronous delivery of viral genomes to host cells, and thoroughly validated the model using additional, independent experiments. Model analysis was used to predict the efficacy of different classes of inhibitors and identified sensitive substeps of replication that could be targeted by current and future therapeutics. A protective replication compartment proved to be essential for sustained RNA replication, balancing translation versus replication and thus effectively limiting RNA amplification. The model predicts that host factors involved in the formation of this compartment determine cellular permissiveness to HCV replication. In gene expression profiling, we identified several key processes potentially determining cellular HCV replication efficiency.
| Hepatitis C is a severe disease and a prime cause for liver transplantation. Up to 3% of the world's population are chronically infected with its causative agent, the Hepatitis C virus (HCV). This capacity to establish long (decades) lasting persistent infection sets HCV apart from other plus-strand RNA viruses typically causing acute, self-limiting infections. A prerequisite for its capacity to persist is HCV's complex and tightly regulated intracellular replication strategy. In this study, we therefore wanted to develop a comprehensive understanding of the molecular processes governing HCV RNA replication in order to pinpoint the most vulnerable substeps in the viral life cycle. For that purpose, we used a combination of biological experiments and mathematical modeling. Using the model to study HCV's replication strategy, we recognized diverse but crucial roles for the membraneous replication compartment of HCV in regulating RNA amplification. We further predict the existence of an essential limiting host factor (or function) required for establishing active RNA replication and thereby determining cellular permissiveness for HCV. Our model also proved valuable to understand and predict the effects of pharmacological inhibitors of HCV and might be a solid basis for the development of similar models for other plus-strand RNA viruses.
| Hepatitis C virus (HCV) infection is a major global health problem, with approximately 170 million chronically infected individuals worldwide and 3 to 4 million new infections occurring each year [1]. Acute infection is mostly asymptomatic, however, it develops into a chronic infection in about 80% of patients, and then is a leading cause of liver cirrhosis, hepatocellular carcinoma and subsequent liver transplantation [2], [3], [4]. A significant fraction of patients cannot be cured even with modern combination therapies, partially due to ab initio non-responsiveness, but also due to the emergence of drug-resistant HCV quasispecies.
HCV is an enveloped plus-strand RNA virus and belongs to the Flaviviridae family. Upon entry into the host cell, its 9.6 kb genome is translated by a cap-independent, internal ribosomal entry site (IRES) mediated mechanism into a single large polyprotein. Viral and cellular proteases co- and post-translationally cleave this precursor into ten mature viral proteins, comprising three structural proteins (core, E1, E2), the ion channel p7 as well as the six non-structural (NS) proteins NS2, 3, 4A, 4B, 5A and 5B [5]. The five “replicase” proteins NS3 to NS5B are essential and sufficient for intracellular genome replication. NS3 comprises an RNA helicase and a protease domain, the latter of which, together with the co-factor NS4A, forms the major viral protease NS3/4A, liberating itself and all other replicase proteins from the polyprotein precursor. NS4B, together with other NS proteins, induces membrane alterations, observable as convoluted, vesicular membrane structures known as the membranous web and believed to act as the sites of RNA replication [6], [7]. The exact architecture and topology of these structures, and particularly their structure-function-relationship, is not fully understood yet. However, for Dengue virus (DV), a related flavivirus, the three-dimensional makeup of the membrane rearrangements has been solved recently [8]. There, numerous small, vesicular invaginations into the rough endoplasmic reticulum (ER) serve as a protected environment for genome replication. NS5A is a phosphoprotein important both in RNA replication and particle assembly and/or release. NS5B, the RNA-dependent RNA polymerase (RdRP), is the core enzyme of the replicase complex. In order to amplify the viral RNA, NS5B first synthesizes a complementary (i.e. negatively oriented) strand from the plus stranded genome, putatively resulting in a double-stranded (ds) intermediate [9]. From this negative strand template, NS5B then transcribes progeny plus strands. Given the ∼10-fold higher number of plus strands over minus strands within the host cell, this most likely occurs in a repetitive manner [10]. Newly synthesized plus strands are then released by an unknown mechanism from the replicative compartment and can then either be directed to encapsidation into assembling virions, or re-enter the replicative cycle by serving as templates for further translation and subsequent incorporation into a new replication complex.
It is interesting to note that although HCV establishes a persistent infection, it does not have a latent phase; throughout the course of the infection, which can be decades long in many patients, there is constant production of viral RNA, proteins and infectious particles. In most viral infections, presence of non-self structures, such as dsRNA or viral proteins, is readily detected by sensors of the immune system, leading to the production of type I interferon (IFN) and activation of the adaptive immune response [11]. Also in case of HCV, innate as well as adaptive immune responses are elicited, however, by means of various complex interactions with cellular processes, the virus is capable to blunt these defense mechanisms and thus is able to persist [12]. This ability of HCV to maintain low profile persistence is most likely intimately linked to its tightly controlled RNA replication; for the closely related bovine diarrhea virus (BVDV), which can be converted from a persistently to an acutely replicating form, a direct correlation between excessive RNA replication and the induction of cytopathic effects has been described [13]. To comprehensively study these complex and highly dynamic processes that can only inappropriately be addressed by individual experiments, an eminent approach is mathematical modeling. Consequently, a basic model of HCV infection dynamics was published almost 15 years ago [14] and has since led to the development of several related models of HCV infection and therapy dynamics [15], [16], [17], [18], [19], [20], [21]. However, all of these models described the long-term dynamics at the level of cell populations, organs and even organisms (patients), and did not take intracellular processes such as genome translation and the actual RNA replication into account. With the development of subgenomic HCV replicons, detailed studies of intracellular RNA replication became possible [22], [23]. A thorough quantitative analysis of persistent subgenomic replicons in Huh-7 cells [10] led to the development of a first mathematical model of intracellular steady state RNA replication [24]. Further models addressed the effect of potential drugs on viral replication [25] or included the production of virus particles [26], [27]. However, all published models were solely based on measurements of steady state replication. In contrast, to understand how the virus on the one hand manages to efficiently (and quickly) establish itself within a host cell before the cell is able to mount an antiviral response, while on the other hand, it is strictly limiting its own amplification, static (steady state) data is not sufficient but needs to be complemented by information about the dynamic aspects of replication. Previous studies on replication kinetics in cell culture in fact point to a highly dynamic initial phase of RNA replication in the first few hours after genome transfection or infection, which then reaches a steady state within 24–72 hours [22], [28], [29]. Actual amplification kinetics and the absolute levels attained in the steady state vary largely between different experimental systems and are mainly determined by the permissiveness of the employed host cell [30], [31], [32] and by the viral isolate [30], [32], [33].
Therefore, in our present study we quantitatively followed the onset of intracellular RNA replication within the first couple of hours upon introduction of HCV genomes into the host cells. Based on these data we developed a comprehensive mathematical model capable of precisely describing both, the dynamic and the steady state phases of viral replication. We then used this model to study aspects of the viral replication cycle that cannot directly be accessed experimentally.
To assess the dynamics of HCV RNA replication, we performed quantitative, time resolved measurements of strand specific viral RNA and polyprotein concentrations over 72 h after viral RNA transfection into Huh7 cells. To achieve sufficiently strong replication that can be measured reliably, we used subgenomic reporter replicons carrying the firefly luciferase gene in front of the viral proteins [28] (figure 1A), and we synchronized the onset of replication to the largest feasible extent by using electroporation to instantaneously introduce in vitro transcribed replicon RNA into the cells. As host cellular factors play a critical role in determining the efficiency of viral replication [30], [31], we used two different cell lines: Huh7-Lunet is a clonal cell line of exceptionally high permissiveness for HCV RNA replication [34], whereas a low passage of standard Huh-7 cells (Huh-7 lp) replicates HCV RNA to significantly lower levels, as has been described previously [30]. Over the course of 72 hours we then followed HCV replication, measuring plus-strand and minus-strand RNA by strand specific quantitative Northern blotting and firefly luciferase activity as a highly sensitive surrogate marker of viral protein translation, since luciferase expression was under the control of the HCV IRES (see figure 1A). Of note, luciferase activity correlates with the amount of viral protein translated, but does not allow discrimination between cytoplasmic NS proteins and proteins inside the RC.
Upon transfection of replicon RNA into Huh7-Lunet cells, the RNA was instantly translated into protein and at the same time was rapidly degraded (figure 1B). Consequently, after a first peak, translation also leveled off or even decreased slightly, while negative strand RNA first became detectable at 4–8 hours post transfection. From around 8 hours on, synthesis of new positive strand RNA then exceeded its degradation, and levels of both, positive and negative strand RNA as well as of viral protein started to increase rapidly (note the logarithmic scale in figure 1B). A steady state was finally reached at around 30 hours post transfection (in Huh7-Lunet), which was stable until the end of the observation at 72 hours.
In order to comprehensively understand the observed HCV replication dynamics and its underlying molecular processes, we set up a mathematical model of the intracellular HCV RNA replication. Dahari and colleagues developed a similar model previously, upon which we could build here [24]. Briefly, our model comprises all relevant molecular species (RNA, proteins, ribosomes, etc.), and describes each step in the RNA replication cycle, such as translation, protein maturation and the formation of the actual RNA replication complex, as reactions of the involved molecules using differential equations based on standard mass action kinetics. Of note, the establishment of a vesicular replication compartment (RC) by viral proteins (in concert with cellular functions) within which RNA replication takes place is reflected in the model by partitioning of the respective molecular species into distinct “cytoplasmic” and “replication compartment” pools; e.g. only cytoplasmic HCV RNA (RPcyt) can be translated by ribosomes, but not HCV RNA within the replication compartment (RP). Model equations (eq.) of our final model are given in the materials and methods section and a schematic illustration is shown in figure 2C. The original model of Dahari was solely based on steady state measurements of viral RNA and protein concentrations in a stable replicon cell line [10], and accordingly was not capable of explaining the dynamic phase during the establishing of replication as observed in our experimental data, even after re-fitting all model parameters (high permissive cell line; total sum of squared residuals χ2 = 8.69, compare supplementary figure S1). From this finding it was evident, that modifications to the model were required in order to accurately capture the initial dynamics of HCV RNA replication, as it can be observed upon transfection of viral genomes into “naïve” cells.
Based on biological reasoning, we extended and modified Dahari's original model at two steps of the replication cycle. For one, to account for ab initio replication in our setting (in contrast to pre-formed steady-state replication), we introduced one additional RNA species Rpunp, representing the transfected “naked” replicon RNA and an according processing step (rate k0), subsuming any re-folding, association with RNA-binding proteins and other processes that might take place and be required before in vitro transcribed RNA assumes a translation-competent state (eq. 1 and 2). This processing corresponds to viral genomic RNA being released into the cytoplasm upon actual infection. We furthermore allowed RNA degradation to be different (presumably higher) for the “unprocessed” transfected RNA (μpUnp) as compared to “processed” or cell-derived RNA (μpcyt). The second step that we updated to reflect the current biological understanding of the molecular processes was the initiation of minus strand RNA synthesis (which in the model is assumed to correspond to the formation of the replicative compartment, as discussed later). It has been described for HCV, but also for other viruses [35], [36], [37], [38], that the formation of a productive replicase complex requires the viral polymerase (NS5B) and possibly other NS-proteins to be supplied in cis. This means that for reasons not yet fully elucidated, NS5B cannot initiate RNA synthesis from a free, cytosolic RNA genome, but only from the very RNA that it has been translated from. This implies a tight spatio-temporal coupling of (poly)protein production and initiation of RNA replication, i.e. initiation can only occur immediately after translation/polyprotein processing and therefore in close proximity to the translation complex (TC). As our model does not account for spatial effects (such as diffusion), we approximated this cis-process by requiring an active translation complex instead of free, non-translating RNA (RPcyt) for the initiation of minus strand RNA synthesis (RIP, eq. 7). This cis-triggered formation of the replicative compartment consequently is the only route for uptake of viral genomes and also NS proteins (Ecyt) into replication vesicles. This is a major change to Dahari's original model, in which cytosolic RNA (Rpcyt) and NS proteins (Ecyt) could freely and independently enter the compartment. This model, comprising Dahari's original model with the described extensions, we then considered our base model.
We then tested, whether our base model would be capable of explaining the measured replication dynamics. We therefore fitted the model to the experimental data from the high permissive Huh7-Lunet cells. In fact, this resulted in a significantly better fit as compared to the original model (Dahari: χ2 = 8.69, base model: χ2 = 2.12) and was capable of adequately describing both, the highly dynamic initial phase as well as the ensuing steady state of viral RNA replication (figure 2A).
Having established a base model for HCV replication, we next wanted to assess which factors could explain differences observed between high and low permissive cell lines. In our experimental measurements for two differently permissive cell lines, Huh7-Lunet (high permissive) and Huh-7 lp (low permissive), replication reached a steady-state within the period of observation (72 h), however, plateau levels of viral protein, plus-strand RNA and minus-strand RNA differed by approximately one order of magnitude; further, the onset of the net increase of plus-strand RNA was delayed significantly in the low permissive cells and also the minimum concentration of plus-strand RNA reached during net degradation in the first hours after transfection were significantly lower in low permissive cells (compare figure 1B and C). As both, Huh7-Lunet and Huh-7 lp cells, were transfected with the same subgenomic HCV replicon, these differences must be due to differences between the host cells. In order to reflect this host influence also in our model, we tested different steps in the HCV RNA replication cycle that do or could feasibly depend on a host process: (A) efficiency of RNA entry or initial RNA processing; (B) the number of ribosomes available for RNA translation; (C) RNA degradation in the cytoplasm (possibly including antiviral processes such as activation of RNaseL); (D) polyprotein translation or maturation (i.e. cleavage); (E) the formation of the replicative compartment/initiation of minus-strand synthesis; (F) RNA synthesis or (G) RNA degradation inside the replication compartment; or (H) the export of newly synthesized RNA into the cytoplasm. To evaluate these alternatives for their capacity to explain the differences in dynamics and steady-state levels between the two cell lines, we fitted our base model simultaneously to the experimental data from both cell lines, leaving only the parameters free to differ between high and low permissive cells that, in the respective hypothesis (A) to (H), depend on the corresponding host factor; all other parameters were constrained to be identical between the two cell lines. We found that hypotheses (A), (B), (C), (D), (F), (G) and (H) could not explain the above described qualitative difference in replication dynamics: while (C) and (H) did lead to a steady-state but could not reproduce the lower plateau RNA levels in Huh-7 lp, hypotheses (A), (B), (F) and (G) altogether failed to establish a steady-state in low permissive cells in the course of the simulated time period of 80 h (supplementary figure S2). In order to identify the best fitting hypothesis, we also quantitatively assessed the capability of each hypothesis to fit both data sets by calculating χ2 over all data points from the two time series, as well as Akaike's information criterion (AIC), which additionally takes into account the number of unconstrained parameters (figure 2B). While parameter differences in the RNA synthesis inside the RC, i.e. hypothesis (F), led to the lowest overall χ2 value, hypothesis (E)– assuming a difference in the formation of the RC and initiation of RNA synthesis– led to a slightly larger χ2 (5.84 vs. 5.60) but a significantly lower AIC (−21.31 vs. −0.68). Moreover, hypothesis (E) reached a steady-state within 80 h, while (F) did not. This comparison therefore identified the initiation of minus strand RNA synthesis (i.e. the formation of the RC) as the step in the model, at which alteration of a single reaction rate suffices to optimally transform replication dynamics from high permissive cells into the dynamics found in low permissive cells.
Biologically, this step is highly complex and not thoroughly understood yet. After translation and polyprotein processing, reorganization of host cell endomembranes is triggered by viral NS proteins such as NS4B, which has been shown to be a key player in the formation of membrane convolutions at the rough endoplasmic reticulum. These vesicular membrane structures, dubbed the membranous web, have been reported to be the site of HCV RNA replication [7], providing a distinct replicative compartment for the viral replicase machinery. However, the molecular mechanisms leading to the formation of productive replication vesicles are not known. Nonetheless, it is clear that host factors must be required in this complex process, for example proteins involved in membrane biogenesis and reorganization, as well as signal transducers and regulatory molecules; and also the initiation of minus strand RNA synthesis might require a cellular co-factor. It appears plausible that limited abundance of one of these factors in some cells might be responsible for their lower permissiveness for HCV replication. Therefore, we next wanted to include this host factor as an explicit species in our model, which is required for RC formation/minus strand initiation. For that purpose, we subsumed all these possible host determinants by one unspecified host factor HF (see figure 2C), which we assumed to interact with viral NS proteins (Ecyt, e.g. NS4B or NS5A) and with actively translated HCV RNA (TC) to create replication vesicles and to allow for initiation of minus-strand RNA synthesis (being part of the minus-strand initiation complex RIP, see eq. 7 and figure 2C). In addition, we made the (non-crucial, see supplementary figure S3 and supplementary table S5) assumption that HF is only catalyzing the reaction without being consumed.
With this additional modification to the mathematical description of the formation of replication compartments, and calibration of the model to the experimental data from both cell lines (constraining parameters and initial values to biologically meaningful ranges taken from measurements or literature wherever possible), excellent agreement between the model and experimental data was achieved, both, for high and low permissive cells with an overall χ2 of 2.01 and AIC of −112.31(figure 3A and B). We therefore considered this our final working model, illustrated in detail in figure 2C. Briefly, the model comprises 13 molecular species in two distinct compartments, the cytoplasm and a replicative compartment (RC), and is parameterized with 16 parameters, corresponding to reaction rates, as well as three non-zero initial values: the initial concentration of HCV RNA (Rpunp), the initial concentration of the host factor (HF), as well as the total number of ribosomes available for viral RNA translation (Ribotot). The full system of differential equations and detail on the modeling procedure can be found in the Materials & Methods section; more detail on parameter optimization and analysis are given in supplementary text S1.
Interestingly, analysis of the fitted parameters showed that the concentration of the host factor was more than 10 fold higher in highly permissive Huh7-Lunet cells than in low permissive Huh-7 lp cells. This difference led to slower formation of the replication compartment in Huh-7 lp cells, which in turn resulted in the observed delay in early viral replication and in decreased steady state levels in these cells. Based on our model and computational analysis, we therefore propose that a host process is critically involved in the formation of replication vesicles and/or the initiation of minus-strand RNA synthesis, turning this into the rate-limiting step for HCV RNA replication in low permissive cells.
While the model could be very well fitted to the original replication data, we then wanted to corroborate its applicability for predicting replication dynamics also under distinct conditions that were not part of the calibration process. For this purpose, we performed additional, independent experiments using mutant HCV replicons with defects at defined stages of the replication cycle. We predicted the impact of such defects on viral replication a priori using the model, and retrospectively compared the results with the experimental data in order to assess the validity of model predictions. This approach of introducing targeted mutations into the HCV genome interfering with distinct functions in the viral RNA replication cycle allows validation of individual steps in the model, thus step-wise reconfirming model assumptions and parameters.
As a test of the translation phase of the model, we measured viral plus-strand RNA and protein expression using a replication deficient replicon harboring a deletion of the catalytic triad (GDD motif) of the NS5B polymerase. The measured RNA and protein data thus reflect only the effect of translation and degradation in the cytoplasm, with no RNA replication occurring. We predicted the impact of this intervention with our model by setting the formation rate kPin of the plus strand replication initiation complex RIp to zero (eq. 3, 5 and 7), thus completely switching off polymerase activity at the earliest possible point, while leaving all other model parameters unchanged. Notably, our model predictions of this intervention matched the experimental data from both, Huh7-Lunet and Huh-7 lp cells, validating our model of cytoplasmic RNA degradation and translation (Figures 4A and B). The fact that the experimental measurements showed almost identical RNA decay dynamics and viral protein (luciferase) levels in high and low permissive cells is also direct experimental confirmation of our modeling based assessment above, that differences in permissiveness cannot be related to RNA “processing” or degradation, or to ribosome availability or protein translation in the cytoplasm (hypotheses (A), (B), (C) and (D) tested above).
We next focused on validating the RNA replication steps of our model. For this purpose, we utilized chimeric replicons with heterologous 5′- or 3′-NTRs derived from a different genotype [22]. We previously showed that these chimeric replicons exhibit decreased replication efficiency due to inefficient initiation of plus-strand synthesis (in case of the 5′-NTR exchange) or minus-strand synthesis (in case of the 3′-NTR exchange) [22]. We predicted the effect of these interventions with the fitted model by decreasing the parameters kpin and k4m for the 3′-NTR exchange (eq. 3, 5, 7, 8 and 9), and k5 and k4p for the 5′-NTR exchange (eq. 7, 8 and 9), corresponding to the rates of the minus- and plus-strand initiation and synthesis, respectively (for reference, see figure 2C). Comparison of our prediction with experimental measurements demonstrated that in both cases the model qualitatively agreed with the experimental data. Consequently, upon refitting of these parameters to the new data, the model was capable of quantitatively describing the perturbed replication kinetics (figure 4C). Furthermore, the model correctly predicted the impact of the respective NTR-exchanges onto the ratio of plus- to minus-strand RNA at the steady state (figure 4D). Predictions for both NTR-exchanges were in close quantitative agreement with our previously published experimental observations, which showed an 8.7∶1 (simulation 9.0∶1) ratio between plus- and minus-strand for the wildtype, 16.1∶1 (11∶1) for the 3′-NTR-chimera, and 4.7∶1 (4.8∶1) for the 5′- chimera [22].
Taken together, our model was able to correctly infer the effects of targeted interventions at different steps of the replication process, including complete replication deficiency, as well as specific inhibition of plus- or minus-strand RNA synthesis, respectively. We therefore conclude that our model provides a realistic description of HCV RNA replication dynamics, and thus can be confidently used to further study such processes in silico that are difficult or impossible to address experimentally.
Having such a comprehensive and accurate model at hand, we proceeded by applying it to concrete problems in the field of HCV research. The first question we addressed was which sub-steps of HCV RNA replication would be most susceptible to targeted interference. Such processes are potentially attractive targets for the design of new DAAs against HCV. To find out which step in the replication cycle has the biggest impact on the resulting RNA and protein levels, we assessed the relative sensitivity of replication towards alterations of reaction rates in the model. To account for the two clearly discernable phases of replication – the highly dynamic establishing phase and the steady-state phase – we performed a global sensitivity analysis using the extended Fourier Amplitude Sensitivity Test (eFAST) [39], [40] at an early (4 h) and at a late (72 h) time point. We separately assessed the sensitivities of plus-strand RNA, minus-strand RNA as well as protein levels towards individual and simultaneous changes of 16 rate constants and the three initial values (figure 5 and supplementary figure S4).
For the establishing phase of replication, this analysis showed that the most influential processes are the polyprotein translation (rate k2), the export rate of RNA into the cytoplasm (rate kPout) and the efficiency of plus- (rate k4p) and minus- (rate k4m) strand RNA synthesis inside the replication compartment, respectively (figure 5A). As one would expect, alterations in k2 mainly influence the amount of viral protein (eq. 4 and 6) and only to a lesser degree viral RNA (eq. 2 and 3), whereas k4m mainly affect RNA species (eq. 7, 8 and 9). k4p and kPout in contrast strongly influence RNA and protein concentrations (eq. 8, 9, 10 and 11). Further important steps are the initial “processing” of the transfected RNA (rate k0), since this determines at what time and to what extent RNA is available for translation, as well as the RNA degradation rate μRC inside the replication compartment. The availability of viral RNA for rapid genome replication and the replication process inside the membranous web itself are therefore key determinants of the initial replication dynamics and thus the efficiency of infection, and consequently constitute a very attractive target for anti-viral drugs. Interestingly, the rate of polyprotein translation (eq. 4) naturally has a big impact on viral protein concentration, but only a fairly restricted influence on RNA levels. Furthermore, the cleavage rate of nascent viral polyprotein (eq. 4 and 5, rate kc) only very mildly impacts replication dynamics.
A profoundly different pattern can be observed for the steady state phase. The single most influential parameter determining viral RNA and protein levels was found to be the degradation rate of viral RNA inside the replication vesicles μRC (eq. 7 to 11), while most other parameters showed no significant sensitivities (figure 5B and supplementary figure S4). However, it is virtually impossible to influence this parameter by cellular (e.g. innate immune) or pharmacological intervention (except by physical destruction of the membranous structures), therefore making inhibition of viral replication particularly cumbersome once the steady state has been established. Taken together with the results from the early phase, these analyses suggest a key role of the replicative compartment for a successful establishment and maintenance of infection.
In the light of the above findings, pointing to a central role of the membranous web within the RNA replication cycle, we further studied the underlying molecular functions of this compartment. For one, we assessed the importance of its protective character onto the dynamics of viral genome replication. Model fitting led to a more than 4-fold lower RNA degradation rate within the replication compartment (μRC) as compared to RNA degradation in the cytoplasm (μpcyt, see table 1). To simulate the effect of less stringent protection of the RNA inside the RC, we then deliberately increased its degradation rate (μRC) and calculated the resulting levels of plus strand RNA over time (figure 6A). This analysis showed that the degradation rate inside the replicative compartment inversely correlated with the amount of RNA produced at any given time. Interestingly, this correlation was not continuous, exhibiting a threshold of productive RNA replication, constituting a “cliff”, crossing of which prevented the establishing of a (non-zero) steady-state and effectively killing off viral replication (figure 6A, dark blue area, see also supplementary figure S5). This highly instable region with very low (or zero) RNA copy numbers, strikingly, was reached once degradation inside the RC (μRC) was approximately equal to the degradation rate in the cytoplasm (μpcyt). Our model therefore predicts that the viral RNA must be protected from active degradation during replication in order for HCV to maintain robust persistent replication. While it is virtually impossible to reproduce the above findings in a biological experiment (i.e. increasing RNA degradation inside the replicative compartment), previous in vitro data actually showed that viral RNA in the cell, particularly the minus-strand, is highly resistant to nuclease treatment [10], implying that indeed degrading enzymes cannot enter the replication vesicles. Moreover, in inhibitor studies, ongoing HCV replication was blocked by interferon or a pharmacologic NS3/4A inhibitor, leading to rather slow decrease of RNA with a half-life of 12–20 h [41], [42], most likely representing a slow degradation of replication vesicles. In good agreement with these studies, our model predicts a half-life for RNA inside the replicative compartment of 12 h (rate μRC = 0.08 h−1), whereas RNA transfected into the cytoplasm decayed with a half-life of approximately two hours in the experiments using a replication-defective replicon (see figure 4A). Experimentally very hard to address, however, is the degradation rate μpcyt of cytoplasmic HCV RNA generated through replication that might exhibit a different folding or be bound by other proteins as compared to transfected RNA. Yet, it appears highly likely that this degradation rate would more closely match the rate of degradation of transfected, cytoplasmic RNA rather than that of RNA within the membranous replicative environment. In keeping with this plausible assumption, our model predicts a half-life for newly synthesized cytoplasmic RNA of approximately 165 min (μpcyt = 0.363 h−1). Although model estimations for both, μpcyt and μRC, exhibit a rather broad confidence interval, simultaneous modification of both parameters shows that μRC needs to be substantially lower than μpcyt in order to explain the observed kinetics (figure 6B, dark blue area). In terms of viral protein, Quinkert and colleagues showed that in contrast to RNA, only a small fraction (<5%) of NS5B molecules is protease resistant [10]. In line with these observations, our model predicts that the vast majority of viral protein remains in the cytoplasm.
Another important question, which can hardly be addressed experimentally, is the possibility of re-initiation of minus-strand synthesis inside the replication vesicle. While theoretically it is feasible that the replicative machinery re-initiates minus-strand synthesis on newly produced plus-strands inside the replication compartment (eq. 7, second to last term), the alternative hypothesis is that such an initiation event can only happen in cis upon translation in the cytoplasm (see also section on model development above). In fact, when analyzing the calibrated model, we found that the rate constant for this reaction (k3 in eq. 7, see figure 2C for reference) needed to be close to zero (<10−4 h−1*molecules−1) to fit the experimental data, and the concentration of “active” polymerase (E) was severely limiting the rate of RNA synthesis during the initial dynamic phase. This resulted in an extremely low efficiency of internal re-initiation, implying that most or all of the newly synthesized viral plus-strand RNA is exported to the cytoplasm, from where it must be re-imported for further rounds of RNA replication to occur. Hence, our model supports the notion that negative-strand initiation is very different from plus-strand initiation in that it most likely depends on actively translated RNA with the required NS proteins, mainly NS5B, being supplied in cis.
The observed relative shortage of active polymerase within the replication vesicles and the lack of internal re-initiation consequently prevents an exponential amplification of the viral RNA within the replicative compartment. Replication vesicles thus attenuate the rate of viral replication by limiting the availability of the factors required for minus-strand initiation. At the same time, depending on the export rate of newly synthesized plus-strand RNA from the replication vesicles (kpout), they can also exert tight control over protein translation. Newly synthesized RNA can either be exported to the cytoplasm where it can be used for another round of protein translation (or, in an actual infection setting, the assembly of new viral particles), or it accumulates within the replication vesicles; there, however, it cannot be used as a template for minus strand synthesis due to the above described reasons.
Taken together, the development of a membranous replication compartment, by physically separating production of new protein (translation) and the generation of new RNA (replication), therefore constitutes an important additional level of control over the virus' replication kinetics. This high degree of controllability might be one reason for the evolutionary success of membranous replicative structures, as basically formed by all positive strand RNA viruses. In case of HCV, it allows for sustained low-level replication as is required for the establishment of persistence, mainly by restricting availability of the required proteins within the replicative compartment.
Particularly for a persistent virus, tight control over its own replication is essential in order to not overwhelm its host cell and thereby kill it [13]. As we have learned above, the distinct replication compartment plays a central role in this self-limitation. Consequently, we therefore studied, which processes in turn regulate the formation of replication vesicles and eventually lead to the establishment of a steady state. The host factor (HF) in our model has been found to be a requisite for the attainment of a steady state and its concentration was a determinant regulating plateau levels of viral RNA and protein between the two differently permissive cell lines. For that reason, we now systematically assessed the impact of different availabilities of HF onto steady-state levels of viral RNA and protein. For HCV RNA levels, this analysis showed a linear correlation with HF concentrations in the range tested: the more abundant HF was, the more RNA replication took place. Interestingly, however, polyprotein levels exhibited a saturation behavior, reaching a plateau for HF concentrations above 20 “molecules” (note that HF is a virtual species, so “molecules” is an arbitrary unit) (figure 6C). To understand this nonlinear dependence of viral protein on HF levels, we analyzed the model under conditions of varying HF amounts and found that this saturation stems from different factors being limiting for increasing HF levels: in low permissive cells (featuring low HF concentrations of around 4 “molecules”), HF availability is limiting the formation of replication vesicles (eq. 7). Therefore, overall RNA concentrations remain relatively low, leaving polyprotein production at a low but steady level; here, RNA in the cytoplasm is the rate limiting factor for protein translation. In high permissive cells (featuring high HF levels of around 50 “molecules”), in contrast, rapid formation of replication vesicles occurs with an associated rapid increase in viral RNA levels. However, ribosome availability (Ribo) then becomes limiting for protein translation (eq. 3), explaining the plateau seen for viral protein concentrations (figure 6C). Accordingly, the ratio between viral protein (i.e. luciferase) and plus-strand RNA is lower in the steady state in high permissive cells. This is well in line with the experimental data (figure 1, compare B and C).
Interestingly, these findings suggested that the actual mechanisms governing the establishing of the steady state in low permissive and high permissive cells are different. While in low permissive cells the formation of replication vesicles is the limiting step due to a lack of host factor HF, surprisingly the host translation machinery is the bottleneck in high permissive cells.
As differential abundance of the host factor (or host process) HF in our model sufficed to explain the observed difference in HCV replication dynamics between high and low permissive cells, it was intriguing to identify the biological nature of this factor. For that purpose, we set out to compare gene expression profiles of Huh-7 cells of different passage number or clonal origin that we had found to exhibit substantially different permissiveness for HCV RNA replication [22], [30] (figure 7A). We performed full-genomic cDNA microarray (Affymetrix HGU133plus 2.0) analysis in eight such Huh-7 derived cell lines, including the above used Huh7-Lunet and low passage (lp) Huh-7 cells. Figure 7B shows a scatterplot of the normalized gene expression values for these two cell lines. Assuming a direct correlation between permissiveness and the expression of the host factor HF as suggested by the above analysis (compare figure 6B), we fitted a linear model of each gene's expression level to the observed replication efficiencies in all eight cell lines. By this, we could assess how well each individual gene predicts replication efficiency over the full set of cells. On these data, we then carried out an analysis of variance (ANOVA) to identify genes whose expression profiles correlated significantly with replication efficiency. Figure 7C shows the resulting p-values over the degree of differential expression (as log fold-change) between Huh7-Lunet and Huh-7 lp (see also supplementary table S1). We could identify 355 genes, whose expression levels correlated with permissiveness (p<0.2) and which exhibited a difference in expression levels of more than 23% (log fold-change >0.3 or <−0.3) (figure 7C and supplementary table S2). We then subjected these potential HF candidates to bioinformatics analyses in order to identify host cellular processes or pathways, which are over-represented among those genes (supplementary tables S3 and S4). These analyses mainly identified metabolic processes such as lipid metabolism and cell growth and proliferation, which is in line with the notion of HCV RNA replication requiring proliferating cells for efficient replication, at least in Huh-7 cells [43], and numerous reports on its requirement on lipid biosynthesis (reviewed in [44]).
While the number of potential HF candidate genes was too large to be functionally validated individually within this study, we surveyed previously published data on HCV host factors, including a manually curated database of HCV-host interactions (VirHostNet [45]) as well as large-scale siRNA-based screens [46], [47], [48], [49]. Whereas such high-throughput approaches exhibit very high false-negative rates [50] and therefore are not suited to exclude HF candidates from our analysis, their false-positive rate is very well controlled and the identified hit genes are highly reliable. Using these data, we could in fact identify 17 of our HF candidates to be implicated with HCV (table 1; marked in red in figure 7C). Six of these genes (JAK1, LHX2, PIP5K1A, RPS27A, PPTC7 and COPA) were found in siRNA-mediated approaches to directly influence HCV replication, as would be expected for a limiting host factor. Five genes (TF, VCAN, TRIM23, SORBS2 and MOBK1B) were identified in a large-scale yeast-two-hybrid based interaction screening [51] to interact with at least one HCV protein (interaction partner listed in table 1). This, however, does not necessarily indicate that the interaction is essential for RNA replication. On similar lines, six further genes (MCL1, SERPING1, CASP8, PIK3CB, GAB1 and APOB) were previously reported to interact with specific HCV proteins in individual studies. Interestingly, most of them (MCL1, CASP8, PIK3CB and GAB1) were implicated with a modulation of apoptosis and cell survival/proliferation, supporting our above analysis, in which “cell growth and proliferation” was found to be an enriched function among the differentially expressed genes (supplementary tables S3 and S4).
Based on our model prediction of a limiting host factor/process involved in the formation of functional replication compartments and utilizing our transcriptomic analysis of differently permissive cells, further studies should be devised aiming to delineate the exact nature of this factor or process. Identification of a cellular function that is essential for HCV replication but naturally limiting in certain cell lines would be very intriguing in terms of pinpointing novel targets for anti-HCV therapy. Such a factor would promise to be inhibitable without critically affecting host cell viability, while severely compromising HCV replication efficiency.
In the present study, we have developed a mathematical model of the intracellular steps of HCV replication. In contrast to previous models [24], [25], [26] we were not only interested in studying steady state replication in stable replicon cell lines, but specifically addressed the highly dynamic initial phase after RNA genome delivery into the host cell. We therefore performed quantitative, time-resolved measurements of viral protein translation as well as strand-specific viral RNA concentrations in two distinct Huh-7 derived cell lines, exhibiting a vastly different permissiveness for HCV RNA replication [32]. With this data, we tried to re-calibrate the most comprehensive HCV replication model available to date [24], but found that the model was not capable of explaining the observed dynamics and ensuing steady state simultaneously. We therefore modified and extended that model by accounting for the “naked”, unprotected nature of the initially transfected in vitro transcribed RNA and by updating of the formation step of the RC and the initiation of negative strand RNA synthesis to the current biological understanding of this process. Under steady state conditions, as studied by previous models, equilibrium of the viral replication machinery with static ratios between cytosolic viral RNA and NS proteins has been achieved. Therefore, in the model by Dahari and colleagues [24], uptake of viral RNA and protein into the replicative compartment could be described by simple first order import reactions. In our setting, however, concentrations for replication competent viral RNA and NS proteins start from zero and grow dynamically in the course of the experiment. Hence, simple first order import reactions do not suffice if the uptake depends on the abundance of more than one species, which is highly likely given biological evidence. Accounting for the above described cis-requirement for initiation of productive replication complexes [35], [36], [37], which means that an RNA molecule can be used as a template for replication only by an NS5B molecule having been translated from that very RNA, we solely allowed a complex of actively translated plus-strand RNA (i.e. translation complexes TC) and cytosolic NS proteins (Ecyt) to be taken up into the RC.
While these model extensions greatly enhanced the fitting quality to the data of a single cell line, we then identified that step in the model, at which an altered kinetic rate could explain the dynamics found in the second cell line as well. For this purpose, we tested a series of hypotheses, fitting the model simultaneously to the two differently permissive cell lines and allowing only those parameters to differ that would be influenced by the host cell in the respective hypothesis. By this approach, we could exclude various processes, e.g. differences in translation efficiency, altered cytoplasmic RNA degradation or different RNA synthesis rates within replication complexes. It is also biologically plausible, that these processes do not differ between the two examined Huh-7 cells lines and therefore cannot explain the observed differences in permissiveness; both, translation and RNA degradation have been shown before to be comparable across different Huh-7 cells [30], and the polymerization rate of the HCV RdRP NS5B is unlikely to depend on host factors (other than ribonucleotides). In principle, a combination of several such processes might be able to explain the observed behavior; however, following Occam's razor, we considered the simplest solution to be the most likely one. Eventually, we identified the formation process of replicative vesicles to be the best suited step in the model, altering the rate of which sufficed to fit the model to measured data from either cell line. We then introduced a host factor (HF) as a new species in our model, and required viral RNA (in the form TC) and NS protein (Ecyt) to form a complex with it in order to allow for the initiation of negative strand RNA synthesis and the formation of the RC. Assumption of different concentrations of this host factor then was sufficient to very accurately explain the differences in RNA replication permissiveness in the two cell lines. This final model therefore completely satisfied all experimental observations and could also correctly predict the effects of targeted perturbations during extensive validation experiments.
We then used the calibrated and validated model to further study individual steps of the viral lifecycle. Sensitivity analysis was applied to pinpoint the most influential steps, perturbation of which would lead to the greatest impact on replication dynamics and yield. A very interesting first finding was that once steady state replication has been reached, the system proved to be relatively robust towards perturbation of individual sub-steps of replication. The degradation rate of RNA inside the RC was the most sensitive parameter under these conditions, and had a significantly higher influence than all other rates. This parameter, however, can hardly be influenced biologically or therapeutically. Very likely, this robustness is key to HCV's prevailing in the face of cellular stress- and innate immune responses [52], [53], [54], [55]. The actual mechanistic basis of this remarkable robustness so far remains elusive.
In contrast, at an early time point after introduction of HCV genomes into the cell, the system was found to be substantially more fragile with respect to the number of sensitive parameters. This suggests that therapeutic intervention with HCV replication by DAAs would be most efficient in newly infected cells, emphasizing the potential of such drugs for the prevention of reinfection upon liver transplantation. The processes found to be most sensitive during the early phase of replication were polyprotein translation as well as the RNA polymerization rate of NS5B. Of note, polyprotein cleavage by the viral NS3/4A protease was surprisingly little influential. This, however, has been described before, e.g. in a study examining the role of cyclophilin A for HCV replication [56]. In that study, viral mutations conferring resistance to the cyclophilin A inhibitor Alisporivir (Debio-025) were shown to significantly affect the efficiency of polyprotein cleavage without notably affecting RNA replication of the replicon [56]. This could raise some concern about the first (very recently) approved direct acting antivirals for HCV, the NS3/4A inhibitors Telaprevir and Boceprevir [57]: on the one hand, they need to exhibit an extremely high potency of inhibition in order to suppress HCV RNA replication efficiently. On the other hand, there should be comparatively little restrictions to the development of escape mutations rendering NS3/4A resistant to the compounds, owing to the relatively small effect on replication dynamics even in a case where the mutation functionally lowers protease activity as it is predicted by our model. Simply put, the virus can effectively buy itself out of pharmacologic inhibition at only modest fitness costs, and in fact, at least for the first generation of protease inhibitors, this is indeed the case [58], [59]. In contrast, according to our model analysis, HCV should be far more sensitive towards inhibition of the NS5B polymerase activity. In line with this prediction, an NS5B inhibitor (HCV-796) yielded a significantly faster and stronger response when directly compared to a very potent protease inhibitor (BILN 2061), both dosed at the same multiples of their respective EC50s [60]. This difference in efficaciousness could even get potentiated when considering the development of escape mutations. Particularly for nucleoside/nucleotide analogues, which target the catalytically active center of NS5B, all so far observed resistance mutations have a negative influence on its polymerase activity [61]. Based on our model, however, lowering NS5B activity is predicted to have a pronounced impact on overall replication efficiency, thereby substantially increasing the fitness costs for such escape mutations. In fact, despite being “genetically easy” (i.e. single nucleotide exchanges suffice) such resistance mutations against nucleotidic inhibitors have been shown to be of negligible clinical relevance due to their extraordinarily strong impact on replication efficiency [62]. In general, we want to note that a modeling approach as ours can help in estimating and understanding the sensitivity of HCV replication upon (e.g. pharmacologic) inhibition of a particular step in the life-cycle. It cannot, however, generally predict the development of resistance mutations, as the actual number and position of nucleotide/amino acid exchanges required for resistance eventually determine the likelihood of their occurrence and their fitness-cost, respectively.
One simplification that we accepted in developing the model is that the formation of the membranous vesicles is modeled as one step (eq. 7) together with the formation of the actual replicase complexes (i.e. the initiation of negative strand RNA synthesis). This is owing to a lack of an experimental handle for the discrimination of “productive” from empty or non-functional vesicles. In fact, it has been shown that the vesicular membrane structures are formed by viral NS protein also in the absence of RNA replication [6], [63]. Therefore it seems likely that initiation of RNA synthesis will depend on the formation of membrane alterations, but still represents a distinct step in the formation of an active replication site. However, in this two-step scenario, membranous vesicles would form based on the concentration of cytosolic NS proteins (Ecyt) and a host factor (HF), and replication complexes (Rip) would mainly depend on Tc (and possibly Ecyt and HF) and the availability of vesicles. In effect, formation of productive replicative vesicles would again depend on those three species, TC, Ecyt and HF and should in principle be compatible with our simplified one-step model.
On similar lines, for reasons of simplicity, our model considers only one single, large replication compartment. This assumption is clearly not correct, as numerous sites of virus induced convoluted membrane structures have been observed in HCV replicating cells [7] and each cell holds approximately 100 negative strand RNAs (i.e. markers for productive replication complexes) on average [10]. However, the approximation with a single large replicative compartment should be adequate provided the real number of vesicles is large enough for formation or loss of individual vesicles not to lead to significant sudden changes of viral RNA and protein availability in the cytoplasm. As measurements of replication are technically limited to bulk assessments and cannot probe individual vesicles, for the time being this point cannot be addressed more adequately. Similarly, there might also be (and likely is) heterogeneity among cells in terms of kinetics and absolute numbers. Also here, probing individual cells for plus and minus strand RNA as well as for polyprotein production is almost impossible with today's technology, and consequently, our model represents an approximation of the average cellular behavior in a larger population of cells.
Curiously, a central result of our study was the conclusion that the assumption of a key host factor was essential to fit our model to the dynamics of RNA replication. This factor was important to explain RNA replication in Huh-7 cells, but might not be as limiting in other HCV permissive cells, e.g. primary human hepatocytes. Moreover, in a physiological setting, restrictions in other steps of the viral life cycle, e.g. sub-threshold receptor levels during entry [64], [65] or a limitation in the apolipoprotein system required for particle secretion [66] might play critical roles as well. Importantly, also the innate immune response (and on a larger time-scale also the adaptive one) poses severe restrictions on viral replication via effector genes, whose molecular identity and functions have only recently begun to be identified [67], [68]. These influences would need to be included in a future, fully comprehensive model of HCV replication. For our present model, based on Huh-7 cells, however, we have so far neglected any impact by the innate immune system, as we could previously demonstrate that presence or absence of functional immune recognition of HCV by the (Huh-7 derived) host cell does not have a measurable effect on its permissiveness [32].
Still, for RNA replication in this single most important cell culture system for HCV, we found a limiting host function involved in the formation of the replication compartment to be crucial to explain the observed replication kinetics. The molecular function of this host factor is still unclear; one or more cellular proteins could be involved, taking part in the formation of the membrane alterations or in the initiation of RNA synthesis. Even a more general condition such as stress tolerance could serve as the host requirement proposed by our model. Since this host factor(s)/condition(s) HF was sufficient to model the varying RNA replication efficiencies in different Huh-7 populations, we performed gene expression profiling to identify genes potentially defining permissiveness. While our analysis identified 355 genes, whose expression correlated with the degree of permissiveness of the respective cell line, there were no single factors or well-defined pathways that stood out significantly. In order to test the limiting nature of these identified factors for HCV RNA replication, one would have to individually overexpress those genes in low permissive cells and assay for an enhancement in HCV replication. Whereas this was beyond the capacity of our current study, we made use of extensive publicly available data on cellular interaction partners of HCV (VirHostNet [45]) and high-throughput RNAi-based knock-down studies [46], [47], [48], [49] in order to recognize genes that had been implicated with HCV before. This approach identified 17 cellular genes whose expression levels on the one hand correlated well with permissiveness for HCV replication, and that, on the other hand, were either reported to at least interact with an HCV protein, or were shown to have a direct impact on HCV replication upon knock-down (table 2). While for this small sub-set of genes a reliable functional link to HCV could therefore be established, we cannot exclude any of the remaining differentially expressed genes as potentially crucial host factors for HCV; this is true even in spite of a virtually genome-wide coverage of the published screening studies, as such approaches are characterized by extremely high false-negative rates [50]. Therefore, comprehensive future studies need to exploit the information contained in our transcriptomic analysis, systematically testing those host factors for an impact on HCV replication that most significantly correlated with permissiveness.
Already during model development, but also throughout our model analyses, the formation and function of the membranous replication compartment was found to be crucial for successful viral HCV replication. Previous literature as well as our model analysis imply that membrane alterations serve at least three distinct purposes. For one, they provide a protected environment for RNA replication, shielding this very sensitive process from the host cell degradative machinery as also shown experimentally before [10]. Without this protection, the viral RNA would quickly be degraded, and replication, according to our model, would become highly vulnerable to stochastic effects due to very low molecule numbers. In fact, should cytoplasmic RNA degradation be only slightly stronger than our mean estimate for μpcyt (but well within its confidence interval), e.g. upon stress or under conditions of an activated immune response, the system would cross a threshold and replication would die off inevitably. Therefore, to compensate for such a lack of protection of the replication machinery, HCV would have to develop a completely different amplification strategy, most likely involving a much higher rate of RNA synthesis in order to maintain sustained replication. This, very likely, would not be compatible with low-level, low profile replication as required for persistence [13]. Secondly, sequestration of viral replicative intermediates, such as double-stranded RNA, into membranous compartments also shields them from recognition by ubiquitous pattern recognition receptors of the intrinsic innate immunity (which, as described above, is neglected by our current model). A third important aspect, however, is the fact that this strict compartmentalization allows for a tight control of viral RNA replication versus protein translation. By limiting the amount of viral and/or host protein inside, the replicative compartment not only protects, but paradoxically also attenuates RNA replication. Presumably, this serves to limit replication to levels sustainable by the cell and permitting low-level persistent replication over a long period of time with very limited detection by the immune system. At the same time, by controlling the amount of newly synthesized RNA released into the cytoplasm, the vesicles indirectly control the amount of protein translation and, in an in vivo situation, particle formation, as was also suggested by another modeling approach [26].
We provide the first comprehensive modeling of the entire RNA replication cycle of a positive strand RNA virus, from the onset of RNA replication to steady state levels. However, membranous replication sites are a hallmark of all positive strand RNA viruses with very different replication strategies. In case of HCV the membranous replication compartment seems to have a rather limiting role in virus RNA replication, probably contributing to viral persistence and chronic disease. In contrast, most positive RNA viruses replicate fast, cause acute diseases and are cleared by the immune system (e.g. the closely related flaviviruses such as Dengue or West Nile virus). Interestingly, in the related group of pestiviruses, pairs of viral isolates have been found, replicating either in a non-cytopathic/persistent or in a cytopathic/acute manner [69]. Upon integration of cellular mRNA sequences into their genomes, dramatically enhancing the efficiency of viral RNA replication, these biotypes switch from well-controlled, persistent infection to an aggressively replicating, cytophatogenic phenotype [70]. Also in case of Sindbis virus, cytopathic replication can be switched to persistence by a single point mutation [71]. Both examples demonstrate a tremendous flexibility to adapt the concept of membranous replication compartments to various replication strategies. It would therefore be highly interesting to use our model as a blueprint for modeling replication kinetics of closely related positive strand RNA viruses following a lytic/acute replication strategy, e.g. Dengue virus or West-Nile-virus. Comparing the principles governing replication of such a virus to the here described strategy of HCV could offer a completely new approach to examining– and eventually comprehending– the general requirements allowing viruses to establish chronicity.
Another obvious yet intriguing direction into which our presented modeling approach could be developed, is extending it to comprise the full infectious virus life cycle, including particle production and secretion, receptor binding and cell entry. In fact, two very recent publications studied RNA replication kinetics upon HCV infection [6], [29] and found a dynamic behavior extremely reminiscent of what we describe here for subgenomic replicons: the initially present RNA is rapidly degraded early upon infection and then starts to replicate exponentially at around 6 to 8 hours post infection, which is reflected in both, plus- and minus-strand RNA signals. This similarity to the kinetics observed in our experiments is remarkable, as initial RNA concentrations are about two to three orders of magnitude less in the infection (roughly 1–50 genomes per cell) as compared to our transfections (∼4.000 genomes per cell). The single major difference to the here described situation in a replicon setting is the increasing excess of plus-strand RNA over the minus-strand for late time points (e.g. 50-fold excess at 72 h) which seems to be due to decreasing minus-strand levels, while plus-strand RNA basically maintains a steady-state [29]. It is intriguing to speculate that this phenomenon might reflect partitioning of the plus-strand RNA into translation/replication on the one hand, and particle assembly/genome encapsidation on the other hand. As encapsidated genomes would no longer be available for initiation of new replication complexes, minus-strand RNA levels should consequently decrease over time. In order to adapt our model to an actual infection setting, however, we will need to switch to a stochastic model to deal with extremely low copy numbers of RNA per cell. Such situations can be addressed mathematically using the Gillespie algorithm, provided appropriate single cell measurements are available. The model could then also be extended to describe the extracellular steps of the viral life cycle, up to receptor binding and cell entry, which could finally allow for very precise simulation of viral spread through a population of naïve cells. Such a comprehensive model would be highly valuable to examine and predict the effects of therapeutic intervention with viral entry or release as compared to inhibition of intracellular steps of replication. Even more importantly, it could be suited to finally link our fine-grained molecular model of HCV replication to the very interesting patient-level models of HCV infection and therapy dynamics [14], [72], and thereby open up new avenues to rationally designing novel therapeutic strategies, but also to understanding the effects of molecule-scale events onto the progression of a complex disease.
All cells were maintained in supplemented Dulbecco's modified Eagle medium (DMEM) as described previously [10]. Huh-7 low passage refers to naïve Huh-7 cells, passaged less than 30 times in our laboratory, see also Binder et al. [32]. Huh7-Lunet and Huh-7/5-2 are highly permissive clonal cell lines [32]. Huh7-Lunet NP (unpublished) refers to a derivative of Huh7-Lunet, which is significantly less permissive than its parental cell line.
For kinetic analyses of HCV RNA replication, the genotype 2a (JFH1 isolate) constructs pFKi389LucNS3-3′_dg_JFH (wild-type) and pFKi389LucNS3-3′_dg_JFH/ΔGDD (replication deficient) [73] were used, as well as the NTR-chimeric constructs pFK-I341PI-Luc/NS3-3′/JFH1/5′Con (5′-NTR exchange) and pFK-I341PI-Luc/NS3-3′/JFH1/XCon (3′-NTR exchange) [22]. Permissiveness of cell lines was assessed using a genotype 1b (con1) replicon, using the plasmid pFK-I341PI-Luc/NS3-3′/Con1/ET/∂g. In vitro transcription of HCV replicons was performed as described previously [22], [30]. Briefly, plasmid DNA was purified by phenol/chloroform extraction and transcribed with 0.9 U/µl T7 RNA polymerase (Promega). RNA was then purified by DNase (Promega) digestion, extraction with acidic phenol and chloroform and room temperature isopropanol precipitation. RNA concentration was determined spectrophotometrically and integrity was confirmed by agarose gel electrophoresis.
Cells were transfected with in vitro transcribed HCV RNA by electroporation as described previously [22]. For determination of host cell permissiveness (figure 7), 5 µg of RNA were used for electroporation and cells were seeded into 6-well plates (1/12 electroporation per well). Samples were lysed at 4, 24, 48 and 72 h post transfection and stored at −80°C until measurement of luciferase activity. For time resolved quantitation of HCV replication, 4×106 cells were transfected with 10 µg of HCV RNA, corresponding samples were pooled and cells were seeded into 6-well plates for luciferase assays as described above or into 10 cm cell-culture dishes at a density of 4×106 cells per plate (2×106 cells/plate for time points 48 h and 72 h) for RNA preparation and Northern blotting. For the 0 h RNA sample, 4×106 cells were washed twice with DMEM directly after electroporation, pelleted and lysed in guanidinium isothiocyanate. Other samples were lysed at the indicated time points (2, 4, 8, 12, 18, 24, 48 and 72 h) and lysates were stored at −80°C until further processing.
For determination of HCV replication by luciferase activity measurement, all samples of one experiment were frozen at −80°C upon harvesting and thawed simultaneously prior to luciferase detection. Measurements were performed as described in Binder et al. [22], with all samples measured in duplicate. Luciferase activity was normalized to the input activity assessed at 2 h (kinetic experiments) or 4 h (permissiveness determination) post electroporation, to correct for transfection efficiency.
RNA preparation and Northern blotting were performed according to established procedures [22]. In essence, total cellular RNA was isolated from guanidinium isothiocyanate lysates by a phenol/chloroform based single-step protocol and denatured in glyoxal. Samples were analyzed by denaturing agarose gel electrophoresis and Northern hybridization. For strand specific detection of HCV RNA, radioactively labeled riboprobes encompassing nucleotides 6273 to 9678 of the JFH1 sequence were generated by T7- (minus-strand detection) or T3-polymerase (plus-strand detection) mediated in vitro transcription of plasmid pBSK-JFH1/6273-3′ [34]. Signals were recorded by phosphorimaging using a Molecular Imager FX scanner (BioRad, Munich, Germany) and quantified using the QuantityOne software (BioRad). To determine absolute molecule numbers, signals were quantified using serial dilutions of highly purified plus- and minus-strand in vitro transcripts of known quantity, which were loaded onto the same gel. Cross-hybridization of minus-strand probes with the plus-strand standard was observed to a low extent and corrected for.
Permissiveness of eight Huh-7 derived cell-lines was assessed using a standard luciferase replication assay as described above. Total cellular RNA of untransfected cells was then isolated by Trizol extraction according to the manufacturer's protocol (Invitrogen, Karlsruhe, Germany), and gene expression was measured using the Affymetrix Human Genome U133 Plus 2.0 platform. Data were normalized in R/Bioconductor using RMA normalization. Genes were filtered using the variance-based (IQR) filter in nsFilter, and log2 fold-changes between high and low permissive cells were computed. We then fitted a linear model to the data, predicting replication efficiency in the eight cell lines from the corresponding gene expression values. ANOVA was used to assess statistical significance of individual genes. Hit selection was done using a relatively low threshold of 0.2 on the p-value and a log fold-change of at least 0.3, corresponding to a change in expression of approximately 25%. Resulting genes were intersected with published RNAi screening [46], [47], [48], [49] and virus-host protein interaction [45] data as described, yielding a list of 17 host factors that are differentially expressed between the high and low permissive cells, that correlate with replication permissiveness in the eight cell lines used, and that have previously been shown to be associated with HCV infection or replication. Genes were then mapped to pathways and annotated further using DAVID version 6.7 [74], [75] and IPA (Ingenuity Systems, www.ingenuity.com).
We developed a mathematical model using ordinary differential equations based on mass action kinetics. The model is subdivided into two compartments: 1) initial RNA processing, translation into the polyprotein and polyprotein processing (cleavage) occur in the cytoplasm, and 2) viral genome replication takes place inside of the replication compartment. A graphical summary of the model is shown in Figure 2C. The following set of equations was used to describe the processes in the two compartments:
Cytoplasm(1)(2)(3)(4)(5)(6)Here, Rpunp (eq. 1) represents the number of plus-strand RNA molecules entering the cell upon transfection. This transfected RNA is processed into translation competent Rpcyt (eq. 2) at rate k0, describing, for example, transport and structural re-folding processes. The processed plus-strand RNA Rpcyt interacts with ribosomes Ribo at a constant rate k1 to form translation complexes Tc (eq. 3), which degrade at rate μTc. Ribosomes are recovered when translation complexes Tc degrade with rate μTc. Note that, as the total number of ribosomes in the cell (Ribotot) is assumed constant, the number of ribosomes available for translation is given by Ribotot – TC, and it is not necessary to introduce a separate equation for ribosomes. Unprocessed and processed RNAs Rpunp and Rpcyt degrade with rate constants μpunp and μpcyt, respectively (eq. 1 an 2). For simplicity, we assume that 10 ribosomes simultaneously translate the same HCV RNA [76], therefore, Ribotot represents complexes consisting of 10 ribosomes. Viral polyprotein P is formed from Tc at an effective rate k2 (eq. 4). When the translation of polyprotein is complete, the translation complex dissociates into plus-strand RNA and ribosomes at rate k2. Newly produced polyprotein is cleaved with rate kc into the mature viral nonstructural (NS) proteins Ecyt (eq. 5). NS proteins degrade at rate μEcyt. Eventually, plus-strand RNA and NS proteins, most notably the polymerase NS5B, interact in cis and together with NS proteins in trans (Ecyt) as well as a cellular factor HF to form a replication complex within the induced vesicular membrane structure. This cis interaction of Rpcyt and translated NS proteins is realized in the model by requiring active translation complexes Tc instead of free Rpcyt for the formation of replication complexes. The host factor HF catalyzes the formation of RIp, at the rate kPin. Once RIp is formed, ribosomes are freed again at rate kPin. This leads to the ternary reaction TC+ECyt+HF→RIp+RIbo, simultaneously describing formation of the replication compartments and initiation of minus strand RNA synthesis, compare also supplementary text S1 and supplementary figure S6. In turn, HF is freed again when RIp degrades or upon completion of minus strand synthesis. As the total number of host factor molecules in the cell is assumed constant, we can replace HF by HF(0) – RIp, where HF(0) is the total number of HF molecules in the cell. Lastly, since we use a luciferase readout to measure polyprotein concentration, we furthermore include a luciferase marker L in the model, which is produced at the same rate as the polyprotein (k2), however does not require further processing and degrades with rate μL (eq. 6).
Replication compartment(7)(8)(9)(10)(11)RIp is the minus-strand RNA initiation complex (eq. 7), which contains a plus-strand RNA serving as template for the synthesis of minus-strand RNA. Minus strand RNA is synthesized from RIp at rate k4m, yielding double stranded RNA Rds (eq. 8). We assume minus-strand RNA to be always bound to its complementary plus-strand in a double-stranded replicative intermediate. When the production of minus-strand RNA is complete, RIp dissociates into Rds, HF and viral NS protein E (eq. 9). Next, Rds interacts again with E to form a plus-strand RNA initiation complex, RIds (eq. 10), to initiate the synthesis of new plus-strands, Rp, with a constant rate k4p, and dissociates into Rds and E. Newly synthesized plus-strand RNA, Rp (eq. 11), then leaves the replication compartment at rate kPout to participate in translation, or interacts with the polymerase E and host factor HF to again form the minus-strand RNA initiation complex RIp at rate k3. For simplicity, we assume that the RNA RIp, Rds, RIds and Rp, and proteins E all degrade with rate μRC.
Reaction rates in the model were taken from literature as far as known, or estimated by fitting the model to the experimental data. Following Dahari et al [24], we used a value of k2 = 100 polyproteins per hour per polysome for protein translation. RNA replication was assumed to occur at a rate of k4m = k4p = 1.7 viral RNA molecules per hour per replication complex, assuming plus- and minus-strand synthesis to occur at the same rate [23], [77], [78]. Based on an estimated half-life of Luciferase of approximately 2 hours, we estimated the corresponding degradation rate to be μL = 0.35 h−1 [79], [80]. We furthermore estimated the NS protein half-life in the cytoplasm to be around 12 hours, corresponding to a rate of μEcyt = 0.06 h−1 [76], [81], [82]. We observed from model calibration that the optimization would yield values with μTc>μpcyt, violating the expectation that RNA in translation complexes should be more stable than free RNA in the cytoplasm. We hence added the constraint μTc/μpcyt = 0.5, enforcing a 2-fold higher stability of RNA that is actively translated. We furthermore observed a low sensitivity of model output with respect to parameters k1, kc, k3 and k5, compare figure 5, and hence fixed these parameters based on manual model analysis, for details see supplementary text S1.
Estimation of the remaining 7 model parameters, 3 initial values and a scale factor to convert luciferase measurements into polyprotein molecule numbers was done using multiple shooting, as implemented in the PARFIT package [83], [84], [85]. We simultaneously minimized the least squares prediction error on the high and low permissive cells in log-concentration space, using all individual measurements in the objective function. An additional scaling factor was introduced in the optimization problem to convert luciferase measurements for the viral polyprotein to molecule numbers. All model species containing viral plus-strand RNA or minus-strand RNA, respectively, were added for comparison with the experimental data, yielding Rptot = Rpunp+Rpcyt+Tc+RIp+Rds+RIds+Rp for the total plus-strand RNA and RMtot = Rds+RIds for the total negative strand RNA concentrations. Ratios of RNA as reported in literature were furthermore used to constrain the optimization [10]. As some species attain very low values, we compared results of the approximation using differential equations with a stochastic solver (supplementary figure S7). For details of the parameter estimation and objective function used see supplementary text S1. Obtained model parameters and confidence intervals are shown in table 1.
To test our model for structural identifiability, we performed a local identifiability analysis at obtained optimal parameter values using SensSB [86]. Results of this analysis are shown in Supplementary Figure S8. High correlation between two parameters means that a change in the model output caused by a change in one parameter can be compensated by an appropriate change in the other parameter. This then prevents the parameters from being uniquely identifiable despite the output being very sensitive to changes in individual parameters. Parameters for which values were known from literature or which were fixed were also included in this identifiability analysis, to assess their effect on results. These parameters are indicated in grey in the Figure; several of these parameters are highly correlated with other parameters, thus reiterating the importance of experimental measurements for them. Importantly, the identifiability analysis indicates that most of the parameters that had to be calibrated from data showed low correlation to other parameters only, indicating an overall satisfactory identifiability of the model and, in particular, no indication of structural non-identifiability in the model with correlation values close to ±1.
We furthermore calculated confidence intervals on estimated model parameters using the covariance matrix of the parameters, as described in supplementary text S1. Most of the kinetic reaction rates had reasonable standard errors and confidence bands, while larger uncertainties were observed for the initial values, compare table 1. This sloppiness is typical for models in systems biology [87], [88]. Based on our aim to develop a predictive model and not uniquely identify individual reaction rates, our assessment was that the model is sufficiently identifiable for our purpose.
Global sensitivity analysis was performed using the extended Fourier Amplitude Sensitivity Test (eFAST) [39], [40]. This algorithm calculates the first and total-order sensitivity indices of each parameter, and assesses the statistical significance of these sensitivity indices by a method based on dummy parameters. For details, we refer to Saltelli et al [89]. In brief, for a given model y = f(x) with scalar y and input vector x = (x1, …, xn), the first order sensitivity index with respect to xi is the expected amount of variance that would be removed from the total output variance, if we knew the true value of xi, divided by the total unconditional variance:Si is a measure of the relative importance of the individual variable xi in driving the uncertainty in the output y. In contrast, the total sensitivity index with respect to a variable xi measures the residual output variance if only xi were left free to vary over its uncertainty range, and all other parameters were known:STi is a measure of how important a parameter is in determining the output variance, either singularly or in combination with other parameters. To assess the significance of obtained indices, eFast furthermore calculates the first and total order sensitivity index for a dummy parameter that is not part of the model. Indices that are not significantly larger than this dummy parameter index should not be considered different from zero [39].
Figures 6 and S4 show the resulting eFAST total order sensitivity indices of viral plus- and minus-strand RNA concentrations and viral polyprotein concentration with respect to the 16 model parameters and three initial values at two different time points, early in the viral lifecycle and after attainment of the steady state levels.
|
10.1371/journal.pntd.0007548 | The effect of amantadine on an ion channel protein from Chikungunya virus | Viroporins like influenza A virus M2, hepatitis C virus p7, HIV-1 Vpu and picornavirus 2B associate with host membranes, and create hydrophilic corridors, which are critical for viral entry, replication and egress. The 6K proteins from alphaviruses are conjectured to be viroporins, essential during egress of progeny viruses from host membranes, although the analogue in Chikungunya Virus (CHIKV) remains relatively uncharacterized. Using a combination of electrophysiology, confocal and electron microscopy, and molecular dynamics simulations we show for the first time that CHIKV 6K is an ion channel forming protein that primarily associates with endoplasmic reticulum (ER) membranes. The ion channel activity of 6K can be inhibited by amantadine, an antiviral developed against the M2 protein of Influenza A virus; and CHIKV infection of cultured cells can be effectively inhibited in presence of this drug. Our study provides crucial mechanistic insights into the functionality of 6K during CHIKV-host interaction and suggests that 6K is a potential therapeutic drug target, with amantadine and its derivatives being strong candidates for further development.
| Chikungunya fever is a severe crippling illness caused by the arthropod-borne virus CHIKV. Originally from the African subcontinent, the virus has now spread worldwide and is responsible for substantial morbidity and economic loss. The existing treatment against CHIKV is primarily symptomatic, and it is imperative that specific therapeutics be devised. The present study provides detailed insight into the functionality of 6K, an ion channel forming protein of CHIKV. Amantadine, a known antiviral against influenza virus, also inhibits CHIKV replication in cell culture and drastically alters the morphology of virus particles. This work highlights striking parallels among functionalities of virus-encoded membrane-interacting proteins, which may be exploited for developing broad-spectrum antivirals.
| Chikungunya fever is a severe and debilitating illness caused by the mosquito-borne arbovirus, Chikungunya Virus (CHIKV) [1–4]. Infections are generally non-fatal, but this virus has been much in the limelight lately, due to its rapid spread and outbreaks worldwide [2–4]. Although, India is endemic for CHIKV; outbreaks do occur, the latest one reported in New Delhi in 2016 [4]. Currently, there are no specific therapeutics against CHIKV infections, the treatment being primarily symptomatic [1].
CHIKV, like other alphaviruses, is an enveloped RNA virus with particle diameter ranging between 65–70 nm [5]. The viral genome is organized into structural and non-structural protein-encoding regions [5]. The structural protein cassette is composed of glycoproteins (E1-E2-E3), capsid protein (C), and 6K, which was recently shown to have a transframe variant (TF) [6, 7]. 240 copies each of glycoproteins E1 and E2; as well as the capsid protein, are arranged in accordance with T = 4 symmetry [8]. Although the roles of capsid and envelope proteins in the life cycle of CHIKV are fairly well studied [9–14], reports pertaining to the direct functional characterization of 6K are rare, making it the least understood amongst all CHIKV structural proteins. A recent report suggests that 6K is a prime target for mounting CTL- mediated immune response in the host, indicating its significance as a therapeutic target [15]. One contributing factor for the lack of direct biochemical characterization of CHIKV 6K is its extreme intrinsic hydrophobicity, as well as cytotoxicity [7, 16], which makes the production of sufficient quantity of functionally active, recombinant protein fairly challenging. In fact, the majority of the studies on 6K from other alphaviruses—demonstrating the biochemical properties and role in viral life cycle—have been carried out by analysis of mutated virus [17–23] or by RNA expression at the cellular level [7, 24], with only a few studies characterizing recombinant 6K [16, 25]. Further, the molecular details of membrane association by 6K and the exact role of this activity in promoting virus budding remains unknown.
Recently, a transframe (TF) variant of 6K was identified, which is generated from the 6K gene as a result of a (-1) ribosomal frameshift. TF has the same N-terminal domain as 6K, but a different C-terminus [6, 7]. 6K and TF are produced during infections by most alphaviruses, and both are thought to be essential for virus budding, although only TF appears to be packaged within virions while 6K is probably retained at the membranes of infected cells [6, 7]. Additionally, studies with SINV have highlighted the role of palmitoylation for the localization of TF to the plasma membrane [26].
The membrane permeabilization activity of alphavirus 6K places it in the category of viroporins. Nucleation of aqueous passageways by viroporins allows movement of ions and small molecules, which in turn, facilitates virus entry, replication, and egress. Some well-characterized members of the viroporin family include M2 of influenza A virus, Vpu of HIV, p7 of HCV, and 2B of picornaviruses [27]. The role of viroporins in sustaining infections, and their significance as targets for drug development, is illustrated by the antiviral activity of amantadine, which targets the ion-channel forming protein M2 and prevents Influenza A infections [28].
Given the recent scenario of frequent outbreaks of Chikungunya fever in different parts of the world [1–4], and the role of 6K in supporting alphavirus infections [17–23], we attempted to functionally characterize CHIKV 6K and assess its potential as a target for drug development. Using a combination of electrophysiology, cryo-electron microscopy, biophysical techniques, and molecular dynamics simulations, we demonstrate that CHIKV 6K interacts with membranes in multifaceted ways, leading to permeability as well as vesicle fusion. We show that CHIKV 6K exists in oligomeric forms, and forms ion channels in membranes. In addition, we demonstrate that virus-like particles (VLPs) of CHIKV show a marked deviation from their usual morphology when treated with amantadine; and that the inhibitory activity of amantadine extends to CHIKV replication in cell culture. Thus, it can be strongly emphasized that 6K is critical in CHIKV biology and is a potential therapeutic target for the treatment of Chikungunya fever.
The 6K proteins of alphaviruses are extremely hydrophobic in nature with more than 50% of the residues having a positive value on the hydrophobicity scale (Fig 1A). A GRAVY (grand average of hydropathy) value of 1.006 also indicates that the protein has an overall hydrophobic nature. This, as well as the presence of transmembrane regions predicted by several servers (Fig 1B), were likely responsible for our inability to generate CHIKV 6K alone, in soluble form, by recombinant expression in bacteria.
A fusion of CHIKV 6K with GST (Glutathione S Transferase) was successfully expressed and purified from E. coli Rosetta (pLysS cells). We found that IPTG induction and subsequent bacterial growth at a relatively lower temperature of 18°C, is an essential requirement for maximizing expression. GST-6K was initially purified by pulldown with GST beads (Fig 1C), followed by size-exclusion chromatography, which generated two separate peaks corresponding to GST and GST-6K (Fig 1D). The GST-6K peak was fairly broad and the molecular weight corresponding to the primary peak fraction (Fig 1D, indicated with an arrow) was calculated to be 190.98 kDa, which indicated the formation of a hexamer (monomer MW = 33.5 KDa). Since the recombinant protein appeared to have the propensity to break into its constituent parts (Fig 1D), GST-6K was utilized within 24–48 hours post purification for every experiment.
The ability of GST-6K to form ion channels in vitro was investigated using an electrophysiology setup as described in Materials and Methods. The ion channel activity of GST-6K was measured as a flow of current across an otherwise intact DPhPC membrane (Fig 2A and 2B). Fig 2A and 2B (i, iii, v and vii) represent the current versus time traces of GST-6K incorporated into the lipid bilayer membrane at applied membrane potentials of -100 mV and +100 mV respectively. The corresponding all point histograms are shown in Fig 2A and 2B (ii, iv, vi and viii). Addition of GST-6K to the membrane caused spikes in the current trace, which displayed the typical stepped nature of the opening and closing of an ion channel. This showed that GST-6K successfully incorporated within the bilayer and formed ion-conducting, stable channels, which were functional under the aforementioned membrane potentials, and opened at both negative and positive voltages.
Altogether, observations were made at four different protein concentrations of 0.2, 0.85, 1.2 and 2.6 ng/μl. The lipid (DPhPC) concentration remained fixed in all cases, as the formation of bilayer membrane in the BLM cup demands an exact amount of lipid. Increasing concentration of GST-6K coincided with the appearance of multiple open states with higher currents, at both positive and negative membrane potentials. These observations can be best explained as follows. GST-6K forms pore on the membrane by oligomerization (S1 Fig), however, the number of monomers which oligomerize to form a pore is not fixed. Increase in protein concentration may result in the generation of higher-order oligomers, leading to the formation of bigger pores in the membranes and higher currents (S1 Table). Generation of multiple oligomeric states is well evidenced in pore-forming proteins [29]. Similar experiments with only GST as negative control do not result in the formation of ion channels, as seen by us (S2 Fig) and others [30]. This indicates that the ion-channel forming activity is exclusively due to 6K.
Since GST-6K appeared to be capable of forming ion-channels in membranes, we attempted to determine if it preferentially damages target membranes. Liposomes mimicking the lipid composition of ER and plasma membranes encapsulating the fluorescent dye sulforhodamine B, were generated as described [31, 32]. The ability of GST-6K, at a concentration ranging from 0.1 μM-10 μM, to disrupt these liposomes, was tested using standard methods. A distinct increase in disruption of ER mimicking liposomes was observed with concentrations of 0.5 μM (and higher) of GST-6K (Fig 3A). Transmission electron micrographs of GST-6K treated, vs non-treated ER-mimicking liposomes, displayed a clear difference in morphology (Fig 3B and 3C), with the latter image (Fig 3C) showing liposomes with their surfaces decorated with proteinaceous material. Incubation of plasma membrane mimicking liposomes with GST-6K did not cause any significant release of dye (Fig 3A), or alteration of surface features (Fig 3D and 3E), indicating the inability of GST-6K to effectively rupture these membranes.
To further pinpoint the role of specific lipids in GST-6K-mediated membrane damage, similar experiments were conducted with SulfoB-encapsulating liposomes composed of DOPC (Dipalmitoylphosphatidylcholine) and cholesterol in a 1:1 molar ratio. GST-6K, even at relatively higher concentrations of 5 and 10 μM, was incapable of disrupting these liposomes (Fig 3A). These observations, taken together, indicated that CHIKV 6K is probably unable to damage membranes rich in cholesterol. Free, purified GST did not display any significant ability to disrupt membranes, indicating that the GST tag did not influence the effect of CHIKV 6K on membranes in any way (Fig 3A).
Incubation of ER-mimicking liposomes with GST-6K and subsequent visualization using transmission electron microscopy, revealed that the protein dotted the surface of liposomes (Fig 3C). Manual picking of particles and 2D classification (Fig 3F) indicated that the morphology of the structures formed was not uniform. Thus, it appears that GST-6K may induce the formation of a range of oligomeric structural units on the surface of membranes.
The inability of CHIKV 6K to cause leakage in plasma membrane specific liposomes was somewhat surprising, as the stipulated role of 6K during virus particle egress implies some association with the plasma membrane [21]. We, therefore, tested the cellular location of 6K tagged with EGFP (Enhanced Green Fluorescent Protein) in mammalian epithelial cells using confocal microscopy. Upon transfection of EGFP-6K into HEK (Human Embryonic Kidney) 293T cells, the GFP fluorescence showed a very high degree of localization with the ER, partial localization with Golgi membranes, but no localization to mitochondria, nucleus or the plasma membrane (Fig 4A–4E). HEK-293T cells were also co-transfected with pCDNA3.1 containing either CHIKV E2 or E1 glycoproteins, along with GFP-6K. The E1 and E2 glycoproteins were expressed in conjunction with an N-terminal myc-tag, which was detected with an anti-rabbit HRP conjugated antibody (Fig 4F and 4G). 6K selectively localized with E2 as opposed to E1 and traversed to the plasma membrane upon simultaneous expression with E2 (Fig 4H). This data, in conjunction with previous instances of E2 trafficking to the plasma membrane by itself [33–35], suggests that the physiological role of 6K in infected cells is quite possibly dependent on E2, which might dictate the trafficking of 6K to the plasma membrane.
In order to better understand the mode of membrane association and ion channel formation by CHIKV 6K, we carried out microsecond scale (1.5 μs) MD simulation studies. In the absence of an experimentally derived tertiary structure for 6K, two different predicted structures - 6k1 and 6k2—were considered (Fig 5A-i and 5B-i respectively). All simulations were carried out for a total of 1.5 μs in the presence of 100 mM concentration of NaCl under the influence of an applied electric field.
When a single molecule of 6k1 was embedded within a pre-equilibrated POPC (1-pamitoyl-2-oleoyl-sn-glycero-3-phosphocholine) membrane the peptide adopted an overall angular conformation with the central alpha-helical segment (14QQPLFWLQALIPLAALIVLCNCLR37) attaining an almost 15-degree tilt with respect to its initial conformation. The peptide retained this conformation for the rest of the simulation period.
However, embedding 6 molecules of 6k1 in the membrane led to the formation of a small stable channel through which ions were observed to pass through (Fig 5A-iv). Secondary structural analysis of the individual monomers (S3 Fig) showed that 6k1 primarily remains helical during simulation time, with the central alpha-helical segment of each monomer spanning through the length of the membrane, thereby providing structural stability to the overall complex. The average size of the channel was found to be ~1 nm, which is sufficient for ions and small molecules like sulforhodamine B to pass through and also corresponds closely to the channel diameter obtained from electrophysiology experiments. We utilized the “Pore Walker” server [36] to detect the residues within the channel and found that the channel lumen was primarily lined by residues Ala1, Thr2, Tyr3, Glu5, Ile24, Pro25, Ala28, Leu32 and Arg37 from each monomer.
Similar simulations studies with the oligomeric form (Fig 5B-ii) of 6k2 produced an interesting outcome (Fig 5B-iii, iv). No stable channel for ion passage, unlike that observed for 6k1 oligomers, were detected in these cases (Fig 5-iv). A plausible explanation for this discrepancy was provided by the difference in the orientation of the N-terminal helix in 6k1 and 6k2 conformations (Fig 5A-i and 5B-i respectively). We generated two potential oligomeric arrangements with the best possible spatial placement of 6k1 and 6k2. The 6k1 oligomer contains the N-terminal helices positioned towards the interior of the assembly (Fig 5A-ii), whereas the 6k2 oligomer has the N-terminal helices projected outwards (Fig 5B-ii). The hydrophobicity plot for 6K (Fig 1A) clearly shows that, except the small N-terminal stretch of approximately 15–20 residues, the rest of the sequence is considerably hydrophobic. Thus, as the simulation progressed, the inward orientation of N-terminal helices of 6k1 probably favored the stabilization of the channel, widening it enough for the passage of small molecules; whereas, for 6K2, the cavity collapsed on itself, thus hindering the formation of a stable channel. Our simulation studies provide indications as to how CHIKV 6K forms ion channels within biological membranes and the dynamics of ion conduction and hints towards the critical role of the N-terminal residues of 6K in the formation and stabilization of the ion channel.
During the visualization of GST-6K interaction with ER mimicking liposomes by transmission electron microscopy (Fig 3C), the formation of a population of liposomes with larger diameter was noticed. Further studies clearly showed the occurrence of liposomal fusion (Fig 6A), which was further investigated using dynamic light scattering (DLS) (Fig 6B). First, 1 mM calcium chloride was utilized as a positive control, as calcium ions are known to facilitate fusion of artificial vesicles in vitro [37]. Addition of calcium chloride resulted in a size-shift of vesicles to larger liposomes (Fig 6B). Upon exposure of ER-mimicking liposomes to 1 μM GST-6K, a distinct overall increase in the diameter of vesicles was noted, indicating the presence of larger sized structures (Fig 6B). This additional, fusogenic ability of CHIKV 6K appears to be similar to that of the M2 channel protein of Influenza A virus [38]. Given these similarities between Influenza A M2 and CHIKV 6K, we attempted to check whether the membrane fusion events orchestrated by CHIKV 6K could be prevented by amantadine, a well-known inhibitor of M2 [38]. The fusion of ER-mimicking liposomes by GST-6K was entirely abrogated by 1μM amantadine (Fig 6B). Addition of the same amount of amantadine also inhibited the release of fluorescent dye from ER mimicking vesicles by GST-6K to a significant proportion (Fig 6C).
The effect of amantadine on the ion channel formation by GST-6K was checked by electrophysiology experiments described previously at both positive and negative membrane potentials (Fig 6D and 6E), addition of 1 μM amantadine to 2.6 ng/μl GST-6K on the bilayer (BLM) resulted in the formation of only closed states (Fig 6D and 6E, iii-iv), while clear close and open states were detected in the absence of amantadine (Fig 6D and 6E, i-ii). This clearly indicates that amantadine abolishes the ion channel activity of CHIKV 6K. It may be mentioned here that given the concentration of the protein the recordings show that 6K forms multi-channels, which are likely to be due to the formation of different oligomeric states of the protein on the BLM.
To understand whether the ability of amantadine to inhibit the membrane activity of 6K translates to any effect on CHIKV particle assembly, we transfected HEK 293T cells with the cDNA encoding the entire structural protein cassette of CHIKV, followed by treatment of transfected cells with 1μM amantadine. Transfection of the structural protein cassette resulted in the formation of CHIKV VLPs, which were purified and observed using cryo-electron microscopy (Fig 7A). While the majority of particles generated from untreated cells (~74%) displayed the usual size and morphology of wild type CHIKV (Fig 7A), particles generated from amantadine-treated cells were relatively smaller, heterogeneous, and in some cases appeared to lack lipid-associated glycoproteins (Fig 7B). Upon detailed visual examination of ~1000 particles from each population, 89.29% of particles generated from amantadine treated cells exhibited aberrant morphology, in contrast to 24.79% aberrant particles generated from untreated cells (Fig 7C). We postulate that the deviation of CHIKV particles from standard size and morphology is possibly due to the detrimental effect of amantadine on the membrane activity of 6K (Fig 6D and 6E). Taken together, our data highlights the necessity of 6K-mediated membrane interaction for correct virus particle assembly.
To check whether the inhibitory activity of amantadine on correct virus assembly extends to the infectivity of virus particles, the ability of CHIKV (strain S 27) to replicate in vero cells was tested in presence and absence of amantadine. First, any cytotoxic effects of amantadine on vero cells was tested with different concentrations of the drug (25–200 μM) for 24 h. MTT assay showed that while ~100% cells were viable upon being treated with 100 μM of amantadine, the cellular viability decreased to 95% upon treatment with 150 μM of the drug, and was further reduced to 88% with 200 μM of the drug (Fig 8A). To check the effect of amantadine on CHIKV replication, a dose kinetics assay was performed (Fig 8B), in which vero cells were either mock infected or infected with CHIKV (MOI 0.1) and treated with different concentrations of the drug (2.5–200 μM). Cell culture supernatants were harvested at 18 hpi and plaque assay, as well as qRT-PCR, were carried out to estimate the virus titer. As observed in Fig 8B, around 58% reduction in virus titer was observed in samples treated with 40 μM concentration, while 77% reduction was observed at 100 μM concentration of amantadine in comparison to control. For further confirmation, qRT-PCR for CHIKV E1 gene was carried out. It was observed that the E1 gene expression was decreased significantly with increased concentration of amantadine as evident from the Ct values shown in Table 1.
Additionally, for assessing the effect of amantadine on CHIKV propagation at different time points post-infection, a growth kinetics experiment was performed (Fig 8C). Vero cells infected with CHIKV (0.1 MOI) and treated with 60 μM of amantadine, 90 minutes post infection, were harvested at 6, 12 and 18hpi, and the harvested cell culture supernatants were processed through a plaque assay to estimate the viral titer. It was observed that there was ~38% reduction in viral titer at 6hpi and ~60% reduction at 12hpi and 18hpi as shown in Fig 8C. Taken together, the results indicate that amantadine can inhibit viral infection and shows significant anti-CHIKV effect under in vitro growth conditions.
Although different virus families have distinct mechanisms for host interaction, however, there exist notable similarities in the processes of entry, uncoating, replication, and egress. Some examples are membrane fusion for cellular entry of enveloped viruses, amphipathic peptide-mediated membrane disruption by non-enveloped viruses, and viroporin-mediated membrane alteration/remodeling for facilitating viral propagation [39–41]. These analogous steps in the virus-host interaction pathway [42, 43] can potentially be targeted for generating broad-spectrum antivirals. This approach may ultimately be more cost-effective than engineering separate therapies against specific viral pathogens.
Viroporins constitute a group of virally encoded membrane-interacting proteins that are crucial for initiating and maintaining successful viral infections [41] and consequently are suitable therapeutic targets. The 6K protein from alphaviruses, due to its ion channel forming ability, and its critical role in facilitating virus budding, has been considered a member of the viroporin family [16, 21, 24]; although the analogue protein from CHIKV has not been characterized functionally. Here, we show for the first time that the 6K protein from CHIKV associates into oligomers and interacts with membranes–two essential features which categorize virally encoded proteins as viroporins. Interestingly, our data show that CHIKV 6K interacts with membranes in multifaceted ways–it can induce ion channel formation, allow passage of small molecules like fluorescent dyes, and also facilitate fusion of vesicles. As this kind of complex membrane association was earlier identified for the M2 protein of Influenza A Virus, also categorized as a viroporin, we attempted to test whether an existing, FDA-approved and marketed M2 ion channel inhibitor–amantadine- can affect the functionality of CHIKV 6K. Our data showed that indeed, amantadine at a concentration of 1μM [45] was able to inhibit ion channel activity, vesicle fusion and membrane permeabilization properties of 6K in vitro. We further hypothesized that if 6K is required for correct budding of CHIKV particles from cells, the inhibition of 6K functionality by amantadine would lead to the production of virus particles with a significant degree of aberrance in morphology, similar to the phenotype earlier observed upon deletion of 6K in other alphaviruses [20, 21]. We found that indeed, amantadine at a concentration of 1μM was sufficient to cause structural aberrations in a majority of CHIKV virus-like particles budding out of cells. This effect of amantadine on the activity of CHIKV 6K is surprising, given the lack of primary sequence similarity between 6K and M2 [44]. However, there could be similarities in the tertiary or quaternary folds of M2 and 6K, which might allow amantadine binding and is worthy of further investigation through structural studies.
Given the effect of amantadine on particle morphology, we tested the possibility of utilizing amantadine as an inhibitor of CHIKV infection. Vero cells infected with CHIKV virions (strain S 27), showed a significant decrease in viral titer upon exposure to 5 μM or more amantadine (Fig 7B). This concentration was higher than that required for alteration of particle morphology, which indicates that during infection, there could be other compensatory factors resulting in the generation of infectious particles, or that the morphologically altered particles do retain some infectivity. The IC50 value of amantadine was calculated to be 29.51 μM (Fig 7B), which is comparable to that observed for other antiviral drugs [46]. Additionally, the qRT-PCR assay also confirmed that the expression of viral RNA progressively reduced with increase in amantadine concentration. Moreover, through a growth curve analysis, it was observed that the addition of amantadine inhibited CHIKV infection remarkably at later times in vitro. This strongly implies that 6K is indeed a valuable target for the development of pan-alphaviral inhibitors, using amantadine and its derivatives as the starting point.
Our work also highlights interesting features of 6K localization and functionality, which are essential for understanding the mechanism of CHIKV assembly. Our liposome assays show that CHIKV 6K preferentially disrupts vesicles that mimic the lipid composition of the endoplasmic reticulum (~ 60% disruption) as opposed to those that mimic the lipid profile of plasma membrane (~10% disruption). Likewise, recombinant expression of 6K in mammalian cells results in the protein being localized primarily in ER, highlighting a preference shown by the protein for the ER membrane. Indeed, it appears that enhanced cholesterol composition, as present in the plasma membrane, is a deterrent for 6K; however, existing literature conjectures that 6K forms ion channels in the plasma membrane [16]. In order to rationalize these contradictory phenomena, we hypothesized that 6K can probably traffic to the plasma membrane, only if it has a binding partner. Specifically, previous studies showing that the E2 glycoprotein of other alphaviruses interact with 6K, that any detrimental alteration in 6K hinders effective glycoprotein processing [17, 18, 19, 21], and that E2 is also capable of traversing to the plasma membrane on its own, led us to conjecture that E2 might play a role in conveying 6K to the plasma membrane. To verify this hypothesis, we carried out confocal microscopy studies of EGFP tagged 6K, in presence of E1 and/or E2 glycoproteins; and found that CHIKV 6K strongly colocalizes with E2 as opposed to E1, and that its primary location is altered to the plasma membrane upon co-expression with E2. It is possible that the functionality of 6K in altering ionic homeostasis at the plasma membrane is contingent upon its alliance with E2 during virus budding, and that 6K in presence of E2 utilizes the trans-golgi network from ER to reach the plasma membrane.
Our data shows that CHIKV 6K has the propensity to form oligomers and can integrate with planar lipid membranes to produce ion channels. However, how these oligomeric associations are stabilized can only be answered from high resolution experimental structural data. All-atom molecular dynamics simulation studies indicate that the diameter of channels formed by 6K oligomers is in the range of 1.0–1.10 nm, which is sufficient for the passage of small ions and fluorescent dyes. A 2D classification of GST-6K particles dotting the surface of ER liposomes also indicated heterogeneity in the structure of particles formed on these liposome surfaces. A close analysis of amino acids forming the lumen of the channel identified a sizeable proportion of residues from the hydrophilic N-terminal region of individual monomers. The rest of the polypeptide, which is highly hydrophobic, appeared to provide the necessary structural anchorage required to prevent the collapse of the channel. The validity of this arrangement for channel formation is highlighted by attempted simulations with an oligomeric form of 6K, where the N-terminal regions of individual monomers extended outwards from the channel. In this particular arrangement, the central helical transmembrane segments collapsed on one another, disrupting the formation of the central channel.
Recent developments in the field have indicated that the transframe or TF variant of 6K may also be involved in membrane penetration and may be packaged in virions [7]; while the fate of 6K is to remain associated with cellular membranes. Since the sequence for ribosomal frameshift is inherent in the 6K cDNA, we expect that a minor proportion of recombinant protein produced by us will also have an altered C-terminus. However, this alteration will consist of incorporation of an expression-vector derived stretch of 7 residues, which is entirely different from the actual C-terminal region of TF derived from the CHIKV genome. This minor population of recombinantly generated variant is therefore not expected to have similar functionality as viral genome generated and virus-incorporated TF. In the absence of detailed structural and functional data on 6K and TF, it is at this point impossible to comment on the possibly separate roles played by these components in the life cycle of alphaviruses. Our work highlights the biophysical characteristics and functionalities of CHIKV 6K that are analogous to those displayed by viroporins and attempts to fundamentally characterize its cellular localization and multifaceted membrane interacting abilities. We hope that given the requirement for 6K for the propagation of the important human pathogen CHIKV, efforts will be made to utilize 6K as a drug target to develop therapeutic strategies in the future.
Hydrophobicity plots corresponding to CHIKV 6K sequence, was generated using the ExPASy tool ProtScale (https://web.expasy.org/protscale/), and probable transmembrane domains were identified using DAS [47], TMHMM [48], TOPCONS [49] and PHOBIUS [50].
The cDNA corresponding to CHIKV 6K was obtained from a mammalian expression cassette encoding all structural proteins corresponding to CHIKV strain 37997 (CMV/R CHIKVC-E3-E2-6K-E1) [51]. This cassette was a kind gift from Dr. John Mascola, NIAID (National Institute of Allergy and Infectious Diseases, USA). The cDNA corresponding to 6K was subcloned into the BamHI and XhoI sites of the bacterial expression vector pGEX-6P2 (GE Healthcare), thus generating a construct with an N-terminal GST-tag. For mammalian expression, 6K cDNA was subcloned into the NheI and HindIII sites of pEGFP-N1 (Clontech), resulting in the insertion of an N-terminal EGFP tag. CHIKV glycoproteins E1 and E2 were subcloned into the BamHI/XbaI and EcoRI/XbaI sites, respectively, of pCDNA 3.1(+) (Invitrogen) with N-terminal myc-tag. All constructs were confirmed by sequencing.
Expression of GST-6K was induced in E. coli Rosetta pLysS cells. Cells were grown at 37°C until OD reached 0.8, when the temperature was reduced to 18°C, followed by addition of 1 mM IPTG. After 3 hours of induction, cells were pelleted and the expression of GST-6K was confirmed by western blotting, using an anti-GST antibody (Abcam, USA). The pelleted cells were resuspended in a lysis buffer containing 50 mM Tris pH 7.5, 100 mM NaCl, 1 mM DTT, 10% Glycerol and 1% CHAPS, lysed by sonication, followed by centrifugation to remove cellular debris, and a two-step purification process. In the first step, the soluble fraction was incubated with 200 μl of GST beads (Pierce Glutathione Agarose, Thermo Scientific) for 60 minutes at 4°C, with shaking, to pull down the fusion protein. The bound protein was eluted from beads using 10–20 mM glutathione in lysis buffer. The second step of purification involved size-exclusion chromatography on a Superdex-200 10/300 G/L column, using an ÄKTApurifier 10 (GE Healthcare). A buffer consisting of 50 mM Tris, pH 7.5 and 100 mM NaCl, at a flow rate of 0.5 ml/min was used for SEC elution.
Purified GST-6K was incorporated in the bilayer membrane as described previously [52]. The apparatus for electrophysiology experiments consisted of a polystyrene cuvette (Warner Instruments) with a thin wall separating two aqueous compartments. The polystyrene divider had a circular aperture with a diameter of 150μm. Both compartments were filled with 1 ml buffer containing 10 mM HEPES, pH 7.4, 500 mM KCl and 5 mM MgCl2 and connected to an integrating patch amplifier (Axopatch 200B, Axon Instruments) through a matched pair of Ag/AgCl electrodes. The cis chamber was connected to the head stage (CV-203BU) of the amplifier, while the trans-chamber was held at virtual ground. A solution of DPhPC (Avanti Polar Lipids, USA) and cholesterol (6:1) in n-decane (10 μl) was painted over the aperture to form the membrane. Reconstitution of GST-6K in BLM (Bilayer Membrane) was initiated by adding desired concentration (0.2, 0.85, 1.2, 2.6 ng/μl) of protein in BLM buffer, followed by mixing using a magnetic stirrer. A sudden shift in the membrane current indicated the incorporation of the channel in BLM. For amantadine inhibition experiments, GST-6K was added to a final concentration of 2.6 ng/μl. Amantadine to a final concentration of 1 μM was added to the BLM buffer containing GST-6K only after ensuring that the protein has properly integrated in the membrane. Channel current was recorded using Digidata (1440A, Axon Instruments), Low pass Bessel filter of 2 KHz and the acquisition software CLAMPEX (PCLAMP 10.2, Axon Instruments). The channel current was recorded at fixed applied membrane potentials in the range of -100 to +100mV at a sampling frequency of 10 KHz (temperature between 24–25°C). Open and closed states of the channel (s) were identified as described [52]. Data were analyzed using the software CLAMPFIT (PCLAMP 10.2, Axon Instruments), Origin 5.0 (Originlab Corp. USA) and Matlab.
Generation of liposomes mimicking the membrane composition of the endoplasmic reticulum (ER) and plasma membrane, and liposome disruption assays, was carried out as described [31, 32]. All lipids were procured from Avanti Polar Lipids (Alabaster, AL, USA). For the assay, GST-6K, at a concentration of 0.1 μM to 10 μM, was incubated in the presence of dye-encapsulated liposomes for 30 minutes at 25°C, and end-point fluorescence was monitored at 585 nm. Purified GST was used as a control in all experiments. Experiments were carried out in triplicates on a Perkin Elmer fluorescence spectrophotometer using a quartz cuvette.
DLS analysis was carried out in a Malvern Zetasizer (Nano ZS90). Briefly, freshly generated ER mimicking liposomes were incubated for 30 minutes alone, or in presence of different concentrations of GST-6K, calcium chloride or amantadine. All experiments were repeated thrice.
HEK 293T cells were cultured in 12-well plates in DMEM, supplemented with 5% FBS and 1% Pen-Strep (GIBCO). Transfection was carried out using Lipofectamine 2000 (Life Technologies, USA) according to the manufacturer’s instructions.
Fixing and staining of cells for confocal microscopy were carried out as described [32]. Images were captured on a confocal laser scanning microscope (Leica SP5 Confocal Laser Scanning Microscope) using a 60X oil-immersion objective. Localization of 6K to cellular organelles or with E1/E2 was estimated by calculating the Pearson’s correlation coefficient, which measures the linear correlation between fluorescent channels. Values greater than 0.5 were considered to indicate a high degree of colocalization between fluorescent signals. General processing of images was carried out with the software ImageJ (http://rsb.info.nih.gov/ij/). All experiments were done in triplicates.
For production of VLPs, ~34.2 x 106 adherent HEK 293T cells, maintained as above, were transfected with 32 μg purified plasmid DNA (CMV/R CHIKVC-E3-E2-6K-E1) mixed with 60 μl lipofectamine 2000, and incubated for 72 hours at 37°C, in presence of 5% CO2. Post incubation, media was harvested and subjected to cushioning on 15% sucrose at 1,00,000x g for 2 hours at 4°C. The resultant pellet was resuspended in a buffer consisting of 20 mM Tris, pH 7.5 and 100 mM NaCl.
For negative staining, 4 μl of GST-6K (0.5 μM), mixed with ER liposomes and incubated for 25 minutes, was applied onto glow discharged, 400-mesh carbon-coated copper grids (Agar Scientific Ltd., UK). The sample was incubated for 4 minutes, and the excess drained off on a Whatman filter paper. 3 μl of 2% uranyl acetate was added for 1 minute, followed by washes (3X) with water, and the grid was air-dried for 1 minute. Micrographs were captured at a magnification of 29000x.
For cryo-freezing of CHIKV VLPs, 4.5 μl of purified VLPs (sucrose cushioned@100,000xg), at a concentration of 0.3mg/ml (absorbance at 280 nm), was applied to glow-discharged Quantifoil R2/2 holey carbon grids. Vitrification was carried out using a Vitrobot Mark IV (FEI/Thermofisher), with a blot time of 2.5s, by plunging grids into liquid ethane cooled by a surrounding bath of liquid nitrogen. Grids were transferred to a cryo-holder (Gatan model 626) and visualized on a FEI Tecnai F20 G2 FEG Transmission Electron Microscope, operated at 200 kV. Digital images were recorded on a CCD camera using TIA software under low-dose conditions at a magnification of 50,000x, with a defocus range of -3 to -5 μm.
2D classification of GST-6K was carried out using EMAN 2.2 [53]. Briefly, particles were boxed and extracted from micrographs. A total of 746 particles were selected for generating the 2D classes. All micrographs were CTF corrected and reference-free 2D classification was carried out using standard methods. A total of 6 iterations were done and 16 representative classes were chosen.
Vero cells (African green monkey kidney epithelial cells) and Chikungunya virus (CHIKV, S 27 strain, accession no. AF369024.2) were kindly gifted by Dr. M. M. Parida, DRDE, Gwalior, India. Cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM; PAN Biotech, Germany) supplemented with 5% Fetal bovine serum (FBS; PAN Biotech, USA), Gentamycin, and Penicillin-Streptomycin (Sigma, USA). Amantadine was purchased from Sigma Aldrich (now MERCK).
Vero cells were seeded in 96 well plates (Corning, USA) and after the cells had reached 90% confluence, they were treated with different concentrations of amantadine for 24 hours and incubated at 37°C in 5% CO2. Cellular cytotoxicity assay was performed according to the protocol described earlier [54]. Cellular cytotoxicity was determined in triplicate and each experiment was repeated three times independently.
Vero cells at 90% confluence were grown in 24 well plates in complete DMEM (Pan Biotech, USA). For infection, cells were first washed two times with 1X PBS (Himedia, USA) and thereafter infected with CHIKV, at a Multiplicity of Infection (M.O.I) of 0.1, in serum-free medium. The infected cells were incubated at 37°C in 5% CO2 for 90 minutes with shaking at every 10 minutes interval. After 90 minutes, cells were washed twice with 1X PBS, and incubated with complete DMEM containing different concentrations of amantadine (2.5 μM to 200 μM). Thereafter, cell culture supernatants from CHIKV infected, drug-treated or untreated cells, were harvested at different hpi according to the experiment for viral titer estimation.
Plaque assay was performed according to the procedure mentioned before [55]. Briefly, vero cells seeded onto 6 well plates were infected with different dilutions of the harvested cell culture supernatants mentioned above. After infection, cells were washed twice with 1X PBS and overlaid with semi-solid media (DMEM containing 10% FBS and 2% methylcellulose). The cells were fixed once plaques were visible and countable (4 to 5 days post-infection). For fixing, the cells were first treated with 8% formaldehyde, followed by staining with 8% Crystal violet solution.
Equal volume of samples (Mock, CHIKV infected, and infected as well as amantadine treated) was taken for viral RNA isolation using the QIAamp viral RNA isolation kit (Qiagen, USA) as per the manufacturer’s instructions. RT reaction was performed using the First Strand cDNA synthesis kit (Fermentas, USA) as per the manufacturer instructions. An equal volume of cDNA was used during qRT-PCR for amplifying E1 gene of CHIKV [56].
In the absence of any experimentally determined three-dimensional structure for CHIKV 6K, two different structures predicted by the server Bhageerath-H [57] were utilized for simulation studies. Both structures contain a central alpha-helical segment with two other short helical stretches at the N- and the C-termini. The predicted three-dimensional structures of CHIKV 6K were embedded within a pre-equilibrated 392 POPC lipid bilayer. Peptide molecule(s) were placed perpendicular to the plane of the membrane. The entire system was solvated in water and 100 mM NaCl was added. After equilibration for 100 ns, the system was simulated without position restraints for 700 ns. After 700 ns, an electric field corresponding to 90–110 mV was switched on along the z-axis for facilitating ion conduction through the channel. All simulations were carried out for 1.5 μs, using the GROMACS package v5.1.1 [58]. Specific simulation steps followed has been outlined earlier [31, 59]. Gromacs analysis tools, UCSF Chimera [60] and VMD (Visual Molecular Dynamics) [61] were utilized for data analysis and molecular visualization.
|
10.1371/journal.pgen.1003416 | Differential Association of the Conserved SUMO Ligase Zip3 with Meiotic Double-Strand Break Sites Reveals Regional Variations in the Outcome of Meiotic Recombination | During the first meiotic prophase, programmed DNA double-strand breaks (DSBs) are distributed non randomly at hotspots along chromosomes, to initiate recombination. In all organisms, more DSBs are formed than crossovers (CO), the repair product that creates a physical link between homologs and allows their correct segregation. It is not known whether all DSB hotspots are also CO hotspots or if the CO/DSB ratio varies with the chromosomal location. Here, we investigated the variations in the CO/DSB ratio by mapping genome-wide the binding sites of the Zip3 protein during budding yeast meiosis. We show that Zip3 associates with DSB sites that are engaged in repair by CO, and Zip3 enrichment at DSBs reflects the DSB tendency to be repaired by CO. Moreover, the relative amount of Zip3 per DSB varies with the chromosomal location, and specific chromosomal features are associated with high or low Zip3 per DSB. This work shows that DSB hotspots are not necessarily CO hotspots and suggests that different categories of DSB sites may fulfill different functions.
| For sexual reproduction, meiosis is an essential step ensuring the formation of haploid gametes from diploid precursors of the germline. This reduction in the genome's content is achieved through a specialized type of division, where a single round of DNA replication is followed by two successive rounds of chromosome segregation. The first round separates the homologs. For this to faithfully occur, homologous chromosomes pair with each other and experience recombination, catalyzed by the formation of programmed double-strand breaks (DSBs). Upon their repair, a subset of DSBs will generate crossovers, which result from an intermediate that creates a physical link between homologs and allows their correct segregation by the meiotic spindle. DSBs, as well as crossovers, do not occur randomly along chromosomes but at preferential places called hotspots. To ask if all DSB hotspots also give rise to high crossover frequency, we have systematically compared the map of DSBs with that of a protein, Zip3, which we show preferentially binds to DSB sites that are being repaired with a crossover. We discovered that several DSB hotspots rarely produce crossovers, meaning that the decision to repair a DSB with a crossover can be influenced by specific chromosomal features.
| During meiosis, the programmed formation of DNA double-strand breaks (DSBs) and their repair by homologous recombination ensures that crossovers (CO) occur between homologous chromosomes. COs promote the accurate segregation of homologs at the first meiotic division, thus avoiding aneuploidy, which is a common cause of birth defects and congenital diseases. In all species, two to 30 times more DSBs are formed than COs, indicating that only a subset of all DSBs formed in a cell are repaired through a pathway that will give rise to a CO. The remaining DSBs are repaired by other homologous recombination pathways, such as the synthesis dependent strand annealing (SDSA) mechanism, symmetrical Holliday junction resolution or Holliday junction dissolution [1], that result in non-crossovers (NCOs). In addition, a substantial fraction of meiotic DSBs is also repaired by homologous recombination using the sister chromatid as template, which is not productive for chiasmata and homolog segregation [2]. The repair pathway choice has thus to be tightly controlled to ensure the required number of COs per homolog pair.
DSBs and COs tend to occur more frequently at preferred sites, or hotspots. It is not known whether DSB hotspots are also CO hotspots, or whether DSB repair is modulated by DSB localization on a chromosome. This question could be answered by comparing a high resolution genome-wide map of CO frequencies to the existing high resolution maps of DSBs, for instance in budding yeast (e.g., [3], [4]). Nevertheless, several studies have suggested that the relative contribution of each DSB repair pathway may vary from site to site along the genome. For instance, using a small number of yeast meioses, Mancera et al noted that some sites gave rise to more COs and others to more NCOs per total recombination events [5]. Using a similar approach, Fung and colleagues showed that close to centromeres, COs and NCOs are strongly repressed although DSB activity was reported in these regions, suggesting that DSBs in centromere-proximal chromosomal regions are preferentially repaired by sister chromatid recombination [6], [7]. Analyses of human sperm recombination frequencies revealed that the CO/NCO ratio varied 30 times in the sites under study [8], [9], [10]. Finally, in the fission yeast Schizosaccharomyces pombe, strong discrepancies were found between the DSB map and the CO frequencies [11]. Thus, it is worth investigating if the map of meiotic DSBs truly reflects the map of COs along the genome, and what chromosomal features may influence the choice of DSB repair pathway.
Several factors affect CO formation and their sites of action may reflect how a DSB is repaired. A group of proteins collectively termed “ZMM” is necessary for the formation of about 85% of all COs in budding yeast [12], [13]. During yeast meiosis, the ZMM proteins act by stabilizing the Single End Invasion (SEI) recombination intermediate, which once formed is transformed via capture of the second break end into a double Holliday junction (dHJ) that is mainly resolved as a CO [12], [14], [15]. The ZMM group comprises proteins that act directly on recombination intermediates in vitro, such as the Mer3 helicase, which promotes D loop extension and the Msh4–5 heterodimer, which stabilizes dHJs. This group also includes Zip1, the central element of the synaptonemal complex (SC), as well as Zip2, Zip3, Zip4 and Spo16 that might promote SC formation through Zip1 polymerization between homolog axes [13], [16]. Currently, it is hypothesized that the ZMM proteins, by promoting SC initiation and by directly acting on recombination intermediates, protect the CO-prone recombination intermediates (dHJ) from dissolution by anti-CO proteins, such as Sgs1 [17]. Zip3 has orthologs in C. elegans (ZHP-3) and in mammals (RNF212) and is considered to be a SUMO E3 ligase that sumoylates chromosome axis proteins, thus promoting SC polymerization. Indeed, the Zip3 sequence includes a SUMO Interacting Motif (SIM) and a C3H2C3 Ring-Finger Motif (RFM) that are important for Zip3 in vitro E3 ligase activity and necessary for SC polymerization and correct sporulation [18].
Indirect evidence suggests that ZMMs localize at CO-designated sites, but this has never been demonstrated. ZMMs form foci during meiotic prophase at the time of recombination [16], [19], [20] and the number of Zip3 foci is compatible with CO frequency in wild-type yeast strains [20]. Moreover, in hypomorphic spo11 mutant strains in which the number of DSBs but not of COs is reduced (a phenomenon known as CO homeostasis), the number of Zip3 foci follows the CO variation [21]. Finally, Zip2 foci are non-randomly distributed along chromosomes, like COs [22]. Among the ZMMs, Zip3 seems to be acting earlier because it is required for focus formation of all the other ZMMs [16].
We thus mapped Zip3 binding sites along individual genomic regions and genome-wide during budding yeast meiosis and then determined the features that influence its distribution. We show that Zip3 association with chromosomes is dynamic, occurring first with centromeres, in a DSB-independent manner, then with meiotic chromosome axes upon DSB formation and finally with DSB sites upon joint molecule formation, the preferred intermediate for CO production. These features establish Zip3 as a marker of CO-designated sites. Genome-wide mapping of Zip3 recruitment to DSB sites demonstrates the existence of different types of DSB hotspots based on CO production.
Zip3 localization was previously investigated only by indirect immunofluorescence on chromosome spreads. To investigate Zip3 localization on meiotic chromosomes at about 1-kb resolution, we used chromatin immunoprecipitation (ChIP) and qPCR and yeast strains in which Zip3 was C-terminally tagged at its endogenous locus with three copies of the Flag epitope. Strains expressing the ZIP3-His6-FLAG3 allele showed normal meiotic progression and spore viability (98%, 205 tetrads dissected), showing that the tagged protein is functional. During a meiotic time-course, DSBs monitored at the BUD23 promoter hotspot on chromosome 3 form and reach a maximum at 3–4 hr, before getting repaired (Figure 1A). Zip3 showed a reproducible dynamic localization. It bound first to centromeres from 2 hr after meiosis induction and before DSB formation, then to axis-associated sites and finally to DSB sites, particularly at 4 hr (Figure 1E). At this time, DSB fragments, as detected by Southern blotting, started disappearing (Figure 1A), indicating that DSB ends were already engaged in homologous recombination repair.
As Zip3 might be a SUMO E3 ligase, we investigated whether interaction with SUMO regulated Zip3 binding to the different chromosomal structures. To this aim, we mutated the Zip3 SIM (zip3I96K mutant) or the RFM (zip3H80A mutant) motif. Both mutated proteins were timely induced during meiosis, but they lacked the characteristic lower migrating bands that correspond to sumoylated Zip3 [18] (Figure 1B). In both mutants, early Zip3 binding to centromeres was abolished (Figure 1E), consistent with the previous suggestion that Zip3 recognizes sumoylated proteins at centromeres [18]. Moreover, recruitment to axis-associated and DSB sites was also mostly abolished (Figure 1E) and meiotic progression was impaired in both zip3 mutants (Figure 1D), similarly to what was observed in zip3 null mutants (data not shown). These findings indicate that Zip3 SUMO binding and E3 ligase activities are essential for Zip3 association with chromosomes and all its functions in meiosis. SUMO binding could be directly involved in Zip3 recruitment to all these chromosome locations or indirectly, if required only for the initial Zip3 enrichment at centromeres, and if this is an essential step for the subsequent recruitment of Zip3 to axes and DSB sites.
We then mapped Zip3 binding sites genome-wide using microarrays at 3, 4 and 5 hr during meiotic progression in two independent meiotic time-course experiments (Figure S1A and S2). Genome-wide profiling confirmed the results obtained by ChIP and qPCR (Figure 2A). We then compared the Zip3 maps with the maps of the axis-associated Rec8 cohesin [23] and of Red1, another meiotic axis component that does not show the strong centromere association characteristic of Rec8 [24]. The reference DSB map was the map established by genome-wide mapping of ssDNA in a repair-defective dmc1Δ mutant [3].
At 3 hr after meiotic induction, Zip3 was strongly associated with centromeres, as seen on individual chromosomes (Figure 2A and Figure S3) and in the genome-wide analysis (Figure 2B, Figure S1B and Table 1). All 16 centromeres contained a strong Zip3 peak at less than 1 kb away, and 16% of the 287 Zip3 peaks at this time were found at less than 10 kb from the centromeres. Moreover, 81% of Zip3 peaks at less than 10 kb from a centromere overlapped with an axis-associated Rec8 peak and 38% with a Red1 binding site. At 3 hr, Zip3 was weakly associated with chromosome arms and the Zip3 peaks at more than 10 kb from a centromere coincided with Rec8 (54% peaks) and Red1 (50%) enriched sites (Figure 2B and Figure S3). This is reflected by the overall strong correlation between the Zip3 signal at 3 h and the Rec8 and Red1 profiles (Table 1). At 4 hr, Zip3 association with Rec8 sites diminished (only 35% of its 966 binding sites occurred at Rec8 sites), while its association with DSB sites started to increase (Figure 2B, Figure S4, and Table 1). Concomitantly, the relative Zip3 binding to centromeres decreased (Figure 2B). Finally at 5 hr, Zip3 was almost exclusively associated with DSB sites. Indeed, none of the 557 Zip3 peaks was found at less than 1 kb from centromeres and only 15% of Zip3 peaks coincided with a Rec8 peak at this time (Figure 2B and Table 1).
Thus, during meiosis, Zip3 associates first with centromeres. Centromeric Zip3 enrichment is then progressively reduced, whereas association with axis sites and particularly with DSB sites increases, in agreement with its previously described role in recombination.
To investigate which events triggered these dynamic changes in Zip3 localization we used yeast mutants that affect precise steps of recombination (Figure 3A). Zip3 association with centromeres early in meiosis might occur independently of DSB formation. Indeed, by using the spo11Δ mutant in which DSBs are not formed, we could show that Zip3 associated transiently with centromeres, but not with axis or DSB sites (Figure 3B and 3C: ChIP and qPCR analysis of individual sites; Figure S3 and Table 1: genome-wide analysis). Thus, association of Zip3 with centromeres is independent of DSB formation, whereas DSB formation is required for Zip3 association with the chromosome arms.
Moreover, in the rad50S mutant strain, where Spo11 DSBs are formed but not processed, Zip3 was recruited to centromeres and then chromosome axes, but not to DSB sites (Figure 3B and 3C). In the dmc1Δ mutant that is resection-proficient but deficient in strand invasion, Zip3 was transiently recruited to the axis-associated sites, with kinetics similar to those of wild-type cells, but associated rarely with DSB sites (at least eight times less than in wild-type cells), at the three sites examined (Figure 3B and 3C). Similarly, in the mnd1Δ mutant in which Dmc1 is loaded onto DSB ends but strand invasion does not occur [25], Zip3 was recruited to axes, but not to DSB sites (Figure 3B and 3C). We conclude that DSB formation is sufficient to trigger Zip3 localization at axis sites, whereas strand invasion is required for Zip3 association with DSB sites.
In meiosis, rad52Δ mutants allow strand invasion by Dmc1 filaments, and wild-type levels of the Single End Invasion (SEI) intermediate, a crossover-specific intermediate, but are strongly impaired in the following step, second end capture, which leads to double Holliday junction formation and crossover resolution [26], [27]. In rad52Δ mutants, we detected centromere and axis association delayed but to nearly wild-type levels, but a strongly reduced binding of Zip3 to the three DSB sites (Figure 3B and 3C). This suggests that Zip3 requires the second end capture step, a crossover specific event, for associating with sites of DSB.
Finally, we analyzed Zip3 association with chromosome structures in the ndt80Δ mutant in which dHJs are formed but not resolved [14]. Zip3 recruitment to DSB sites occurred, at levels even higher than in wild-type, suggesting that dHJ formation is the event that triggers or stabilizes Zip3 recruitment to DSB sites (Figure 3B and 3C). In addition, we reproducibly detected a very strong enrichment on the axis, perhaps a consequence of the aberrant turnover of dHJ intermediates in this mutant. Finally, we noticed that Zip3 remained bound with DSB sites longer than in wild-type (Figure 3B).
This mutant analysis reveals that Zip3 associates with DSB sites only when they are engaged in dHJ intermediates, which are the CO precursors. Therefore Zip3 association with DSB sites can be considered as a marker for CO sites.
We next investigated the role of Zip1, which is the central element of the SC and was previously described as not necessary for Zip3 focus formation [16], [20], in Zip3 localization by ChIP and qPCR analysis. In the absence of Zip1, Zip3 was recruited to centromeres, although less than in wild-type cells, and to axis-associated sites, but only rarely to DSB sites (about 10-fold reduction, Figure 3B and 3C). This may be linked to the suggested role of Zip1 in stabilizing the Smt3 chains that are good binding substrates for Zip3 ([18] and Discussion).
Key events of meiosis are regulated by several kinases that are activated at different steps of meiosis. As Zip3 is phosphorylated in a DSB-dependent manner in meiosis ([18] and Figure 4A), we asked whether the dynamic Zip3 localization on chromosomes could be regulated by changes in its phosphorylation status. The CDK kinase Cdc28, together with the Cdc28-associated cyclins Clb5 and Clb6, is necessary for meiotic replication, DSB formation and SC formation [28] and can phosphorylate Zip3 in vitro [29]. In vivo, post-translational modifications of Zip3 are reduced in a clb5 and clb6 mutant [18], suggesting that Zip3 may be a CDK target. We mutated the six S/T-P CDK consensus motifs of Zip3 to A-P motifs (Figure S5) and found that mutant and wild-type Zip3 were similarly recruited and that meiotic divisions and spore viability were unaffected (Figure S5 and data not shown), demonstrating that Zip3 phosphorylation by CDK has no role in normal meiosis.
We next investigated the role of Zip3 phosphorylation by the Tel1/Mec1 kinases, the budding yeast homologs of ATM/ATR. Tel1 and Mec1 are activated upon meiotic DSB formation and play important roles in several key meiotic processes, such as DSB end resection, inter-homolog recombination and regulation of meiotic prophase checkpoint [30]. To this aim, we mutated the four S/T-Q consensus motifs for Tel1/Mec1 to A-Q motifs (zip3-4AQ mutant). This led to a decrease of the low migrating forms of Zip3 due to phosphorylation (Figure 4B). Many of the Mec1-dependent phosphorylated proteins are substrates for the PP4 phosphatase, including histone H2A129 or the Zip1 protein in meiosis [31]. We found that the Zip3 lower migrating forms accumulated in a pph3Δ catalytic subunit PP4 phosphatase mutant, but not in a double zip3-4AQ pph3Δ mutant (Figure 4C). Together, these findings provide strong indication that Zip3 is phosphorylated by the Mec1/Tel1 kinases during meiosis.
We next investigated the meiotic phenotypes of the zip3-4AQ mutant. Meiotic progression, spore viability (97%, 149 tetrads) and kinetics of DSB formation and repair were as in wild-type (Figure 5A and data not shown). At centromeres and axis sites, Zip3-4AQ was normally recruited. However, at the three tested DSB sites, loading of mutant Zip3 was 2- to 3-fold reduced in comparison to wild-type Zip3 (Figure 5B). Thus, the Mec1 consensus phosphorylation sites of Zip3 are important for its localization or stabilization on recombination intermediates. The reduced recruitment of Zip3-4AQ may result in lower CO frequencies. Indeed, in the EST3-FAA3 interval flanking a strong DSB site on chromosome 9, fewer COs were formed in the zip3-4AQ mutant than in the wild-type ZIP3 strain (Figure 5C and 5E). To test whether COs were reduced also at other loci, we performed tetrad analysis in a strain that contains genetic markers on chromosome 3, 7 and 8 to measure the genetic distances in three intervals per chromosome. Genetic distances were significantly reduced in three of the nine intervals tested, demonstrating the effect of the zip3-4AQ mutation on CO frequency (Figure 5D and Table S1). The observation that the genetic distance was reduced at two intervals on chromosome 3 (the smallest chromosome tested) and at none on chromosome 7 (the largest chromosome) suggests that perhaps smaller chromosomes are more affected by the Zip3 mutation (Figure 5D).
The residual association of Zip3-4AQ with DSB sites and the reduced CO frequency were still sufficient to promote full spore viability. We thus investigated whether the Zip3 S/T-Q motifs become essential for spore viability when DSBs are reduced. However, a mutant with reduced DSB levels did not show increased spore lethality when combined with the zip3-4AQ mutant (Figure S6). Finally, we hypothesized that the features of part of the COs in the zip3-4AQ mutant and of COs associated with wild-type Zip3 may be different. We thus measured CO frequency in the mus81Δ strain (wild-type Zip3), in which the alternative CO pathway is inactivated [32], and in the double zip3-4AQ mus81Δ mutants by physical analysis of the EST3-FAA3 DSB site with flanking markers. In our hands and at the hotspot examined, mutation of MUS81 did not affect CO formation in both strains, and CO was even slightly stronger in each case compared to its MUS81 counterpart (Figure 5E). We conclude that mutating Mec1/Tel1 consensus phosphorylation sites of Zip3 decreases its association with DSB sites and reduces CO frequency, and that the remaining CO are not dependent on the MUS81 pathway.
In wild-type meiosis, Zip3 loading was not comparable at all DSB sites (see Figure S4). Specifically, although there was a high correlation between DSB and Zip3 sites at 4 and 5 hr after meiotic induction, Zip3 was enriched at DSB sites to various degrees (Figure S7).
To test whether variations in Zip3 loading at DSBs correlated with changes in recombination frequencies, we chose DSB sites with differential Zip3 binding and flanked them with hemizygous recombination markers (Figure S8) to assess both DSB and CO frequencies. In the wild-type strain, we chose a DSB site with strong Zip3 enrichment (EST3-FAA3) and three sites with relatively lower Zip3 accumulation (ATG2-LAP3, COG7-LEU1 and ISF1-ADH3) (Figure 6A and Figure S7). The introduction of the flanking markers slightly lowered the DSB frequency in the interval (Figure S9) and we thus compared CO and DSB frequency in strains containing the flanking markers (Figure 6A and 6B and Table S2). The CO/DSB ratio varied among the sites and paralleled their relative Zip3 enrichment as measured on the ChIP-chip profiles: the three low-Zip3 DSB sites showed between 2.5 and 5 times less COs per DSB than the EST3-FAA3 DSB site (Figure 6A and 6B).
To investigate whether such differential loading could be observed also in a situation where the DSB profile and number were changed, we compared the genome-wide maps of DSBs and Zip3-Flag binding sites in the set1Δ strain, in which DSBs are reduced and redistributed to new sites [33]. ChIP followed by qPCR indicated that Zip3 localized at DSB sites at 6 h and 7 h after meiotic induction, as expected because DSB formation is delayed by about 2 hours in this strain [33] (Figure S10A). Conversely and like in the wild-type strain, few Zip3 binding sites coincided with Rec8 sites at the 6 and 7 h time-points (Figure S10B and S11). Moreover, like in wild-type cells, Zip3 loading onto DSB sites was variable. For instance, PES4, a strong set1Δ DSB site, was highly enriched in Zip3, whereas ARG3, another strong set1Δ DSB site, was not (Figure S10C). We flanked each of these two sites by hemizygous markers (Figure S8) and measured crossover frequencies. Similarly, like in the wild-type strain, the high-Zip3 PES4 site showed 2.2 times more COs per DSB than the low-Zip3 ARG3 site (Figure S10C).
These results are consistent with a positive effect of Zip3 loading on DSB repair by CO and shows that in the genome, there are DSB sites that are less bound by Zip3 and less frequently repaired by CO than the average.
We then asked whether specific chromosome features were associated with these variations in Zip3 binding at DSB sites. We first investigated Zip3 loading at DSB sites close to centromeres as it was reported that inter-homolog CO frequency is usually low close to centromeres, although DSBs can form close to centromeres [7]. On several chromosomes, Zip3 did not bind to centromere-proximal DSBs (Figure S12) and, on average, the relative Zip3 signal at DSB sites close (less than 10 kb) to centromeres was significantly lower than in the rest of the genome (Figure 7A).
To extend the analysis beyond centromere regions, we defined from our mapping data two categories of DSB sites. Among the 400 strongest DSB sites previously determined in the resection-proficient dmc1Δ strain (without DSBs at less than 10 kb from a centromere), we identified “low-Zip3” DSB sites (n = 166 sites) and “high-Zip3” DSB sites (n = 142 sites) (see Protocol S1 for details on the classification). In these two DSB populations, the mean DSB signal was not statistically different (Wilcoxon test, p = 0.13). Similarly, several chromosome features, such as distance from a telomere or a centromere, and replication timing, were also not different (not shown). However, the strength of DSB signal measured in the resection-defective rad50S mutant was lower at low-Zip3 DSB sites than at high-Zip3 DSB sites [3] (Figure 6C and Figure 7B). Analysis of DSB formation by Southern blotting at the three low-Zip3 DSB sites ATG2-LAP3, ISF1-ADH3 and COG7-LEU1 (Figure 6D and 6E) and the low-Zip3 set1Δ DSB site ARG3 (Figure S10D) confirmed that at these sites fewer DSBs were detected in the rad50S than in the dmc1Δ background. By contrast, the high-Zip3 EST3-FAA3 and the high-Zip3 set1Δ PES4 DSB sites showed similar DSB frequency in both backgrounds (Figure 6D and 6E and Figure S10D).
When we classified the DSBs in the rad50S mutant as high (157 sites) or low (113 sites), based on the peak signal intensity like we did for the Zip3 peaks, we found that the Zip3 signal was significantly lower in low-rad50S DSBs (Figure 7C). Overall, 66 DSB sites were present both in the low-Zip3 DSB and the low-rad50S DSB category, that is more than expected by chance (p<0.01, Pearson's Chi-square test). This further strengthens our observation that at least a subset of low-Zip3 DSB sites also shows reduced DSB formation in the rad50S mutant, suggesting that they have distinct properties.
The second chromosomal feature that varied between high- and low-Zip3 DSB sites was the distance from an axis-associated site, defined as a Red1 peak (Figure 7B). Low-Zip3 DSB sites were significantly more distant from an axis site than high-Zip3 DSB sites (median distance from a Red1 peak: 5599 bp and 3660 bp, respectively). Conversely, no difference in the distance from an axis-associated site was observed between low and high rad50S DSB sites (Figure 7C). Furthermore, the low-Zip3 DSB sites that were NOT low rad50S DSBs were still much further away from an axis site than the high-Zip3 DSB sites (5709 bp and 3660 bp from a Red1 peak, Figure S13).
We confirmed this observation in the set1Δ strain, in which the 200 strongest set1Δ DSB sites were classified as high- and low-Zip3 DSBs. High-Zip3 and low-Zip3 DSB sites did not show significant differences in their mean dmc1Δ DSB ChIP-chip signal (p = 0.66), but the low-Zip3 DSBs were significantly further away from a set1Δ Rec8 peak or a Red1 peak than the high-Zip3 DSB sites (Figure S10E).
Thus, we can distinguish two different categories of low-Zip3 DSB sites: sites with reduced DSB formation in the rad50S mutant and sites that are far from an axis-associated site, suggesting that proximity to an axis site favors DSB binding by Zip3 and resolution as a CO (see Discussion).
Here we show that the ZMM protein Zip3 interacts dynamically with chromosomes, associating first with centromeres, then with chromosome axes upon DSB formation, and finally with DSB sites on the recombination intermediates engaged in CO formation. We thus propose that Zip3 is a molecular marker of CO sites. We then demonstrate that Zip3 association with chromosomes requires its SUMO E3 ligase motifs, thus implying that SUMO recognition and transfer are needed for Zip3 interaction with chromosomal proteins. Zip3 phosphorylation sites by Mec1/Tel1 kinase are also important for Zip3 full loading on DSBs and CO formation. Finally, we show the existence of DSB sites that are rarely bound by Zip3 and that produce fewer COs than the average of DSB hotspots. These low-Zip3 DSB sites are sensitive to the effect of the rad50S mutation and tend to be away from an axis-association site, where the recombination process takes place.
A recent study showed that the proteins necessary for DSB formation reside on the chromosome axis, rather than at the sites of DSB formation in loop sequences [24]. This suggests that at the time of DSB formation, DSB hotspot sequences are already located on the chromosome axes. Indeed, using ChIP assays, we found that Zip3 first associates with axes and DSB sites, and later during the recombination process (when dHJs are formed at the pachytene stage) it becomes almost exclusively associated with DSB sites. We propose that at this stage, the recombination intermediates are located in the inter-homolog space and are detached from the axis, as previously seen cytologically in Sordaria [34]. Although recombination takes place close to the axis, axis-associated sites might be less immunoprecipitated by ChIP, because Zip3 is less intimately linked to these sites than to DSB sites.
Our ChIP analysis of Zip3 localization in yeast mutants that affect defined steps of recombination indicates that DSB formation is sufficient to trigger Zip3 localization at axis sites, whereas Zip3 associates with DSB sites only when they are engaged in dHJ intermediates. Our results are in apparent contrast with previous cytological findings about Zip3 foci in various mutants. In the rad50S mutant, many Zip3 foci co-localized with Mre11, which associates with DSB sites in this strain [20], [35]. However, we found that Zip3 does not associate with DSB sites in this mutant. The previously described foci could correspond to Zip3 loading on chromosome axes where Mre11-enriched DSB sites may also be located in the rad50S mutant. Similarly, Zip3 foci were previously detected in the dmc1Δ mutant [36], whereas in our study Zip3 was normally associated with axis sites, but very little with DSB sites. This was not due to experimental artifacts due to a differential ability to immunoprecipitate Zip3 in these mutants, since we observed constant Zip3 recovery during the whole time-courses after immunoprecipitation (data not shown). These discrepancies underscore the complementarity between ChIP approaches and cytological studies and show that similar patterns of foci can underlie completely different protein localizations along chromosomes, as revealed by our study.
The early Zip3 association with axes following DSB formation could be due to Zip3 binding to cleaved DSB sites that are located on the axis, or to a generalized recruitment of Zip3 on chromosome axes, maybe through interaction with a protein phosphorylated upon DSB formation. Our ChIP-chip data favor the second explanation because axis sites close to strong DSB sites were not more enriched in Zip3 and Zip3 binding to axes was rather homogenous along chromosomes (data not shown). The protein responsible for Zip3 loading onto axis sites could be an axis protein that is phosphorylated by the Tel1/Mec1 kinases, such as Hop1 [37].
We observed a reduced recruitment of Zip3 to all chromosomal regions in the zip1Δ mutant. It was proposed that at centromeres, Zip1 stabilizes Smt3 chains, made by other SUMO ligases acting in early meiosis, thus favoring Zip3 binding to centromeres. Our data confirm previous cytological observations [38] and suggest that Zip3 loading at centromeres may be a consequence of Zip1 localization at centromeres early in meiosis. Indeed, Zip1 association with centromeres is Zip3-independent and early centromere coupling mediated by Zip1 does not require Zip3 [39]. Our results in the zip3 SUMO ligase and the zip1Δ mutants are consistent with a previously proposed model [18]: after the initial Zip3 recruitment to DSBs, which requires its SUMO binding motif (our results), Zip1 binds to and stabilizes the Smt3 chains deposited by Zip3. This in turn induces a second wave of Zip3 recruitment to DSB sites via its SUMO binding motif [18]. Indeed, in the zip1Δ mutant, Zip3 association with DSB sites was strongly decreased.
Interestingly, Zip3 foci persisted more on DSB sites in the ndt80Δ mutant than in the wild-type. The ndt80Δ mutant accumulates non-cleaved dHJs and thus our data are consistent with the proposed role of Zip3 and the ZMM in general to stabilize the crossover-designated intermediates from D-loop dismantling and later from dHJ dissolution by activities exerted by anti-crossover factors such as Sgs1 [40]. Strikingly, Zip3 association with the axis site reached very high levels in ndt80Δ cells. This may be due to a change of structure within the synaptonemal complex that persists in this mutant and that alters the association of sites undergoing dHJ with axis-associated sites, and renders these closer to strong DSB sites and thus more closely associated with Zip3. It would be interesting to determine if this increase of Zip3 association is seen for all axis-associated sites or only those that are close to strong DSB sites.
We detected a negative influence of the centromere on the relative binding of Zip3 to DSB sites. However, Zip3 binding was not abolished, although these regions show few CO and NCO events and have been suggested to repair their DSBs mostly using the sister chromatid [7]. A previous study showed that during DSB repair by sister chromatid recombination, the formation of associated joint molecules still depends on the ZMM protein Msh4 [2]. Similarly, we found that when a DSB is forced to be repaired using the sister chromatid, it still binds to Zip3, albeit to a lesser extent than when it is repaired by the homolog (unpublished results). Thus, DSBs might bind to Zip3 also very close to centromeres if they form joint molecules with the sister chromatid, explaining why we see residual Zip3 association with DSB in these regions.
In the rest of the genome, we detected qualitative differences among DSB sites. Specifically, for a chosen set of sites, we show that the CO frequency per DSB can vary from one DSB site to another and that this behavior can be predicted based on the relative Zip3 enrichment at the site. These DSB hotspots have peculiar properties: they form DSBs at a lower frequency in the rad50S mutant (our results and [3]) and they tend to overlap with coding regions (our results and [4]). Previous studies showed that in an artificially late replicating chromosomal region, meiotic DSBs also formed later. Interestingly, DSB formation at these sites is affected in rad50S mutants [41]. In the rad50S mutant DSB formation is impaired at many regions [3] and by extension these could be naturally late-occurring DSBs. Indeed, these “low-rad50S” DSBs tend to occur later, but the asynchrony of meiotic time-courses makes it difficult to reproducibly detect a delay ([3] and data not shown). Based on these data, we can hypothesize that the low-Zip3 DSBs that we have studied are naturally late-forming DSBs. This would imply that in a given chromosomal region, early-forming DSBs are the preferred substrate for CO designation. Indeed, CO designation is a very early event, much earlier than Zip3 association, which we defined as a CO marker in this study. Upon early DSB formation, the CO designation of one DSB might relieve the chromosomes from the experienced stress, thus locally disfavoring further CO designations and explaining CO interference [12], [42]. Thus, a DSB formed later in this region will have little chance to be chosen as a CO event. We also found that besides the rad50S effect on DSB frequency and the possibly associated differential timing of DSB formation, low-Zip3 DSBs are more distant from an axis-associated site. For their repair, and likely also for their formation, DSB sites interact with the chromosome axis, particularly where the Red1 and Hop1 proteins reside, and cytological studies showed that the association between Zip3 and Hop1/Red1 occurs prior to SC polymerization, likely at the future CO sites [36]. We propose that a DSB site away from the axis will be less efficiently brought or kept on the axis, making it less favorable for CO designation.
Our data have important implications for the control of meiotic recombination and genetic distances at the level of DSB formation and repair outcome. It will be interesting to investigate whether the DSB sites with low CO frequency we identified are NCO hotspots being repaired via the homolog or if they are repaired via the sister chromatid and whether they are preferential binding sites for anti-crossover activities. These extra-DSB sites rarely repaired as crossover may be either used early for homolog pairing, which precedes crossover formation, or conversely, they may be later “safety” DSB made in case insufficient early DSB go into crossover (Figure 7D).
Our work paves the way for further studies in other organisms, especially in mammals where the number of DSB largely exceeds that of COs.
All yeast strains (Table S4) are derivative of the SK1 background. They were produced by direct transformation or crossing to obtain the desired genotype. Details of strain construction are in Protocol S1. All transformants were confirmed to have the flanking marker at the correct locus by PCR analysis to discriminate between correct and incorrect integrations.
Synchronous meiosis in liquid culture was performed as described [43]. Progression through meiosis was monitored by scoring nuclear divisions after DAPI staining.
Western blotting was performed as described [23] using the mouse monoclonal anti-FLAG antibody M2 (Sigma, 1∶1000), except for detecting phosphorylated Zip3 (Figure 4 and Figure S6) where samples were separated in 10% 150∶1 acrylamide-to-bisacrylamide gels. Dephosphorylation assays were carried out as described [18], using calf intestinal alkaline phosphatase in the presence or not of 20 mM of the phosphatase inhibitor sodium orthovanadate.
For genetic distances on chromosomes III, VII and VIII, haploids were mated at 30°C on YPD supplemented with 1% Adenine for 5 hr before replica-plating on solid sporulation medium (1% potassium acetate) and incubated at 30°C for at least 48 hr. For recombination between hemizygous drug resistance markers, diploids were grown on YPD plates and then replica plated on sporulation medium at 30°C for at least 48 hr. Asci were dissected on YPD supplemented with 1% Adenine and replica-printed to the appropriate media to check for marker status. P (parental), NPD (non-parental) and T (tetratype) were scored to calculate the genetic distances as described in [44]. For calculation of the map distance, standard error calculations were performed using the Stahl Lab Online tools (http://www.molbio.uoregon.edu/~fstahl/). For calculation of the ratio between CO and DSB per cell, we divided the % of cells that received a CO (genetic distance * 2) by the % of cells that received a DSB (%measured DSB frequency * number of chromatids per cell, i.e. 4).
Genomic DNA was prepared, analyzed and DSB or CO frequency was determined as described [23]. The used restriction enzymes and probes are in Protocol S1.
For each time-point, cells were processed and ChIP was performed as described [23], using 2 µg of the mouse monoclonal anti-FLAG antibody M2 (Sigma) and 30 µL Protein G magnetic beads (New England Biolabs). Quantitative PCR was performed using immunoprecipitated DNA or whole-cell extracts as described [23]. DSB1, DSB2 and DSB3 sites were chosen according to the genome-wide mapping of [3]. DSB1 is in the promoter of BUD23 on chromosome 3; DSB2 is in the promoter of ECM3 on chromosome 15 and DSB3 in the promoter of RIM15 on chromosome 6. Axis site was chosen from the Rec8 binding data of [45], on chromosome 3. Negative control site is neither a DSB site nor a Rec8 site, and is located in the promoter of CDC39, on chromosome 3. Primer positions are in Protocol S1. All time-courses and ChIP assays were repeated at least twice from independent experiments and gave similar results.
Immunoprecipitated DNA and whole-cell DNA were amplified, labeled and hybridized to Agilent 44 k yeast whole genome oligonucleotide arrays as described [33]. Microarray images were read using an Axon 4000B scanner and analyzed using the GenePix Pro 6.0 software (Axon Instruments). Files were converted to text files and analyzed using the R software. The signal intensities of profiles were normalized, by dividing all values by the mean of the lowest 10% ratio probes of the array (decile normalization, as described [24]). In this way, the 10% lowest values fall below 1, so that everything below and around this value can be interpreted as background. The resulting normalized data were next denoised and smoothed, as described before [23]. Raw data from [33], [3] and [24] were reanalyzed as described before [23]. Peaks were identified after denoising and smoothing with a 2 kb window (except for the data by [24], where a 300 bp window was used), and compared as described [23]. In the set1Δ Zip3-Flag 6 and 7 hr ChIP-chip assays, a very high signal was obtained, and we adjusted the threshold to 5 to obtain a number of Zip3 peaks comparable to that of the other experiments. High Zip3 DSB sites were DSB sites that coincide with a Zip3 peak the signal intensity of which differed by less than 50 ranks from that of the DSB site; Low Zip3 DSB sites were DSB sites either not bound by Zip3 or that coincide with a Zip3 peak the signal intensity of which was at least 100 ranks lower than that of the DSB site.
For the chromosome coordinates, we used the Saccharomyces Genome Database features (http://downloads.yeastgenome.org/curation/chromosomal_feature/) of the last update from July of 2010.
The ChIPchip data generated in this study have been deposited at the Gene Expression Omnibus database, accession number GSE40563. Processed data for all chromosomes are provided in Table S3.
|
10.1371/journal.pntd.0006277 | Validation of rK39 immunochromatographic test and direct agglutination test for the diagnosis of Mediterranean visceral leishmaniasis in Spain | Visceral leishmaniasis (VL), the most severe form of leishmaniasis, is endemic in Europe with Mediterranean countries reporting endemic status alongside a worrying northward spread. Serological diagnosis, including immunochromatographic test based on the recombinant antigen rK39 (rK39-ICT) and a direct agglutination test (DAT) based on the whole parasite antigen, have been validated in regions with high VL burden, such as eastern Africa and the Indian subcontinent. To date, no studies using a large set of patients have performed an assessment of both methods within Europe.
We selected a range of clinical serum samples from patients with confirmed VL (including HIV co-infection), Chagas disease, malaria, other parasitic infections and negative samples (n = 743; years 2009–2015) to test the performance of rK39-ICT rapid test (Kalazar Detect Rapid Test; InBios International, Inc., USA) and DAT (ITM-DAT/VLG; Institute of Tropical Medicine Antwerp, Belgium). An in-house immunofluorescence antibody test (IFAT), was included for comparison. Estimated sensitivities for rK39-ICT and DAT in HIV-negative VL patients were 83.1% [75.1–91.2] and 84.2% [76.3–92.1], respectively. Sensitivity was reduced to 67.3% [52.7–82.0] for rK39 and increased to 91.3% [82.1–100.0] for DAT in HIV/VL co-infected patients. The in-house IFAT was more sensitive in HIV-negative VL patients, 84.2% [76.3–92.1] than in HIV/VL patients, 79.4% [73.3–96.2]. DAT gave 32 false positives in sera from HIV-negative VL suspects, compared to 0 and 2 for rK39 and IFAT, respectively, but correctly detected more HIV/VL patients (42/46) than rK39 (31/46) and IFAT (39/46).
Though rK39-ICT and DAT exhibited acceptable sensitivity and specificity a combination with other tests is required for highly sensitive diagnosis of VL cases in Spain. Important variation in the performance of the tests were seen in patients co-infected with HIV or with other parasitic infections. This study can help inform the choice of serological test to be used when screening or diagnosing VL in a European Mediterranean setting.
| Visceral leishmaniasis is the most severe form of leishmaniasis, a disease transmitted through the bite of an infected sandfly. Although the biggest burden of leishmaniasis is in eastern Africa and the Indian subcontinent, the disease is also endemic in parts of Europe. Previous studies have looked at performance of diagnostic methods, but not in great detail on samples derived from a European setting. Using a large set of samples from a national reference laboratory in Madrid, Spain, we assessed a leishmaniasis rapid test and a direct agglutination test for serological diagnosis of visceral leishmaniasis in Europe. Both tests were effective at diagnosing VL, but important differences were seen when testing patients co-infected with HIV or with other parasitic infections. This study can help inform which diagnostic tests are suitable for use in a European Mediterranean setting.
| Visceral leishmaniasis (VL) is a life-threatening disease caused by protozoan parasites of the Leishmania donovani complex. It is widely endemic in South America, eastern Africa and Asia as well as in the Mediterranean basin [1]. More than 500 million people are at risk of acquiring leishmaniasis worldwide, with approximately 90% of the cases arising in rural areas of Bangladesh, Brazil, Ethiopia, India, Somalia, South Sudan and Sudan [2]. In Europe, nine countries report cases of VL annually accounting for less than 2% of the global burden [3], where cases are mostly confined to the Mediterranean countries, but a spread towards northern Europe is being reported as a result of a range of factors, including vector and parasite migration, and changes to the environment and climate [4]. In Spain, a VL outbreak of unprecedented magnitude occurred in the southwest of the capital Madrid between 2009–2013 [5,6], and the country was recently listed among the top 14 VL high-burden country [2].
Facing a possible (re-)emergence of leishmaniasis in Europe, it is important for national public health institutions to have established guidelines for clinical diagnosis of VL to support primary health care and epidemiological surveillance [3,7]. Parasitological confirmation through culturing and/or microscopy remains the gold standard for diagnosis, and gives the clearest indication of parasitic infection. The sensitivity of parasitological confirmation, however, depends on the sample used, where spleen and bone marrow aspirates yield the best results but these are obtained through invasive sampling procedures, with inherent complications, besides presenting variable sensitivity [8]. In addition, the absence of parasites in tissue sample does not necessarily indicate absence of infection. Nucleic acid amplification tools have shown to be more sensitive than microscopy or culture for VL diagnosis, even when using peripheral blood samples [9]. This technology is already available in many hospitals and reference centers in VL-endemic countries in Europe; unfortunately there is a consistent lack of standardization and a very high number of different protocols [9].
Serological tools provide a good diagnostic accuracy as long as they are used in combination with a standardized clinical case definition for VL [1]. Serological tests vary in the target antigen (whole parasite or recombinant protein), ease-of-use (rapid dipstick or necessity for some laboratory infrastructure), sensitivity, specificity, and cost. Underlying HIV infections, or other forms of immunosuppression, however, can affect their sensitivity [10].
The rK39 immunochromatographic test (rK39-ICT) and the direct agglutination test (DAT) have been widely validated in the VL endemic regions of eastern Africa and the Indian subcontinent, with rK39-ICT demonstrating varying sensitivity and specificity depending on the geographical setting [11–13].To our knowledge, no studies using a large set of patients have performed an assessment of both methods on human samples within Europe. To establish evidence on serological VL diagnostic performance in this region, we assessed the sensitivity and specificity of rK39-ICT, DAT and IFAT using historical serum samples collected in Spain from 2009–2015.
The serum samples used in this study are anonymized and are part of a registered collection, as described below. No ethical approval was required.
The study was conducted at the WHO Collaborating Centre for Leishmaniasis, National Centre for Microbiology, Instituto de Salud Carlos III, Madrid, Spain (WHOCCL-ISCIII), which is also the national reference laboratory for leishmaniasis.
We used historical serum samples stored at -70°C at the WHOCCL-ISCIII. These samples are part of a collection registered at the National Biobank Register-Section Collections, Spain, with collection Reference ID: C.0000898. The serum samples in the collection are anonymized.
Samples from suspected VL cases are derived from patients with clinical suspicion of VL as defined in the protocol of the Spanish national network for epidemiological surveillance [14], and were referred from different hospitals to the WHOCCL-ISCIII for diagnosis from 2009–2015. Briefly, a suspected VL case in Spain is defined as a patient who presents with irregular prolonged fever plus splenomegaly and/or weight loss, which may be accompanied by hepatomegaly, lymphadenopathy, leukopenia, anemia and thrombocytopenia. Each suspected cases had multiple samples (whole blood, serum, bone marrow) taken to facilitate diagnosis. While in this study only serum samples were tested, we used all laboratory and clinical results available from each patient to classify them as “case” or “non-case”, and therefore define the reference diagnostic result (see parasitological confirmation below). Samples from VL suspects were further divided according to the HIV status of the patients. In addition, we chose samples from patients who were diagnosed with malaria, Chagas disease or other parasitic infections, as well as serum from healthy individuals (blood donors) from Spain, Belgium and Germany. All samples were anonymized and diagnostic test operators were blinded to the nature of the serum sample.
The rK39-ICT (Kalazar Detect Rapid Test, Inbios International Inc., WA, USA), and the DAT with freeze-dried antigen (ITM-DAT/VL; Institute of Tropical Medicine, Antwerp, Belgium) were performed according to manufacturer’s instruction; with 20 and 1 μl serum respectively. DAT was performed by using the screening method, samples with a titer ≥ 1:3200 were considered positive [15]. An in house IFAT was performed by following a standard method [16]: the antigen was prepared from promastigotes of the L. infantum international reference strain MHOM/FR/78/LEM-75. Antibody binding was detected using fluorescein isothiocyanate-conjugated sheep anti-human immunoglobulin G (heavy and light chains). One μl serum was used. The threshold titer for positivity was ≥1/80. Test results were interpreted and recorded on a standardized form by at least two observers at the minimum reading times, where each observer was blinded to the other’s reading. Any test returning an invalid result or lack of agreement between observers was repeated.
As part of routine diagnosis VL suspect patients are tested at the WHOCCL-ISCIII by nested PCR of blood and bone marrow samples, bone marrow Giemsa microscopy and blood and bone marrow NNN culture, following procedures described elsewhere [17,18]. A serum sample from a VL suspect was defined as pertaining to a case when there was parasitological confirmation of Leishmania in blood and/or bone marrow aspirate in samples taken within 21 days before or after the serum sample was taken.
The statistical software R was used with the ‘epiR’ package to determine sensitivity, specificity, positive and negative predictive values [19,20]. Exact binomial confidence limits were calculated for test sensitivity, specificity, and positive and negative predictive values.
STARD checklist and workflow are provided as supplementary materials, S1 Table and S1 Fig respectively.
A total of 743 samples from 2009–2015 were tested, of which 405 were suspected VL cases, and 338 samples as control group. Most samples were taken from March 2013 to January 2015 (Fig 1A) in hospitals from different regions in Spain, mostly from Madrid and the Mediterranean coast (Fig 1C).
Seventy percent of suspected cases were male, with an average age of 41 years (Fig 1B). Seventy-six patients were HIV positive, while 12 patients had immunosuppression related to organ transplantation (n = 11) and Crohn’s disease (n = 1). The composition of serum samples and the diagnostic test results is detailed in Table 1.
The sensitivity, specificity, positive and negative predictive values of each test are given in Table 2.
A sub group analysis according to the HIV status of the VL suspect patients is shown in Table 3.
The estimated sensitivity of rK39 was 78.0% [70.8–85.2] for all 405 suspected VL patients. Of the 95 HIV-negative VL cases, 79 were correctly diagnosed by rK39 giving a sensitivity of 83.1% [75.1–91.2]. The sensitivity dropped to 67.3 [52.7–82.0] in individuals with underlying HIV infection. Of the 602 negative samples (338 control subjects and 264 non-confirmed VL suspects), rK39 gave 1 false positive in a malaria patient.
The estimated sensitivity of DAT was 86.5% [80.5–92.5] for all 405 suspected VL patients. The sensitivity of DAT in HIV-negative VL suspects was 84.2% [76.3–92.1], rising to 91.3% [82.1–100.0] in individuals with underlying HIV infection. Of the 602 negative samples (338 control subjects and 264 non-confirmed VL suspects), DAT gave false positive results for 41 serum samples, including 2 individuals with malaria and two individuals with other parasitic infections.
The estimated sensitivity of IFAT was 79.4% [72.4–86.4] for all 405 suspected VL patients. The sensitivity of IFAT in HIV-negative VL suspects was 84.2% [76.3–92.1] dropping to 79.4% [73.3–96.2] in VL suspects with HIV. Of the 602 negative samples (338 control subjects and 264 non-confirmed VL suspects), IFAT gave false positives for 2 non-confirmed VL suspects and 15 patients with Chagas disease.
In this study, we assessed the sensitivity and specificity of rK39-ICT, DAT and IFAT on a varied set of historical serum samples collected in Spain from 2009–2015. The diagnostic performance of rK39-ICT and DAT has been largely evaluated in highly endemic country settings [12], with variable results in different geographic locations for rK39-ICT [21], and the performance in a European setting remains largely unknown. A multicenter study compared different diagnostic tests using samples from 26 HIV-negative and 11 HIV-positive VL patients from southern France [22]. This study evaluated a different rK39-ICT (IT-LEISH Bio Rad Laboratories, France) and DAT (the same as in our study), among other serological tests, and obtained a sensitivity of 88.5% for both in HIV-negative VL patients and 54.5% for DAT and 81.8% for rK39-ICT in HIV/VL patients. In a study from Italy with a sample size of 94 patients with suspected VL (21 patients were confirmed VL cases), the reported sensitivity of rK39-ICT was 52.4%, using a different manufacturer than in our study [23]. These results differed from our study, which are more in agreement with other large scale evaluations of rK39-ICT and DAT that show higher sensitivity for DAT in a setting with high HIV-co-infection rate [24].
The evaluation of diagnostic tools for visceral leishmaniasis in Europe is important as the burden of VL remains an issue for European public health officials [3]. We began addressing this issue using a large assembly of samples. Our results show lower sensitivity estimates when compared to published results on serological assay evaluation in South East Asia, the Americas and eastern Africa regions. In a meta-analysis of diagnostic performance, the combined sensitivity estimates of DAT and the rK39-ICT were 94.8% and 93.9%, respectively [25], while in our study, the estimated DAT and rK39-ICT sensitivity was at 86.5% and 78.0%, respectively. In a WHO led evaluation of rK39-ICT, the sensitivity estimates varied greatly from region to region: 67.6% in eastern Africa, 84.7% in Brazil and 99.6% on the Indian subcontinent [21].
The different performance of serological tests between European samples and those tested elsewhere is most likely due to the epidemiological landscape. Patients residing outside of Europe will have different anti-VL immunoglobulin titers, different age patterns of infection, immune and/or nutritional background and/or are exposed to higher parasite diversity [21]. A study analyzing L. donovani strains from African and Asian origin revealed extensive genetic diversity in coding sequences of rK39 homologues, which may provide an explanation for the different performance of rK39-ICT across regions [26]. In the Mediterranean region VL is caused by different genetic variants of L. infantum [27,28], whether this has an effect on the performance of rK39-ICT would be an interesting subject of study.
Our study sample reflects a population living in a southern European member state where samples are routinely submitted for laboratory diagnosis after clinical suspicion of VL. Of the 405 samples submitted between 2009 and 2015 and analyzed in this study, 34% were classified as cases. To expand our sample population and assess the performance of serological tests with respect to cross-reactivity, we further selected confirmed VL negative serum samples with varying immunological exposures, including Chagas disease, caused by another member of the Trypansomatidae family, as well as German and Belgian blood donors who are less likely to have had previous parasite exposure.
Although sensitivity and specificity varied between diagnostic tests, we found that the in house IFAT, rK39-ICT (Kalazar Detect from InBios International, Inc.) and DAT (ITM-DAT/VL, Institute of Tropical Medicine, Antwerp) are valid for VL diagnosis in Europe. The choice of test, however, is according to the epidemiological context and intended application. In the context of VL in Europe we find three main applications for serological tests: seroprevalence studies, clinical diagnosis and outbreak response tools.
Seroprevalence studies involve large-scale screening of samples to determine the burden of disease in a given population. Previous prevalence studies in Europe on blood donors have used different immunological and/or molecular tests [29–31]. Based on our results, we find DAT performed best for seroprevalence studies, with the ability to batch process samples, the acceptable costs, and the specificity and sensitivity values at 86% and 85%, respectively.
For clinical diagnosis, the choice of test depends partly on the immunological status of the patient. In our study, co-infection with HIV reduced the sensitivity of rK39-ICT, making it less applicable for a point-of-care test in HIV individuals in Spain. During the community outbreak of VL in Madrid, 16 out of 160 reported VL cases (10%) had HIV [5], and could therefore have been missed if rK39 was used as sole diagnostic. In a series of 73 VL patients (66% immunocompetent) from that outbreak, another rK39-ICT (SD BIOLINE Leishmania Ab, Standard Diagnostics, Inc., South Korea) showed 67% sensitivity and 100% positive predictive value [32]. To our surprise, in our study the DAT showed higher sensitivity in HIV-positive patients. Although a higher sensitivity in this group is somehow unexpected, it is important to highlight that DAT has returned acceptable sensitivity in the diagnosis of VL in HIV-positive patients, being superior to other serological tests [10,24,33–35]. It is difficult for us to find an explanation to this, and it could be suggested that the observed discrepancy may be due to the difference in the number of patients in each group; being only relevant for rK39-ICT (the only test using a single antigen), for which the different performance according to the HIV status is especially marked. A study specifically designed to assess differences in the diagnostic performance of the tests according to the HIV status would be necessary to address this properly. We did not conduct a separate analysis in patients with other immunosuppressive conditions as previous studies have shown that the diagnostic sensitivity of serological tests is not decreased in patients receiving solid organ transplants, which were 11 out of 12 of our suspected VL cases with immunosuppressive conditions other than HIV [36–38].
During outbreak settings such as those seen in Madrid in 2009–2013, point-of-care tests like rK39-RDT have the benefit of portability, simplicity and the speed of result, allowing quick identification and control of infection clusters. In our study we used all laboratory (PCR, culture, serology) and clinical results available from each patient to classify them as “case” or “non-case”. The reported sensitivities of the serological tests included in this study justify the algorithm proposed for VL diagnosis in the WHO European region, where rK39-ICT is first used in VL suspected cases and can be complemented with other serological or parasitological tests to ensure accurate diagnosis [3]. The rK39-ICT is a simple, fast, commercially available test that uses a less invasive sample. The application of this test for VL diagnosis and subsequent treatment of confirmed cases with liposomal amphotericin B, the reference treatment for VL in the WHO European Region [3], has shown to be cost-effective for Mediterranean VL management in Morocco [39].
In terms of cross reactivity, we found that all true negative serum samples from blood donors from Belgium, Germany and Spain were diagnosed as negative for VL by DAT and rK39-ICT. Some false positive results were obtained with Chagas disease patients and those with other parasitic infections, this was particularly pronounced for the IFAT, a widely used serological test for VL diagnosis in Europe. This can be explained by serological cross reactivity between trypanosomatids [40]. In order to account for this, other infections such as Chagas disease, malaria or other parasites should be routinely discarded to increase diagnostic accuracy. This is particularly important in diagnosing a patient who has resided in or visited a country endemic for other parasitic disease, such as is common in the Spanish migrant populations [41].
To the best of our knowledge this is the first large-scale evaluation of rK39-ICT and DAT for VL diagnosis in Europe. These results can inform public health practitioners in the region on the strengths and limitations of serological diagnosis. In addition to serology, however, PCR diagnosis should always be considered for confirmation of infection, and for the added benefit that molecular characterization brings.
Finding appropriate diagnostic solutions to VL is not only important to contain the burden of this Neglected Tropical Disease, but it will also help in the implementation of the United Nations Sustainable Development Goal of Universal Health Coverage [42].
|
10.1371/journal.pntd.0007250 | A polyvalent coral snake antivenom with broad neutralization capacity | Coral snakes of the genus Micrurus have a high diversity and wide distribution in the Americas. Despite envenomings by these animals being uncommon, accidents are often severe and may result in death. Producing an antivenom to treat these envenomings has been challenging since coral snakes are difficult to catch, produce small amounts of venom, and the antivenoms produced have shown limited cross neutralization. Here we present data of cross neutralization among monovalent antivenoms raised against M. dumerilii, M. isozonus, M. mipartitus and M. surinamensis and the development of a new polyvalent coral snake antivenom, resulting from the mix of monovalent antivenoms. Our results, show that this coral snake antivenom has high neutralizing potency and wide taxonomic coverage, constituting a possible alternative for a long sought Pan-American coral snake antivenom.
| Coral snakes are distributed in the Americas form Southern United States to Argentina. These snakes cause envenomings that, despite not being common, often lead to death. The antivenoms currently produced to treat accidents caused by these snakes have limitations regarding the number of species venoms they could neutralize. Here, we present an antivenom with a wide spectrum of neutralization, when compared to other Anticoral antivenoms. Nevertheless, more studies are still necessary to evaluate its neutralization capacity against the venoms of other species. This antivenom has great potential, as it neutralizes the lethal effects of some of the most common Micrurus species in the Americas.
| Coral snakes of the genus Micrurus and Micruroides represent a highly diverse neotropical monophyletic assembly of about 80 species distributed from the southern United States to northern Argentina [1].
Although uncommon (1–2% of the snake bites in the Americas) [2–4], Micrurus envenomation can be lethal due to the presence of potent toxic factors, mainly neurotoxins, causing peripheral paralysis resulting in respiratory failure [5]. The neurotoxic activity of coral snake venoms is mainly due to the presence of non-enzymatic competitive inhibitors of acetylcholine receptors at the neuromuscular junction known as α-neurotoxins of the three-finger (3FTx) protein superfamily and phospholipase A2 (PLA2) enzymes with pre-synaptic activity [5]. These two components have been revealed as the most abundant components in Micrurus venoms and vary in their proportion according to the species [5,6].
Snake antivenom production takes several stages and therefore considerable amounts of venom, in order to guarantee the quality of the medicament [7–10]. First, the toxicity of the venoms used for immunization must be determined (e.g. median lethal dose), then, animals (i.e. horses, goats) are inoculated with non-lethal doses of venom to produce a hyperimmune serum and subsequently, potency trials (e.g. median effective dose) must be carried out at different times in order to test the efficacy and stability of the product [9,10]. Micrurus snakes have relatively small sizes, which results in low venom yields, are difficult to find in the field and to maintain in captivity for extended periods of time. These aspects constitute serious setbacks for gathering sufficient amounts of venom for the production of coral snake antivenoms [11].
Antivenoms capable of neutralizing the toxic activities of a large range of heterologous Micrurus venoms have been long sought. Initially, as a mean to use antivenoms derived from snakes capable of yielding large amounts of venom against the toxic activities of snakes considered a public health threat but yielding very low amounts of venom per individual [12,13]. Later, as a way to produce antivenoms capable of neutralizing the lethal activities of a wide range of coral snake venoms that could be used in the Americas [8]. However, although antibody cross-reactivity has been widely observed between monovalent antisera and heterologous Micrurus venoms, in many cases resulting the ability of the antivenom to neutralize the lethal activity of the heterologous venom [14–16], in a number of cases and despite cross-reactivity, antivenoms are unable to neutralize the lethal effect of heterologous venoms [8,15,17,18]. In the Americas, anti-coral snake antivenoms are produced by the Instituto Nacional de Producción de Biolo’gicos (ANLIS) “Dr Carlos Malbrán” in Argentina, the Clodomiro Picado Institute (ICP) in Costa Rica, the Butantan Institute in Brazil, Instituto Bioclon in Mexico [19] and Laboratorios Probiol in Colombia [20]. However, while the antivenoms produced in Central America can neutralize the lethal activities of M. nigrocinctus, M. mosquitensis, M. dumerilii, M. fulvius, M. clarki, M. alleni and M. tener, they are unable to neutralize the lethal activities of M. mipartitus, M. surinamensis, M. spixii and M. pyrrhocryptus [21–24]. Likewise, those produced in South America, while able to neutralize the lethal activities of M. frontalis, M. corallinus, M. pyrrhocryptus, M. fulvius, M. nigrocinctus and M. surinamensis, are unable to neutralize the lethal activities of M. altirostris, M. ibiboboca, M. lemniscatus and M. spixii [25–27].
Based on the large extent of cross-reactivity between elapidic antivenoms and elapidic heterologous venoms and the cross neutralization of the lethal activity of a Notechis scutatus antivenom against the lethal activity of the M.fulvius venom [28], polyvalent anti-elapidic antivenoms have thus been considered as an alternative for the long sought development of a Pan-American anti-coral snake antivenom. In fact, a pentavalent anti-elapidic antivenom developed by CSL Limited in Australia using as antigens Notechis scutatus, Pseudechis australis, Pseudonaja textilis, Acanthophis antarcticus and Oxyuranus scutelatus venoms has been shown to neutralize the lethal activities of M. corallinus, M. frontalis, M. fulvius, M. nigrocinctus and M. pyrrhocryptus [29].
Here we report the production of a horse polyvalent anti-coral (Micrurus) snake antivenom derived from the mixing of monovalent antivenoms against M. dumerilii, M. mipartitus, M. isozonus and M. surinamensis venoms. The polyvalent antivenom is capable of neutralizing the lethal activity of M. dumerilii, M. mipartitus, M. isozonus, M. surinamensis, M. medemi, M. lemniscatus and M. spixii venoms thus constituting a promising Pan-American anti-coral antivenom.
The lyophilized venoms were obtained from the venom bank at the Instituto Nacional de Salud (INS) de Colombia, Bogotá. Venoms were kept frozen at -40°C. Species included in the study were chosen based on venom availability and inclusion on different Micrurus phyletic lineages [30,31]: M. mipartitus (Middle Magdalena Valley–MMV) of the bicolored group; M. dumerilii (MMV), M. medemi (Orinoco Basin—OB) from the monadal group and M. isozonus (OB), M. lemniscatus (OB), M. surinamensis (OB) of the triadal group (Fig 1). All venoms used were obtained from Colombian specimens.
Eight mixed breed horses were used with weights between 325–370 kg and between four to six years old. Horses were kept in the open, in pasture enclosures in a farm of the INS in Bojacá, Cundinamarca, Colombia, under veterinary care. Horses were vaccinated against tetanus and equine influenza, dewormed for gut helminths and washed to remove potential external parasites. Hematological, hepatic and kidney health was tested every six months and only horses with healthy organs until the last inspection were used for immunization. Mice CD-1 ICR strain, of 16–20 g, were obtained from the animal facility at the INS, Bogotá.
Experiments followed ethical procedures established in the protocols for animal experimentation at the INS (INT-R04.0000.01) and by the World Health Organization [9,10]. Animal experimentation was approved by the Institutional Committee for Animal Use and Care at the National Health Institute (Comité Institucional para el Cuidado y Uso de los Animales en el Instituto Nacional de Salud -CICUAL-INS), resolution 0052 of 2018.
Hyperimmune horse sera was obtained following the World Health Organization (WHO) guidelines [9,10] and the internal immunization protocol defined by the INS. In order to evaluate the immunogenicity of individual venoms and the capacity of individual antisera to cross neutralize heterologous venoms, experimental monospecific antivenoms were produced with the venom of four Micrurus species: M. dumerilii, M. isozonus, M. mipartitus and M. surinamensis. For each species venom, two horses were used. The immunization scheme for each horse lasted for up to three months, with injections administered every 5 to 15 days. For the first immunization, the venom was dissolved in Freund’s adjuvant (Becton Dickinson), whereas the remaining ones were dissolved in saline solution 0.85% (SS). Each injection had a volume between 0.5–5 mL, with 15–20 mg of venom, depending on the venom´s toxicity. Once the immunization scheme was completed, animals were bled to test whether there were quantifiable titers following neutralization procedures (see below). When appropriate antivenom neutralization titers were attained (≥3 LD50), horses were bled through puncture in the jugular vein. Up to eight liters of blood were collected in sterile plastic bags with anticoagulant, and plasma separation from cells was made by gravity. Cells were subsequently reinjected back into the horses for a better and faster recovery. Plasma was subsequently purified by means of precipitation with ammonium sulfate and sterilizing filtration, in order to obtain the concentrated antivenom immunoglobulin solution and stored at 2–8°C [32].
Polyvalent antivenom was produced by mixing of monovalent antivenoms and diluted to reach neutralization titers of 0.3 mg/mL of M. dumerilii and M. surinamensis, 0.8 mg/mL of M. mipartitus and 2 mg/mL of M. isozonus. This antivenom corresponds to the "Antiveneno Anticoral Polivalente", produced by the Instituto Nacional de Salud (INS), batch number 15AMP01, with expiration date of March of 2018.
Protein concentration was determined by the Kjeldahl method [10,33], following standardized protocol INS (MEN-R04.6020–010). Values correspond to grams per 100 mL and are expressed as percentage.
Protein content for antivenoms were 10.8% for anti-dumerilii, 8.2% for anti-mipartitus, 9.3% for anti-isozonus, 9.4% for anti-surinamensis and 8.1% for the polyvalent.
Venoms derived from the seven species studied showed a wide variation in lethality. Venom from M. mipartitus showed the lowest lethality (1.87 μg/g), whereas M. isozonus (0.35 μg/g) venom displayed the highest one (Table 1).
Monovalent antivenoms showed appropriate neutralization titers against homologous venoms. M. dumerilii and M. isozonus showed the lowest and highest titers, respectively (Table 2). The anti-dumerilii antivenom neutralized the lethality of M. isozonus and M. mipartitus venoms, with higher titers than those against the homologous venom, but with low titers against M. surinamensis venom. Anti-mipartitus antivenom showed low neutralization activity against M. dumerilii and moderate against M. isozonus and M. surinamensis venoms. The anti-isozonus antivenom displayed low neutralization titers against M. dumerilii, moderated against M. surinamensis and high against M. mipartitus venoms. Finally, the anti-surinamensis antivenom showed low neutralization capability against all heterologous venoms.
The antivenom showed a high capacity of neutralizing the effect of both homologous and heterologous venoms (Table 2). Neutralization capacity against homologous venoms, was lowest against M. surinamensis, and highest against M. isozonus. The antivenom was able to neutralize the lethal effects of heterologous venoms derived from M. spixii (1.58 mg/mL), M. lemniscatus (0.58 mg/mL) and M. medemi (0.68 mg/mL). Surprisingly, its neutralization titers against the heterologous venoms tested were higher than the titers against the homologous venoms derived from M. dumerilii and M. surinamensis. Noteworthy, the neutralization titer against the M. spixii venom was the second highest (Table 2).
Protein content of some of the monovalent antivenoms surpass the upper limit of 10% recommended by WHO [10](e.g. anti-dumerilii 10.8%). Nevertheless, the polyvalent antivenom used as therapy, has a protein content below this limit (8.1%). This value is higher than the 5.5% reported for the antivenom produced by the Instituto Nacional de Producción de Biológicos, Argentina and 4% reported for the Coralmyn, Bioclon, Mexico[26]. Such differences in protein content might be associated to the polyvalence of the antivenom and to the relatively high neutralization titers. It is believed that high protein concentration might increase the probability of adverse reactions [10]. Additionally, the relatively high neutralization titers compensate for this, since less medicament is required, therefore diminishing the total amount of protein administered to the patient.
Our results show a wide variation within the seven venoms tested and important differences as compared with the LD50 values found for the same species in other studies (Table 1). Estimations of the LD50 for the venom of a given species varied within studies, to the extent that the maximum value was almost 12 times the value of the minimum measurement (i.e. M. surinamensis; Table 1). It is difficult to explain the amount of variability within a species, given the number of variables that may influence the final results. Methods to estimate LD50 values vary according to several factors: mice weight and strain, volume of administration, venom treatment (e.g. dried vs lyophilized), inoculation route (e.g. intravenous vs intraperitoneal), etc. All these variables have proven to influence the final results [43]. On the other hand, differences in venom lethality may be the result of geographical variation [5]. For example, the venoms of M. dumerilii in this and other studies come from the middle Magdalena River Valley region of Colombia and LD50 are relatively similar among studies (Table 1). In the case of M. surinamensis, where LD50 varied widely, venoms originated from specimens captured over a large geographical distribution in the Orinoco and Amazonas basins [26,37,38]. Different regions may differ in many aspects (e.g. climate, geography) that may influence venom quality. Moreover, results by the same authors [26,41] for M. surinamensis from the same region, apparently using the same methodology, reached different results (Table 1). Therefore, at this point, conclusions regarding what is influencing differences in venom lethality may be hasty. Future efforts should be made to standardize procedures among laboratories in order to get comparable results.
As shown here, previous works found that monovalent antivenoms neutralize the lethal effects of homologous venoms [8,13,14,44] (Table 2). All monovalent antivenoms described in this study showed some degree of cross neutralization. Likewise, Cohen and collaborators [13,14], produced experimental monovalent antivenoms in rabbits by immunization with the venom of M. dumerilii, reaching high titers when neutralizing the homologous venom and moderate titers against two (M.fulvius and M. spixii) out of the seven venoms studied. In our trials, all monovalent antivenoms showed low neutralization titers against the lethal effect of M. dumerilii venom, contrary to other reports showing that this venom was neutralized by three (M. frontalis, M. fulvius, M. nigrocinctus) out of the four heterologous monovalent antivenoms tested [13,14]. On the other hand, the anti-surinamensis serum, as reported by several studies, showed low cross-neutralization [44]. Herein, we tested for the first time the neutralization capability of M. mipartitus and M. isozonus monovalent antivenoms: the first only showed high cross neutralization titers against M. isozonus and the second only against M. mipartitus (Table 2). It should be noted that Cohen and collaborators [14] tested the anti-dumerilii antivenom against the venom of a subspecies called M. mipartitus hertwigii, but this taxon is currently recognized as M. multifasciatus [45].
Our results show that cross neutralization does not operate in both directions. As stated before, anti-dumerilii antivenom showed high titers against M. mipartitus and M. isozonus, but low titers were recovered from anti-isozonus antivenom against M. dumerilii venom (Table 2). This observation is not new, other works using monovalent antivenom have found similar results [8,13,14,44]. This is an important fact that must be accounted for when designing antivenoms or eventually, when choosing antivenoms for envenomation treatments. For example, the antivenom produced in Costa Rica, which is produced using M. nigrocinctus venom as an antigen, neutralizes the lethality of M. dumerilii [23], one of the coral snakes involved in a large proportion of coral snake bite accidents in Colombia but the anti-dumerilii monovalent antivenom does not neutralize the activities of the M. nigrocintus venom [14].
Our data shows that the INS coral antivenom has good direct and cross neutralization titers (Tables 2 and 3). Particularly, the neutralization titers against all the heterologous venoms were higher than those against the homologous M. dumerilii and M. surinamensis venoms, as measured by either the amount of venom or the number of neutralized LD50s. Currently available Latin American coral snake antivenoms have shown different neutralization capabilities. The Brazilian, Instituto Butantan (raised against M. corallinus and M. frontalis), has proven to properly neutralize the venom of five species, but was ineffective against five [16,29,44,46]. Costa Rican monovalent antivenom (antiM. nigrocinctus), produced by Instituto Clodomiro Picado, has shown to be efficient against five species but unable to neutralize the venom of other two [21–24,47,48]. Mexican Coralmyn monovalent antivenom (against M. nigrocinctus), manufactured by Bioclon Laboratory, neutralizes the venom of three species, but is ineffective against four [25–27]. Finally, the monovalent Argentinian antivenom (raised against M. pyrrhocryptus) produced by Instituto Nacional de Productos Biológicos, has been shown to neutralize the venom of four species, but unable to neutralize the venom of other two [26]. The INS antivenom presented herein has wide neutralization capability against seven species. Further neutralization experiments against a wide range of Micrurus venoms are highly desirable.
The different neutralization range between the INS antivenom and other Latin American antivenoms is likely associated to the fact that most Micrurus antivenoms are mono or bivalent, whereas the INS is a mixture of antibodies raised against four phylogenetically different species. An early experimental polyvalent antivenom produced by Bolaños et al. [37] showed somehow similar results. This antivenom was raised against venoms derived from M. pyrrhocryptus (referred as M. frontalis pyrrhocryptus), M. multifasciatus (referred as M. mipartitus hertwigi) and M. nigrocinctus; and was able to neutralize the lethal effect of homologous and heterologous venoms (M. fulvius, M. dumerilii, M. frontalis, M. corallinus, M. spixii, M. mipartitus, M. alleni and M. lemniscatus. However, it was unable to neutralize the venom from M. surinamensis. Contrarily, an experimental polyvalent antivenom produced by Tanaka et al. [44], as a mixture of monovalent antivenoms raised against M. spixii, M. frontalis, M. corallinus, M. altirostris and M. lemniscatus, showed limited neutralizing efficacy.
Antivenoms from Brazil, Costa Rica and Mexico have not included the venom of M. surinamensis in their immunization schemes, and have very low or no neutralization capacity against this venom. The antivenom we developed includes the venom of this species in the immunization scheme, and displays high neutralization titers against the lethal effects of the M. surinamensis venom (Table 3). Given the particularities of this venom and the inability of heterologous antivenoms to neutralize M. surinamensis venom, the inclusion of venom derived from this species as an immunogen is important in the production of antivenoms in countries where this species occur, such as Brazil, Ecuador, Peru and Venezuela, in order to provide proper therapeutic alternatives [49]. Surprisingly, the commercial monovalent antivenom produced in Argentina, raised against M. pyrrhocryptus, proved to be effective against this species, which is another therapeutic alternative for this difficult to neutralize species venom (Table 3) [26]. Another antivenom that apparently neutralizes the venom of M. surinamensis is the one produced by Probiol [20]. This antivenom, derived from the immunization with M. lemniscatus, M. spixii and M. surinamensis venoms, claims to neutralize the venoms from M. mipartitus, M. surinamensis, M. dumerilii, M. medemi and M. spixii [20]. Nevertheless, the titers of neutralization are not known and no information is provided for the neutralization capacity against the homologous venom from M. lemniscatus.
When comparing the INS antivenom neutralization capacity against the species tested with other antivenoms, INS antivenom showed higher titers with respect to both the amount of venom and the number of median lethal doses neutralized, except for M. surinamensis which is more efficiently neutralized by the antivenom from the Instituto Nacional de Producción de Biológicos, Argentina. (Table 3). All studies here compared appraised the neutralization ability of antivenoms against three LD50, except for Tanaka et al. [16,44], that challenged against two, which might imply that Butantan’s antivenom might have lower neutralization capability. Additionally, our results proved that the INS antivenom neutralizes with high efficacy the lethality of a broad range of Micrurus species venoms (Table 3). These properties are desirable in the clinical practice. First, because with such titers, less amounts (volume and protein) of medicament are needed and the probability of adverse reactions reduces. Second, because a wide taxonomic coverage is always desired, since most of the time there is no appropriate identification of the species causing the accidents.
Comparisons among the neutralization capabilities of antivenoms, as for LD50 toxicity measurements, is difficult. Trials among studies vary widely in methodological aspects like the strain of mice, weight, challenging doses (i.e. from 2–5 LD50), value determination method (e.g. Spearman-Kärber, Probits) or route of injection (e.g. intraperitoneal vs. intravenous). Nevertheless, even if neutralization values vary, the fact that the tested antivenoms are or are not able to neutralize the studied venoms is hardly obscured.
The outcomes of this study show that INS antivenom is the best therapeutic alternative to treat coral snake envenomation in Colombia. Furthermore, this antivenom is the closest version of a long sought Pan-American anti-coral snake antivenom. Because most of the coral snake species whose venoms are neutralized by this antivenom are present in other south American countries, where no coral snake antivenom is produced, like Ecuador, Peru and Venezuela [50,51], or even Brazil, where the antivenom produced has a restricted efficacy for some species [16,44], this antivenom represent a treatment alternative for coral snake envenomation. Additionally, this antivenom might work in North America, given that cross neutralization of anti-M. dumerilii antivenom against M. fulvius venom has been reported [13,14]. On the other hand, the ability to neutralize the venom of Central American species remains to be proven, since only anti-dumerilii antivenom have been tested against M. nigrocinctus venom with negative results [14].
As aforementioned, the design of coral snake antivenoms has been hampered by low venom yields and unpredictable cross neutralization. Production of monovalent experimental antivenoms, evaluation of cross neutralization capacity and finally mixing of appropriate monovalent antivenoms to the desired neutralization titers is an effective approach for the production of polyvalent antivenoms. This way, producers might maximize limited resources (venom) while gaining knowledge on venom immunogenicity and sera cross reactivity.
Despite our promising results, various aspects must be accounted for. Around eighty species of Micrurus occur in the Americas, of which close to 30 occur in Colombia. We have tested the neutralization capacity against the venoms of only seven species. Even if those are the ones more often involved in accidents, there is a substantial number of questions that require our understanding. Examples of this are the spectrum of neutralization of this antivenom, the neutralization capacity against independent activities, such as neurotoxicity and myotoxicity and the best formulation of venom combinations required to produce an antivenom with high and broad neutralization capacities. An understanding of these aspects might also come from clinical results. Finally, this warrant large collaborative efforts to standardize neutralizations tests for comparative purposes and test anti-coral snake antivenoms produced in the Americas against a large number of Micrurus venoms.
|
10.1371/journal.ppat.1003500 | The Footprint of Genome Architecture in the Largest Genome Expansion in RNA Viruses | The small size of RNA virus genomes (2-to-32 kb) has been attributed to high mutation rates during replication, which is thought to lack proof-reading. This paradigm is being revisited owing to the discovery of a 3′-to-5′ exoribonuclease (ExoN) in nidoviruses, a monophyletic group of positive-stranded RNA viruses with a conserved genome architecture. ExoN, a homolog of canonical DNA proof-reading enzymes, is exclusively encoded by nidoviruses with genomes larger than 20 kb. All other known non-segmented RNA viruses have smaller genomes. Here we use evolutionary analyses to show that the two- to three-fold expansion of the nidovirus genome was accompanied by a large number of replacements in conserved proteins at a scale comparable to that in the Tree of Life. To unravel common evolutionary patterns in such genetically diverse viruses, we established the relation between genomic regions in nidoviruses in a sequence alignment-free manner. We exploited the conservation of the genome architecture to partition each genome into five non-overlapping regions: 5′ untranslated region (UTR), open reading frame (ORF) 1a, ORF1b, 3′ORFs (encompassing the 3′-proximal ORFs), and 3′ UTR. Each region was analyzed for its contribution to genome size change under different models. The non-linear model statistically outperformed the linear one and captured >92% of data variation. Accordingly, nidovirus genomes were concluded to have reached different points on an expansion trajectory dominated by consecutive increases of ORF1b, ORF1a, and 3′ORFs. Our findings indicate a unidirectional hierarchical relation between these genome regions, which are distinguished by their expression mechanism. In contrast, these regions cooperate bi-directionally on a functional level in the virus life cycle, in which they predominantly control genome replication, genome expression, and virus dissemination, respectively. Collectively, our findings suggest that genome architecture and the associated region-specific division of labor leave a footprint on genome expansion and may limit RNA genome size.
| RNA viruses include many major pathogens. The adaptation of viruses to their hosts is facilitated by fast mutation and constrained by small genome sizes, which are both due to the extremely high error rate of viral polymerases. Using an innovative computational approach, we now provide evidence for additional forces that may control genome size and, consequently, affect virus adaptation to the host. We analyzed nidoviruses, a monophyletic group of viruses that populate the upper ∼60% of the RNA virus genome size scale. They evolved a conserved genomic architecture, and infect vertebrate and invertebrate species. Those nidoviruses that have the largest known RNA genomes uniquely encode a 3′-to-5′exoribonuclease, a homolog of canonical DNA proof-reading enzymes that improves their replication fidelity. We show that nidoviruses accumulated mutations on par with that observed in the Tree of Life for comparable protein datasets, although the time scale of nidovirus evolution remains unknown. Extant nidovirus genomes of different size reached particular points on a common trajectory of genome expansion. This trajectory may be shaped by the division of labor between open reading frames that predominantly control genome replication, genome expression, and virus dissemination, respectively. Ultimately, genomic architecture may determine the observed genome size limit in contemporary RNA viruses.
| Genome size is the net result of evolution driven by the environment, mutation, and the genetics of a given organism [1], [2]. Particularly mutation rate is a powerful evolutionary factor [3]. The relation between mutation rate and genome size is inversely proportional for a range of life forms from viroids to viruses to bacteria, and it is positive for eukaryotes, suggestive of a causative link [4]–[6]. The genome size of RNA viruses is restricted to a range of ∼2 to 32 kb that corresponds to a very narrow band on the genome size scale (ranging from 1 kb to 10 Mb) across which genome size increase is correlated with mutation rate decrease [7]. This restricted genome size range of RNA viruses was believed to be a consequence of the universal lack of proof-reading factors, resulting in a low fidelity of RNA replication [8], [9].
In the above relation, mutation rate and proof-reading serve as a proxy for replication fidelity and genetic complexity, respectively. Replication fidelity, genome size, and genetic complexity were postulated to lock each other, through a triangular relation [10], in a low state in primitive self-replicating molecules [11]. This trapping, known as the “Eigen paradox” [12], was extended to include RNA viruses [13], providing a conceptual rationale for the small range of genome sizes in these viruses. Recent studies of the order Nidovirales, a large group of RNA viruses that includes those with the largest genomes known to date, provided strong support for the postulated triangular relation [10], [14]. Unexpectedly, they also revealed how nidoviruses may have solved the Eigen paradox by acquiring a proof-reading enzyme. These advancements implied that the control of genome size may be more complex than previously thought, in RNA viruses in general, and particularly in nidoviruses.
The order Nidovirales is comprised of viruses with enveloped virions and non-segmented single-stranded linear RNA genomes of positive polarity (ssRNA+), whose replication is mediated by a cognate RNA-dependent RNA polymerase (RdRp) [15], [16]. The order includes four families - the Arteriviridae and Coronaviridae (including vertebrate, mostly mammalian viruses), and the Roniviridae and Mesoniviridae (invertebrate viruses). The unusually broad 12.7 to 31.7 kb genome size range of this monophyletic group of viruses includes the largest known RNA genomes, which are employed by viruses from the families Roniviridae (∼26 kb) [17] and Coronaviridae (from 26.3 to 31.7 kb) [18], that have collectively been coined “large-sized nidoviruses” [19]. Viruses from the Arteriviridae (with 12.7 to 15.7 kb genomes) [20] and the recently established Mesoniviridae (20.2 kb) [21], [22] are considered small and intermediate-sized nidoviruses, respectively. Nidoviruses share a conserved polycistronic genomic architecture (known also as “organization”) in which the open reading frames (ORFs) are flanked by two untranslated regions (UTRs) [10], [23]–[26]. The two 5′-proximal ORFs 1a and 1b overlap by up to a few dozen nucleotides and are translated directly from the genomic RNA to produce polyproteins 1a (pp1a) and pp1ab, with the synthesis of the latter involving a −1 ribosomal frameshift (RFS) event [27]–[29]. The pp1a and pp1ab are autoproteolytically processed into nonstructural proteins (nsp), named nsp1 to nsp12 in arteriviruses and nsp1 to nsp16 in coronaviruses (reviewed in [30]). Most of them are components of the membrane-bound replication-transcription complex (RTC) [31]–[33] that mediates genome replication and the synthesis of subgenomic RNAs (a process known also as “transcription”) [34], [35]. ORF1a encodes proteases for the processing of pp1a and pp1ab (reviewed in [30]), trans-membrane domains/proteins (TM1, TM2, and TM3) anchoring the RTC [36]–[38] and several poorly characterized proteins. ORF1b encodes the core enzymes of the RTC (reviewed in [39], see also below). Other ORFs, whose number varies considerably among nidoviruses are located downstream of ORF1b (hereafter collectively referred to as 3′ORFs). They are expressed from 3′-coterminal subgenomic mRNAs [40], and encode virion and, optionally, so-called “accessory proteins” (reviewed in [41]–[43]). The latter, as well as several domains encoded in ORF1a and ORF1b, were implicated in the control of virus-host interactions [44]–[48].
In addition to comparable genome architectures, nidoviruses share an array (synteny) of 6 replicative protein domains. Three of these are most conserved enzymes of nidoviruses: an ORF1a-encoded protease with chymotrypsin-like fold (3C-like protease, 3CLpro) [49]–[51], an ORF1b-encoded RdRp [49], [52], [53] and a superfamily 1 helicase (HEL1) [54]–[57] (reviewed in [58]). For other proteins, relationships have been established only between some nidovirus lineages, mostly due to poor sequence similarity. Two tightly correlated properties separate large- and intermediate-sized nidoviruses from all other ssRNA+ viruses, classified in several dozens of families and hundreds of species: a genome size exceeding 20 kb and the presence of a gene encoding a RNA 3′-to-5′ exoribonuclease (ExoN), which resides in nsp14 in the case of coronaviruses [10]. The latter enzyme is distantly related to a DNA proofreading enzyme, and it is genetically segregated and expressed together with RdRp and HEL1 [14], [59]. Based on these properties ExoN was implicated in improving the fidelity of replication in large- and intermediate-sized nidoviruses. This hypothesis is strongly supported by the excessive accumulation of mutations in ExoN-defective mutants of two coronaviruses, mouse hepatitis virus [60] and severe acute respiratory syndrome coronavirus (SARS-CoV) [61], the identification of an RNA 3′-end mismatch excision activity in the SARS-CoV nsp10/nsp14 complex [62], and the high efficacy of a live coronavirus vaccine displaying impaired replication fidelity due to nsp14-knockout [63] (for review see [64], [65]). Although the molecular mechanisms underlying ExoN's function in fidelity control remain to be elucidated, its acquisition by nidoviruses likely enabled genome expansions beyond the limit observed for other non-segmented ssRNA+ viruses [10], [19]. Since ExoN-encoding nidoviruses have evolved genomes that may differ by up to ∼12 kb (from 20.2 kb of Nam Dinh virus, NDiV, to 31.7 kb of Beluga whale coronavirus SW1, BWCoV-SW1), there must be other factors in addition to the proof-reading enzyme that control genome size.
In this study we sought to characterize the dynamics of nidovirus genome expansion (NGE). The NGE is defined by the entire range of the genome sizes of extant nidoviruses, from 12.7 to 31.7 kb, and thus concerns both pre- and post-ExoN acquisition events. Our analysis revealed that ExoN acquisition was part of a larger process with non-linear dynamics, during which distinct coding regions of the nidovirus genome were expanded to accommodate both an extremely large number of mutations and virus adaptation to different host species. Our results indicate that genome architecture and the associated region-specific division of labor [1] leave a footprint on the expansion dynamics of RNA virus genomes through controlling replication fidelity and/or other mechanisms. Eventually, these constraints may determine the observed limit on RNA virus genome size.
Nidoviruses have evolved genomes in a size range that accounts for the upper ∼60% of the entire RNA virus genome size scale and include the largest RNA genomes [10]. What did it take to produce this unprecedented innovation in the RNA virus world? This question could be addressed in two evolutionary dimensions: time and amount of substitutions. Due to both the lack of fossil records and high viral mutation rates, the time scale of distant relations of RNA viruses remains technically difficult to study. Hence, we sought to estimate the amount of accumulated replacements in conserved nidovirus proteins and to place it into a biological perspective by comparing it with that accumulated by proteins of cellular species in the Tree of Life (ToL).
To this end, we used a rooted phylogeny for a set of 28 nidovirus representatives (Table S1), which was based on a multiple alignment of nidovirus-wide conserved protein regions in the 3CLpro, the RdRp and the HEL1, as described previously [10]. The 28 representatives covered the acknowledged species diversity of nidoviruses with completely sequenced genomes [17], [18], [20], [21] and included two additional viruses. For the arterivirus species Porcine reproductive and respiratory syndrome virus we selected two viruses, representing the European and North American genotypes, respectively, because we observed an unusually high divergence of these lineages; for the ronivirus species Gill-associated virus we selected two viruses representing the genotypes gill-associated virus and yellow head virus, respectively, because these viruses showed a genetic distance comparable to that of some coronavirus species [21] (CL & AEG, in preparation). The nidovirus-wide phylogenetic analysis consistently identified the five major lineages: subfamilies Coronavirinae and Torovirinae, and families Arteriviridae, Roniviridae and Mesoniviridae. The root was placed at the branch leading to arteriviruses (Fig. 1A) according to outgroup analyses [10]. Accordingly, arteriviruses with genome sizes of 12.7 to 15.7 kb are separated in the tree from other nidoviruses with larger genomes (20.2–31.7 kb).
We compared the evolutionary space explored by nidoviruses, measured in number of substitutions per site in conserved proteins, with that of a single-copy protein dataset representing the ToL [66] (Fig. 1B). Using a common normalized scale of [0,1], comparison of the viral and cellular trees and associated pairwise distance distributions revealed that the distances between cellular proteins (0.05–0.45 range) cover less than half the scale of those separating nidovirus proteins. (Fig. S1). Unlike cellular species, nidoviruses are grouped in a few compact clusters, which are very distantly related. The distances between nidovirus proteins are unevenly distributed, reflecting the current status of virus sampling: intragroup distances between nidoviruses forming major lineages are in the 0.0–0.25 range, while intergroup distances between nidoviruses that belong to different lineages are in the 0.55–1.0 range. The distances separating the intermediate-sized mesonivirus from other nidoviruses tend to be most equidistant, accounting for ∼15% of all distances in the 0.55–0.85 range. Consequently, nidovirus evolution involved the accumulation of mutations in the most conserved proteins at a scale comparable to that of the ToL. This observation is instructive in two ways. First, it can be contrasted with the conservation of nidovirus genome architecture [58], which emerges in this context as truly exceptional by conventional evolutionary considerations. Second, it makes it plausible that other, less conserved proteins might have diverged beyond the level that can be recognized by sequence alignment, thus establishing limits of the applicability of the alignment-based analysis of nidoviruses. We used both these insights to advance our study further (see below).
To quantify the relation of genome size change and the accumulation of substitutions, we plotted pairwise evolutionary distances (PED) separating the most conserved replicative proteins (Y axis) versus genome size differences (X axis) for all pairs of nidoviruses in our dataset (Fig. 2). It should be noted that the observed genome size differences may serve only as a low estimate for the actual genome size change, since it does not account for (expansion or shrinkage) events that happened in parallel between two viruses since their divergence. The obtained 378 values are distributed highly unevenly, occupying the upper left triangle of the plot. Using phylogenetic considerations (Figs. 1A and S1), four clusters could be recognized in the plot. Genetic variation within the four major virus groups with more than one species (arteri-, corona-, roni-, and toroviruses) is confined to a compact cluster I in the left bottom corner (X range: 0.033–4.521 kb, Y range: 0.051–1.401). Values quantifying genetic divergence between major lineages are partitioned in three clusters taking into account genome sizes: large-sized vs. large-sized nidoviruses (cluster II, X: 0.002–5.433 kb, Y: 3.197–4.292), intermediate-sized vs. other lineages (cluster III, X: 4.475–11.494 kb, Y: 2.896–4.553), and small-sized vs. large-sized nidoviruses (cluster IV, X: 10.536–18.978 kb, Y: 4.159–5.088). Points in clusters I, III and IV are indicative of a positive proportional relation between genome size change and the accumulation of replacements. The off-diagonal location of cluster II can be reconciled with this interpretation under the (reasonable) assumption that the three lineages of large-sized nidoviruses expanded their genomes independently and considerably since diverging from their most recent common ancestor (MRCA). This positive relation is also most strongly supported by the lack of points in the bottom-right corner of the plot (large difference in genome size; small genetic divergence). Overall, this analysis indicates that a considerable change in genome size in nidoviruses could have been accomplished only when accompanied by a large number of substitutions in the most conserved proteins.
Next, we asked whether genome size change could be linked to domain gain and loss. We analyzed the phylogenetic distribution of protein domains that were found to be conserved in one or more of the five major nidovirus lineages [10]. Ancestral state parsimonious reconstruction was performed for the following proteins: ORF1b-encoded ExoN, N7-methyltransferase (NMT) [67], nidovirus-specific endoribonuclease (NendoU) [68], [69], 2′-O-methyltransferase (OMT) [70], [71], ronivirus-specific domain (RsD) (this study; see legend to Fig. S2), and ORF1a-encoded ADP-ribose-1″-phosphatase (ADRP) [72]–[74]. This analysis revealed that domain gain and loss have accompanied NGE (Fig. S2 and Table S2). Particularly, the genetically segregated ExoN, OMT and NMT domains (Fig. 3) were acquired in a yet-to-be determined order during the critical transition from small-sized to intermediate-sized nidovirus genomes. However, the combined size of these domains [10] accounts for only a fraction (49.7%) of the size difference (4,475 nt) between the genomes of NDiV (20,192 nt) and Simian hemorrhagic fever virus (SHFV), which has the largest known arterivirus genome (15,717 nt). The fraction that could be attributed to these and the three other protein domains is even smaller in other pairs of viruses representing different major nidovirus lineages (CL & AEG). This analysis is also complicated by the uncertainty about the genome sizes of nidovirus ancestors that acquired or lost domains.
In order to gain further insight in NGE dynamics, we analyzed large genome areas in which homology signals were not recoverable in the currently available dataset because of both the extreme divergence of distant nidoviruses and the relatively poor virus sampling (Fig. 1). To address this challenge, we developed an approach that establishes and exploits relationships between nidovirus genomes in an alignment-free manner on grounds other than sequence homology. To this end, we partitioned the nidovirus genome according to functional conservations in the genome architecture, using results for few characterized nidoviruses and bioinformatics-based analysis for most other viruses (reviewed in [19]). With this approach, the genomes of all nidoviruses can be consistently partitioned into five regions in the 5′ to 3′ order: 5′-UTR, ORF1a, ORF1b, 3′ORFs, and 3′-UTR (Fig. 3, Table S3). The 5′-UTR and 3′-UTR flank the coding regions and account for <5% of the nidovirus genome size. The borders of the three ORF regions that overlap by few nucleotides in some or all nidoviruses were defined as follows: ORF1a: from the ORF1a initiation codon to the RFS shifty codons; ORF1b: from the RFS signal to the ORF1b termination codon; and 3′ORFs: from the ORF1b termination codon to the termination codon of the ORF immediately upstream of the 3′UTR. As we detail in the Supplementary text (Text S1), the three ORF regions are of similar size but differ in expression mechanism (Fig. 3 top) and principal function. Thus, ORF1a dominates the expression regulation of the entire genome, and ORF1b encodes the principal enzymes for RNA synthesis, while the 3′ORFs control genome dissemination. This region-specific association may be described as a division of labor [1].
We then asked how the different regions contributed to the genome expansion. We initially noted that the intermediate position of the mesonivirus between the two other nidovirus groups is observed only in genome-wide but not in region-specific size comparisons (Fig. 4). In the latter, the mesonivirus clusters with either small-sized (ORF1a and 3′ORFs) or large-sized (ORF1b) nidoviruses. This non-uniform position of the mesonivirus relative to other nidoviruses is indicative of a non-linear relationship between the size change of the complete genome and its individual regions during NGE. Accordingly, when fitting weighted linear regressions for the three regions separately to the six datasets formed by nidoviruses with small and large genomes, support for a linear relationship was found only for the 3′ORF dataset of large nidoviruses; for all other regions a linear relationship was not statistically significant (Fig. S3). These results prompted us to evaluate linear as well as non-linear regression models applied to a dataset including all known nidovirus species (n = 28) (Fig. 5). Two non-linear models were employed: third order monotone splines and a double-logistic regression. In the monotone splines, two parameters – the number and position of knots – determine the regression fit. We identified values for both parameters that result in the best fit (Fig. S4).
Using weighted r2 values, we observed that the splines model captures 92.9–96.1% of the data variation for the three ORF regions. This was a 5–22% gain in the fit compared to the linear model (75.9–90.8%) (Fig. 5). This gain was considered statistically significant (α = 0.05) in two F-tests, a specially designed one and a standard one, as well as in the LV-test for every ORF region (p = 0.019 or better) and, particularly, their combination (p = 9.1e-6 or better) (Table 1). The splines model also significantly outperforms the double-logistic model (p = 0.0014) (Table 1). These results established that the nidovirus genome expanded in a non-linear and region-specific fashion.
Like each region, also the entire genome must have expanded non-linearly during NGE. Revealing its dynamic was our next goal. To this end, we analyzed the contribution of each of the five genomic regions to the overall genome size increase under the three models (Fig. 6 and Fig. S5). The top-ranking splines model (Table 1) predicts a cyclic pattern of overlapping wavelike size increases for the three coding regions (the 5′ and 3′UTR account only for a negligibly minor increase that is limited to small nidoviruses). Each of the three coding regions was found to have increased at different stages during NGE (Fig. 6). A cycle involves expanding predominantly and consecutively the ORF1b, ORF1a, and 3′ORFs region. One complete cycle flanked by two partial cycles are predicted to have occurred during the NGE from small-sized to large-sized nidoviruses. The complete cycle encompasses almost the entire genome size range of nidoviruses, starting from 12.7 kb and ending at 31.7 kb. The dominance of an ORF region in the increase of genome size was characterized by two parameters: a genome size range (X axis in Fig. 6) in which the contribution of a region accounts for a >50% share of the total increase, and by the maximal share it attains in the NGE (Y axis in Fig. 6). For three major regions these numbers are: ORF1b, dominance in the 15.8–19.3 kb range with 72.9% maximal contribution at genome size 17.5 kb; ORF1a, 19.7–26.1 kb and 81.3% at 22.7 kb; 3′ORFs, 26.1–31.7 kb and 89.6% at 29.5 kb (Fig. 6).
Furthermore, the shapes of the three waves differ. The first one (ORF1b) is most symmetrical and it starts and ends at almost zero contribution to the genome size change. This indicates that the ORF1b expansion is exceptionally constrained, which is in line with the extremely narrow size ranges of ORF1b in arteri- and coronaviruses (with mean±s.d. of 4362±86 and 8073±50 nt, respectively; Fig. 4 and Fig. 6). The second wave (ORF1a) is tailed at the upper end and is connected to the ORF1a wave from the prior cycle. This ORF seems to have a relatively high baseline contribution (∼20%) to the genome size change up to the range of coronaviruses. The third wave (3′ORFs) is most asymmetrical (incomplete), as it only slightly decreases from its peak toward the largest nidovirus genome size to which this region remains the dominant contributor (∼77%).
One partial cycle, preceding the complete one, is observed inside the genome size range of arteriviruses and involves the consecutive expansions of ORF1a and 3′ORFs, respectively. Also the main, but still very limited contributions of 5′- and 3′-UTRs (<6%) are observed here. The start of another incomplete cycle, involving the expansion of ORF1b and overlapping with the complete cycle, is observed within the upper end of coronavirus genome sizes.
It must be stressed that nidoviruses occupy different positions on the trajectory that depicts the entire NGE dynamics. For the viruses with large genomes those with smaller genomes represent stages that they have passed during NGE; in this respect the latter may resemble ancestral viruses which have gone extinct. For the smaller genomes those with the larger ones represent stages that they have not reached during NGE. Mesonivirus and roniviruses seem to have been “frozen” after the first (ORF1b) and second (ORF1a) wave, respectively. The third wave (3′ORFs) was due to the genome expansion of coronaviruses and, to a lesser extent, toroviruses (compare the genome sizes of these viruses and the third wave position in Fig. 6). These observations reveal that the constraints on genome size due to genome architecture may be modulated in a lineage-dependent manner.
In this study we provide, for the first time, a quantitative insight into the large-scale evolutionary dynamics of genome expansion in RNA viruses that concerns the upper ∼60% of the RNA virus genome size scale exclusively populated by nidoviruses. In view of the extremely large amount of substitutions that accumulated in the nidovirus genome during evolution, we exploited the functional conservation in the nidovirus genome architecture to partition genomes of nidoviruses into five non-overlapping regions. Using a complex statistical framework, we discovered that consecutive, region-specific size increases must have occurred during NGE. We conclude that the genome size dynamics in nidoviruses may be shaped by the division of labor between ORFs that predominantly control genome replication, genome expression, and virus dissemination, respectively.
Genome size evolution in RNA viruses, unlike that of DNA-based life forms, has received relatively little attention from the research community. The small range of RNA genome sizes might have been perceived as evidence for the lack of meaningful genome size dynamics in RNA viruses. Even if there was any dynamics, its reconstruction could be considered a challenging if not impossible task, since evolutionary signals of distant relationships would not be recoverable, possibly due to the saturation of the genome with substitutions [9], [75]. To our knowledge, genome size increase in RNA viruses has thus far been associated with only a few trends: a concomitant increase of the average size of replicative proteins [76], a reduction of genome compression as measured by gene overlaps [77], and a strong correlation between the presence of helicase [19], [78] and ExoN [10], [14] domains and the genome size in ssRNA+ viruses.
Now, by analyzing NGE, we show that even in the most conserved proteins genome expansion was accompanied by a considerable accumulation of replacements, which may approach saturation (Fig. 2). In other, less conserved proteins this effect is expected to be (much) larger. That relation is in line with the observation that nucleotide substitutions are on average four times more common than insertions/deletions in RNA viruses [7]. Practically, this result indicates that even for the study of a large monophyletic group like the nidoviruses, the power of substitution-based (phylogenetic) analysis is limited. We have overcome this limitation by employing an innovative approach that exploits functional conservation in genome architecture rather than sequence homology. The inferred non-linear dynamics of NGE is supported strongly by different statistical tests. However, in view of the highly uneven distribution of genomes sizes in our dataset, which may be considered a problem, we will provide additional supporting arguments below.
First of all, we note that a virus (called Cavally virus) that is closely related to the unique intermediate-sized NDiV was independently identified in a parallel study [26]. Both viruses share all properties that are critical for this study, including the size of genome and ORFs as well as the assignment of protein domains [21]. These results show that the NDiV characteristics used in our study are reliable. Second, these two mesoniviruses and the very distant roniviruses, which have large genomes, form a monophyletic group (Fig. 1). This clustering correlates with common (molecular) properties, including the infection of invertebrate hosts and the lack of the NendoU domain, which distinguish mesoni- and roniviruses from other vertebrate nidoviruses (Fig. S2) and may apply to other yet-to-be identified members of this group as well. Third, even if we restrict our analysis to small- and large-sized nidoviruses, differences between the size ranges of genomes versus the three ORF regions are already apparent (Fig. 4). Particularly striking are the extremely constrained size of ORF1b in both arteriviruses and coronaviruses as well as the exceptionally large size range of 3′ORFs in large-sized nidoviruses. These constraints contribute prominently to the first and third wave, respectively, of the major NGE cycle (Fig. 6). Thus, the described dynamics of the region-specific genome size increase reflects properties of both mesoniviruses and other nidoviruses, and is expected to be sustained while virus sampling continues.
Poor virus sampling limits the resolution of our reconstruction of domain gain/loss during NGE. For instance, the critically important acquisition of ExoN seems to be tightly correlated with those of two replicative methyltransferases, NMT and OMT (Fig. S2). The fact that NMT and ExoN are adjacent domains of a single protein in coronaviruses (nsp14) whereas OMT resides nearby (nsp16) in pp1ab suggests a link between these domains and indicates that NMT and ExoN may have been acquired in a single event. Furthermore, NMT and OMT were shown to be essential for cap formation at the 5′-end of coronavirus mRNAs [67], [70], [71], with the OMT-mediated modification proposed to be important for the evasion of innate immunity [47]. These enzymes are yet to be characterized in other large-sized nidoviruses.
The ExoN acquisition is a hallmark of the first NGE wave because it is expected to have improved the replication fidelity and, thus, made further genome expansion feasible. In contrast, no domain acquisition with a comparably strong biological rationale could be identified for the second wave. Two aspects, both contrasting the first and second wave, are noteworthy. Firstly, while the first wave seems to reflect genome expansion in a single ancestral lineage that might have given rise to all intermediate- and large-sized nidoviruses (founding event), the second wave encompasses expansions in several lineages that happened in parallel (Fig. S2B). Secondly, evolutionary relations of ORF1a-encoded proteins (underlying the second wave) are not as extensively documented as those for ORF1b (underlying the first wave), since ORF1a proteins in nidoviruses have diverged to a far greater extent. Hence, the domain gain/loss description for the second wave is even less complete than that for the first wave. Most notable is the acquisition of ADRP (formerly termed “X domain” [79]), whose physiological function remains elusive (see Supplementary text S2) and which seems to be part of the second wave in large-sized vertebrate nidoviruses (Fig. 6). Unlike the first and second wave, the third one encompasses changes that predominantly happened during the radiation of a subfamily (Coronavirinae) rather than several families (Fig. 6); they are being analyzed in a separate study (CL & AEG, in preparation). Improved future virus sampling, especially in the genome size range around 20 kb, could be critical for the description of domain gain/loss in ORF1a and its refinement in ORF1b (Fig. S2).
Products of ORF1b, ORF1a, and 3′-ORFs, whose expansion dynamics are reported here, cooperate bidirectionally in the nidovirus life cycle [15], since their functioning depends on each other (Fig. 7, bottom). In contrast, the order in which these regions expanded is unidirectional (Fig. 7 top). It implies a causative chain of events during NGE and suggests, for the first time and unexpectedly, a hierarchy of the three underlying biological processes. To our knowledge, no theory or results published provided a basis for the model describing how genome expansion must have proceeded. Now that the dynamics of NGE have been established, it could be further rationalized using experimental data on the functions of the proteins involved. Importantly and regardless of how plausible these functional considerations might sound, they do not substitute for the evidence of the inferred dynamics presented elsewhere in the paper.
The association of the first wave of domain acquisitions with ORF1b may attest to the universal critical role of replicative enzymes in NGE beyond the 20-kb threshold observed for other ssRNA+ viruses (for discussion see [10]). Regardless in which order the OMT, NMT and ExoN loci were acquired, their products must have been adapted to the core of the RTC that is formed by the ORF1b-encoded RdRp and HEL1-containing proteins [53], [58], [80]. Other, less conserved RTC components are encoded in ORF1a [34], [37], [70], [81]–[83]. It is known that proteins encoded in ORF1a and ORF1b interact in coronaviruses [34], [84], [85] and likely arteriviruses [86], [87]. Some of these interactions, e.g. between nsp10 and nsp14 or nsp16, were shown to be essential for the function of the ORF1b-encoded enzymes [62], [88], [89]. Accordingly, the RTC, already enlarged with the newly acquired ORF1b-encoded subunits, could have triggered and/or become accommodative of the expansion of ORF1a. Additionally, the ORF1a expansion may have been prompted by the need to adapt the expression mechanisms it controls to the changes of the ORF1b-encoded part that had already increased in size and complexity. The final wave of expansion involving the 3′ORFs may have been triggered by the need for virus particle adaptation to accommodate the expanded genome. This plausible link was extensively explored in the literature that implicated packaging head space in the control of genome size in other viruses [90]–[94]). The sizes of genomes and virus particles may also be correlated in nidoviruses, although the evolution of virion size in nidovirus lineages has not been studied to our knowledge. During NGE, a part of the newly acquired genetic material may have been adapted to facilitate both virus-host interactions [46], [48], [95], [96] and coordination between the three ORF regions for the benefit of the processes they control and the life cycle [97]. For instance, in arteriviruses the ORF1a-encoded nsp1 is essential for subgenomic mRNA synthesis and virion biogenesis [86], [87], [98] and a role in transcription was proposed for an ORF1a-encoded domain of nsp3 in coronaviruses [99]. Thus, factors encoded by ORF1a and ORF1b might constrain NGE by controlling the expression of the 3′ORFs region and/or the functioning of its products. This would explain why the 3′ORFs expansion could not have been possible before the expansion of ORF1a and ORF1b. Based on a similar line of reasoning, an extremely tight control of the ORF1b size (Fig. 4) may set the ultimate NGE size limit.
The order in which the three coding regions expanded matches their ranking in terms of sequence conservation, which is evident from the distribution of nidovirus conserved domains across these regions (Figs. 2 and 3). This conservation is inversely proportional to the amount of accumulated substitutions, although a quantitative characterization of the latter aspect is yet to be systematically documented. Genome changes due to region-specific expansion and residue substitution may affect each other, and both may contribute to virus adaptation to the host. In this respect we noticed that viruses with larger genomes, compared to their small-sized cousins, may employ a larger repertoire of proteins for interacting with the host. It is also apparent that large-sized nidoviruses may afford both the acquisition and loss of an ORF as a matter of genome variation in a species (see e.g. [100]–[102]; for review see [103]). Thus, large genomes could provide nidoviruses with an expanded toolkit to adapt upon crossing species barriers and to explore new niches in established hosts.
It is broadly acknowledged that high mutation rates and large population sizes allow RNA viruses to explore an enormous evolutionary space and to adapt to their host [76], [104]. Yet the low fidelity of replication also confines their evolution within a narrow genome size range that must affect their adaptation potential. Above, we present evidence for a new constraint on genome size in RNA viruses. In our analysis of nidoviruses, the conserved genome architecture and associated division of labor emerged as potentially powerful forces that are involved in selecting both new genes and positions of gene insertion during genome expansion. In this respect, the established wavelike dynamics of regional size increase could be seen as the footprint of genome architecture on genome size evolution. Ultimately, these constraints may determine the upper limit of the RNA virus genome size. The reported data point to an important evolutionary asymmetry during genome expansion, which concerns the relation between proteins controlling genome replication, expression, and dissemination, and may certainly be relevant beyond the viruses analyzed here.
Importantly, the major diversification of nidoviruses by genome expansion must have started at some early point after the acquisition of ExoN [10]. From that point on, nidoviruses expanded their genomes in parallel in an increasing number of lineages, each of which may have acquired different domains in the same region. Extant representatives of the major lineages have very different genome sizes and essentially offer snapshots of different NGE stages. It seems that the host range may affect the outcome of this process, since the two families that infect invertebrates are on the lower end of the genome size range in the ExoN-encoding nidoviruses. For yet-to-be described nidoviruses, the genome expansion model can predict the sizes of the three coding regions by knowing the genome size only. The mechanistic basis of this fundamental relation can be probed by comparative structure-function analyses, which may also advance the development of nidovirus-based vectors and rational measures for virus control. Thus, the wavelike dynamics model links virus discovery to basic research and its various applications.
A dataset of nidoviruses representing species diversity from the three established and a newly proposed virus family was used (Table S1). A multiple alignment of nidovirus-wide conserved protein domains (28 species, 3 protein families, 604 aa alignment positions, 2.95% gap content) as described previously [10] formed the basis of all phylogenetic analyses. To put the scale of the nidovirus evolution into an independent perspective, we compared it with a cellular dataset previously used to reconstruct the ToL, for which a concatenated alignment of single-copy proteins was used (30 species, 56 protein families, 3336 aa alignment positions, 2.8% gap content) [66]. The proteins used in the nidoviral and cellular datasets are the most conserved in their group and, as such, could be considered roughly equivalent and suitable for the purpose of this comparative analysis.
Rooted phylogenetic reconstructions by Bayesian posterior probability trees utilizing BEAST [105] under the WAG amino acid substitution matrix [106] and relaxed molecular clock (lognormal distribution) [107] were performed as described previously [10]. Evolutionary pairwise distances were calculated from the tree branches. A maximum parsimony reconstruction of the ancestral nidovirus protein domain states at internal nodes of the nidovirus tree was conducted using PAML4 [108].The quality of ancestral reconstructions was assessed by accuracy values provided by PAML4. The nidovirus genomic sequences are non-independent due to their phylogenetic relatedness [109]. When calculating the contribution of individual sequences to the total observed genetic diversity the uneven sampling of different phyletic lineages must be accounted for. To correct for the uneven sampling we assigned relative weights to the 28 nidovirus species by using position-based sequence weights [110] that were calculated on the alignment submitted for phylogeny reconstruction. The weights were normalized to sum up to one and were used in regression analyses (see below). The sequence weights varied ∼7 fold from 0.017 to 0.116. NDiV, which represents mesoniviruses, showed the largest weight of 0.116 that was distantly followed by those of the bafinivirus White bream virus (WBV; 0.075) and roniviruses (0.06 each); coronaviruses, making up the best-sampled clade, were assigned the lowest weights (0.017 to 0.028 each).
The genome of each nidovirus was consistently partitioned into five genomic regions according to external knowledge (see Results). To model the contribution of each genomic region to the total genome size change, we conducted weighted regression analyses (size of a genomic region on size of the genome) using three models – a linear and two non-linear ones. Position-based sequence weights were used and a confidence level of α = 0.05 was applied in all analyses. The regressions of the different genomic regions were not fitted separately but were joined to produce a genome-wide analysis. The combined contribution of all genomic regions to the genome size change must obviously sum up to 100%. To satisfy this common constraint, in each analysis, regression functions were fitted simultaneously to sizes of the genomic regions by minimizing the residual sum of squares, thereby constraining the sum of all slopes to be not larger than one. The linear model assumes a constant contribution of each genomic region during evolution which was modeled via linear regions.
In the first non-linear model we applied third order monotone splines with equidistant knots [111]. We chose splines because of their flexibility and generality (we do not rely on a specific regression function). The monotonicity constraint was enforced to avoid overfitting which was observed otherwise, and third order functions were chosen to obtain smooth, second-order derivatives. We explored the dependence of the performance of the splines model on variations in two critical parameters, the number of knots and the start position of the first knot. These two parameters define a knot configuration and determine a partitioning of the data into bins. In the first test we evaluated five different configurations generating from three to seven knots. Configurations using eight or more knots resulted in some bins being empty and were therefore not considered. For each number of knots the position of the first knot and the knot distance were determined as resulting in that configuration for which the data points are distributed most uniformly among the resulting bins. The exception was the 3-knot configuration, in which the position of the second knot was selected as the intermediate position in the observed genome size range (22.2 kb). Only configurations with equidistant knots were considered. All probed splines models were evaluated by goodness-of-fit values (weighted version of the coefficient of determination r2). In the second test we evaluated the model dependence on the position of the first knot by considering all positions that do not result in empty bins for the optimal number of knots determined using the approach described above.
As another non-linear model we used a 7-parameter double-logistic regression function that mimics the splines model and more readily allows for biological interpretations. Logistic functions discriminate between two principal states – stationary and growth phases; a double-logistic curve comprises not more than three steady and two growth phases. The “length” of the different phases (in the dimension of the independent variable; e.g. genome size), the steady state values (in the dimension of the dependent variable, e.g. ORF size), and the “strength” of the growth (e.g. the maximum slope of the curve between two steady states) are controlled by the parameters of the regression function. Once estimated, the parameter values can be used to infer genome size intervals for which a particular ORF region is in a steady state as well as critical genome and ORF sizes at the transition between two steady states. Since double-logistic regressions did not converge for the 5′- and 3′-UTRs, linear functions were used for these two genome regions instead.
Linear (null hypothesis) and splines (alternative hypothesis) regression models were compared using standard weighted F-statistics and a specially designed permutation test (see below). To exclude overfitting as the cause of support of the more complex models, we utilized a more sophisticated framework (LV-Test) for the comparison of non-nested regression models (linear vs. double-logistic and splines vs. double-logistic) as detailed in [112]. The test was further modified to include weighted residuals according to virus sequence weights that account for sequence dependence.
Since our null hypothesis (linear model) is at the boundaries of the parameter space, we developed a permutation test to further compare the linear and splines models. To this end, genome region sizes were transformed to proportions (region size divided by genome size), randomly permuted relative to genome sizes, and transformed back to absolute values. These transformations are compatible with the constraints of the null hypothesis and the requirement that region sizes have to sum to genome sizes. Weights were not permuted. The linear and splines models were fit to the permuted datasets and F-statistics were calculated as for the original dataset. The p-value of the test is the fraction of F-statistics of permuted datasets that are larger than the F of the original dataset. It was calculated using 1,000,000 permutations that were randomly sampled out of ∼1029 possible permutations.
Finally, we analyzed the contribution of each genome region to the total change in genome size under the three regression models. The contribution of each region according to a model was calculated as the ratio of change in region size to change in genome size (first derivative of the regression function) along the nidovirus genome size scale. These region-specific contributions were combined in a single plot for visualization purposes.
To conduct all statistical analyses and to visualize the results we used the R package [113].
Accession numbers of virus genomes utilized in the study are shown in Table S1.
|
10.1371/journal.pntd.0007103 | Reprogramming of Trypanosoma cruzi metabolism triggered by parasite interaction with the host cell extracellular matrix | Trypanosoma cruzi, the etiological agent of Chagas’ disease, affects 8 million people predominantly living in socioeconomic underdeveloped areas. T. cruzi trypomastigotes (Ty), the classical infective stage, interact with the extracellular matrix (ECM), an obligatory step before invasion of almost all mammalian cells in different tissues. Here we have characterized the proteome and phosphoproteome of T. cruzi trypomastigotes upon interaction with ECM (MTy) and the data are available via ProteomeXchange with identifier PXD010970. Proteins involved with metabolic processes (such as the glycolytic pathway), kinases, flagellum and microtubule related proteins, transport-associated proteins and RNA/DNA binding elements are highly represented in the pool of proteins modified by phosphorylation. Further, important metabolic switches triggered by this interaction with ECM were indicated by decreases in the phosphorylation of hexokinase, phosphofructokinase, fructose-2,6-bisphosphatase, phosphoglucomutase, phosphoglycerate kinase in MTy. Concomitantly, a decrease in the pyruvate and lactate and an increase of glucose and succinate contents were detected by GC-MS. These observations led us to focus on the changes in the glycolytic pathway upon binding of the parasite to the ECM. Inhibition of hexokinase, pyruvate kinase and lactate dehydrogenase activities in MTy were observed and this correlated with the phosphorylation levels of the respective enzymes. Putative kinases involved in protein phosphorylation altered upon parasite incubation with ECM were suggested by in silico analysis. Taken together, our results show that in addition to cytoskeletal changes and protease activation, a reprogramming of the trypomastigote metabolism is triggered by the interaction of the parasite with the ECM prior to cell invasion and differentiation into amastigotes, the multiplicative intracellular stage of T. cruzi in the vertebrate host.
| Adhesion of Trypanosoma cruzi to distinct elements of ECM involving different surface proteins from the infective stage of the parasite has been described. Despite the relevance of ECM for T. cruzi infection, the signaling pathways triggered in trypomastigotes upon interactions with ECM are less well understood. In previous work we demonstrated the dephosphorylation of proteins, such as α-tubulin, paraflagellar rod proteins and ERK 1/2 in trypomastigotes incubated with either laminin or fibronectin. Further, we described changes in the S-nitrosylation and nitration pattern of proteins from trypomastigote incubated with ECM. To expand our knowledge on ECM triggered parasite signaling we applied quantitative proteomic and phosphoproteomic studies to trypomastigotes incubated with ECM (MTy) compared to controls (Ty). Our results indicate relevant changes in total protein and phosphoprotein profiles in MTy. The kinases implicated in the modifications were suggested by bioinformatic analyses, as well as the number of modifications and the frequency of amino acids per peptide that have been modified. Proteins involved in metabolic processes, including enzymes from the glycolytic pathway, phosphatases and kinases were the most representative groups among the proteins modified by phosphorylation. Quantification of metabolites in MTy and Ty also indicated that glucose metabolism is impaired in trypomastigotes incubated with ECM. The significant inhibition of hexokinase, pyruvate kinase and lactate dehydrogenase activities in MTy associated with phosphorylation levels, strongly suggests that trypomastigotes reprogram their metabolism in response to interaction with the extracellular matrix, an obligatory step prior to host cell invasion.
| The protozoan T. cruzi, the etiological agent of Chagas’ disease, affects a wide range of mammalian hosts, including humans. It is estimated that 8 million people are infected and 25 million are at risk, most of them living in areas of poor socioeconomic development [1]. The cell cycle of T. cruzi involves an invertebrate vector (triatomine bugs) and a mammalian host, and well-defined developmental stages (epimastigotes, metacyclic trypomastigotes, amastigotes and blood trypomastigotes). T. cruzi trypomastigotes, the classical infective stage, invade almost all mammalian cell and tissue types, and invasion is an obligatory step in their life-cycle in mammals. An essential step immediately prior to the mammalian cell invasion is the interaction of the parasite with the surrounding extracellular matrix.
The extracellular matrix (ECM) is a highly dynamic non-cellular three-dimensional macromolecular network, which regulates different cellular functions, such as growth, differentiation or survival [cf. 2]. Its composition includes structural components (“matrisome”, [3]) and elements that can interact with or remodel the ECM [4]. Collagens, proteoglycans and glycoproteins (laminins, fibronectins, thrombospondins, tenascins, among others) constitute the main core of the ECM proteins. ECM-affiliated proteins (such as mucins, syndecans, plexins) and ECM-regulators (such as lysyl oxidases, sulfatases, extracellular kinases, proteases and secreted factors, such as TGFβ, cytokines) were classified as matrisome-associated proteins [4]. Cell-ECM interactions occur mainly by integrins present at the cell surface, which connect the extracellular signals and the intracellular response by the activation of specific signaling pathways [5, 6] and dysregulation of the ECM is associated with the development of several pathological conditions [4,7].
Adhesion of Trypanosoma cruzi and other parasites to distinct elements of ECM has been described to involve different surface proteins from the infective stage of the parasite, of which the gp85/transialidase family plays an essential role (rev. [8]). T. cruzi binds to collagen [9], fibronectin [10–14], laminin [11,15, 16, 17], thrombospondin [11], [18], heparan sulfate [11, 12, 14, 19], galectin-3 [20, 21], as well as TGF-β [22]. Also, remodeling of ECM was observed during infection, with modifications in collagen and fibronectin content, as well as reorganization of laminin [13], at least partially due to T. cruzi proteolytic enzymes [23–25].
Despite the relevance of ECM for T. cruzi infection, the signaling pathways triggered by the trypomastigote-ECM interaction are less well known. Recently, we demonstrated a decrease in S-nitrosylation and nitration in the majority of the trypomastigote proteins when parasites are incubated with ECM [26]. In addition, dephosphorylation of proteins, such as α-tubulin, paraflagellar rod proteins (PFR or PAR), as well as ERK 1/2 was observed in trypomastigotes incubated with either laminin or fibronectin [27], although these do not reflect the entirety of possible interactions between parasite and ECM. To have a better understanding of the process, quantitative proteomic and phosphoproteomic approaches were employed to analyze changes in T. cruzi trypomastigote proteins when the parasites are incubated with ECM. Herein we show important changes in the T. cruzi trypomastigotes proteome, as well as in protein phosphorylation levels upon interaction with ECM. Proteins involved with metabolic processes, phosphatases, kinases and RNA/DNA binding elements were highly represented among the proteins modified by phosphorylation. In particular, a decrease of the glycolytic pathway was suggested by metabolite quantification and measurement of hexokinase, pyruvate kinase and lactate dehydrogenase activities, suggesting an extensive metabolic adaptation of trypomastigotes prior to host cell invasion. Taken together, our data show that not only structural adaptations but also important metabolic changes occur upon parasite interaction with ECM. Understanding these changes may further elucidate the adaptive mechanisms involved in parasite-host interaction.
Trypomastigotes (ECM-treated for 120 min or control) were fixed in 2% paraformaldehyde for 15 minutes at room temperature, pelleted by centrifugation (4,000 x g for 5 minutes), washed twice in PBS, resuspended in PBS, added to a coverslip and dried at room temperature. After permeabilization of the parasites with PBS containing 1% BSA and 0.1% Triton X-100 for one hour at 37°C, anti-phosphoserine, anti-phosphothreonine, anti-phosphotyrosine (Invitrogen–dilution 1:200 for each antibody), anti-PAR monoclonal antibody (1:200) or anti–TcHexokinase (kindly provided by Dr. Ana Cáceres, Universidad de Los Andes, Venezuela) were added and incubated for 1 h at room temperature. After three washes with PBS containing 0.1% Triton X-100, the correspondent secondary antibodies were added (anti-rabbit or anti-mouse-Alexa 555 conjugated (1: 5000); followed by one hour incubation at 37°C. After three washes in PBS-0.1%-Triton X-100, the coverslips were faced under a solution containing 50% glycerol, 50% milliQ H2O 2 mM sodium azide, and 20 μg/mL of 4',6-diamidino-2-phenylindole, dilactate (DAPI-Invitrogen). The images were taken on an ExiBlue™ camera (Qimaging®) coupled to a Nikon Eclipse E 600 optical microscope and deconvoluted using the software Huygens Essential (Scientific Volume Imaging).
Frozen pellets of trypomastigotes (1 x 109 MTy or Ty) were resuspended in 1 mL of Lysis Buffer (30 mM Tris-HCl, pH 7.6, 1 mM EDTA, 0.1% Triton e 0.25 M sucrose, containing phosphatase and protease inhibitors, as described above, disrupted by ultrasonic for 4 x 10 s (frequency of 40%, Thomas GEX 600 apparatus). After centrifugation (10 000 x g, 15 min), the supernatant was separated and employed to measure the enzymatic activities (Hexokinase, Pyruvate kinase and Lactate dehydrogenase). In all cases, the amount of NADH / NADPH was measured spectrophotometrically at 340 nm and its concentration calculated (extinction coefficient = 6.220 M-1 cm-1) [32]. Three independent biological samples were employed. Due to the possible presence of ECM proteins in the MTy samples, the enzymatic activities were expressed by 1x108 parasites. The number of parasites in each experimental point was also estimated by a calibration curve using the amount of paraflagellar rod protein in the Western blotting (anti-PAR monoclonal antibody 1:2000).
Previous studies demonstrated that there is a significant decrease in protein phosphorylation upon parasite interaction with fibronectin or laminin, both components of the ECM. Thus we decided to investigate possible changes in protein phosphorylation that could occur upon trypomastigote interaction with the entire ECM. The experiments outlined in the scheme (Fig 1A) were performed in order to analyze both the proteome and the phosphoproteome of the parasites incubated with ECM.
The proteome and phosphoproteome analyses identified 3,093 proteins and 7,880 phosphopeptides, respectively, with FDR values less than 1% and p scores less than e-7 for peptide identification (Fig 1B). Thirty six proteins (57%) and 212 phosphopeptides (67%) correspond to proteins with unknown function (hypothetical proteins), as described by others in different trypanosomes [27, 35–42], which is in accordance with the number of proteins (49.2%) with unknown functions predicted by the genome sequence of T. cruzi [43] (Fig 1B). Sixty-three non-unique proteins from the proteome data exhibited significant variations, with only 5 showing reduction and 58 showing an increase in their protein level (Fig 1B, S1 Fig, S1 Table). Seventeen proteins showed changes where MTy/Ty ≥1.5, of which nine were hypothetical, including the one with the greatest change (MTy/Ty ≥12.9). Among the proteins with increased expression, it is worth emphasizing members of the gp85/trans-sialidase (TS) family, involved in host cell infection by T. cruzi, (MTy/Ty = 2.2 for one member of group II and Mty/Ty = 4.5 for one member of group IV); small GTP-binding protein rab6 (Mty/Ty = 3.8); ribosomal RNA processing protein 6 (MTy/Ty = 1.98; the splicing factor 3a; and the 2Fe-2S iron-sulfur cluster binding domain containing protein (Mty/Ty = 1.77).
In the phosphoproteomic analysis, only phosphopeptides having MTy/Ty ratio below 0.8 or above 1.2, with p values less than 0.05, have been considered to evaluate the effect of ECM on trypomastigote protein phosphorylation. Among the 303 phosphopeptides selected by these criteria, 69 showed an increase and 234 a decrease in their phosphorylation levels. Of these, 91 (33%) are proteins with known function, of which 19 showed an increase and 72 a decrease in their phosphorylation levels (S2 Table; Fig 2A and 2B). Taken together the data indicate that adhesion of trypomastigotes to ECM overall leads to protein dephosphorylation, in agreement with previous observations with T. cruzi trypomastigotes incubated with ECM components, fibronectin and laminin [27].
Phosphorylation site analysis identified 371 differentially phosphorylated amino acid residues: 275 (74.1%) serine, 83 (22.4%) threonine and only 13 (3.5%) tyrosine residues, in accordance with the abundance of serine/threonine kinases found in T. cruzi [44], the fact that T. cruzi does not express receptor coupled tyrosine kinases but dual specificity kinases [44], as well as with the phosphoproteome data described for different stages of the parasite [42, 45, rev.46]. In silico analysis, as described below, suggested the involvement of different kinases (CAMK, TKL, CMGC, CK1, AGC and others).
To understand better the role of the proteins controlled by phosphorylation during the parasite response to the ECM, GO-terms enrichment analysis was performed. Additional information was obtained on the molecular function and/or sub-cellular localization of 29 proteins and 126 phosphopeptides previously labeled as hypothetical (unknown function) based on version 40 of the TriTrypDB. The data are shown in S1 Table and S2 Table. Most of proteins identified are related to structural function, pathogenicity, metabolism and protein phosphorylation. Both cytoplasm and axoneme seem to be the main localization of the identified phosphopeptides. Among the proteins identified, gp85/trans-sialidase (TS) family members, involved in infection of host-cells by T. cruzi, were enhanced in trypomastigotes incubated with ECM (MTy/Ty = 2.2 for one member of group II and MTy/Ty = 4.5 for one member of group IV). Small GTP-binding protein rab6 (MTy/Ty = 3.8), ribosomal RNA processing protein 6 (MTy/Ty = 1.98), splicing factor 3a and 2Fe-2S iron-sulfur cluster binding domain containing protein (MTy/Ty = 1.77) were also identified in the group of the proteins with an increased expression.
The following proteins are highly represented in the pool of proteins modified by phosphorylation (Fig 2B and S2 Table): proteins involved in metabolic processes (19 phosphopeptides); in phosphorylation/dephosphorylation (such as kinases and phosphatases, 20 phosphopeptides); structural proteins (such as flagellum and microtubule related proteins, 19 phosphopeptides); transport-associated proteins; and RNA/DNA binding elements. This suggests an extensive metabolic adaptation occurring in trypomastigotes prior to host-cell infection.
Of note, enzymes that participate in glucose metabolism, in addition to adenylosuccinate lyase (ADSL) and alanine aminotransferase (ALT) (Table 1; Fig 2C), lipid metabolism (3-oxo-5-alpha-steroid 4-dehydrogenase; putative (pseudogene) and ethanolamine phosphotransferase) (Fig 2B, S2 Table) are modified by phosphorylation.
Since a significant number of enzymes from the glucose metabolism were modified by phosphorylation, their role was further investigated.
Seven enzymes involved in carbohydrate metabolism were identified in the phosphoproteomic analysis: hexokinase (HK); 6-phosphofructo-2-kinase/fructose-2;6 bisphosphatase (PFK2); 6-phospho-1-fructokinase (PFK1)(pseudogene); phosphoglucomutase (PGM); pyruvate phosphate dikinase (PPDK); phosphoglycerate kinase (PGK) and NADH-dependent fumarate reductase (FRD), in addition to adenylosuccinate lyase and alanine aminotransferase (Fig 2C, Fig 3, Table 1, S2 Table). Except for NADH-dependent fumarate reductase (MTy/Ty = 1.5) and adenylosuccinate lyase (MTy/Ty = 2.2), all the others showed decrease in their phosphorylation levels when trypomastigotes were incubated with ECM (Table 1 and Fig 3). Most of these enzymes are localized in the glycosomes, a peroxisome-like organelle essential for trypanosomatids survival and characterized for containing most of the glycolytic/gluconeogenic pathways in kinetoplastids, in addition to enzymes of other metabolic pathways, such as the pentose phosphate pathway, beta-oxidation of fatty acids, and biosynthesis of pyrimidines (rev. [47–49]). Likewise, phosphorylation of enzymes involved in carbohydrate metabolism was described in the proteome and phosphoproteome of the glycosomes in T. brucei and Leishmania donovani [40, 41, 50].
Since the data suggested possible changes in the metabolism of T. cruzi, the metabolite content was analyzed by GC/MS to understand better the response of trypomastigotes upon adhesion to ECM.
GC-MS analysis allowed the identification of 21 metabolites with a significant variation (p < 0.05) in the MTy/Ty ratio (S5 Table), from which significant changes were found for carbohydrate, lipid and amino acid metabolites. Some of these metabolites are substrates or products of the enzymes modified by phosphorylation that are found, although not exclusively, inside the glycosomes (Fig 3A and 3B, Table 1). The following metabolites from the glycolytic pathway were modified in parasites upon incubation with ECM: increase in glucose/galactose (molecules indistinguishable in the GC-MS methodology) with ECM (MTy/Ty = 1.2); decrease in pyruvic acid (MTy/Ty = 0.2) and lactic acid (MTy/Ty = 0.34). Interestingly, metabolites derived from the glycolytic pathway branch and common to TCA cycle: succinic acid (MTy/Ty = 1.37), malic acid (MTy/Ty = 1.21) and fumaric acid (MTy/Ty = 1.15), were increased in parasites incubated with ECM. Succinate is considered the main source of reducing equivalents to the respiratory chain through the action of a NADH-dependent fumarate reductase and it is also (in addition to alanine) one of the main products excreted by trypanosomatids [48, 51], although lactate excretion by T. cruzi [52] may increase, depending on metabolic adaptations. These changes in metabolite level do not appear to correlate with changes in metabolic enzyme levels (S1 Table). Rather, they may reflect modulation of metabolic enzyme activity. Although less representative in the metabolite quantification, an increase in free amino acids in MTy (tyrosine, glycine or isoleucine, MTy/Ty ratio ≈1.4) and fatty acids (mainly palmitic acid, MTy/Ty ratio ≈1.4) may also indicate wider changes in the metabolism of MTy (S5 Table).
To analyze the potential role of metabolic enzyme phosphorylation in modulating the glycolytic pathway in MTy, enzymatic activities of hexokinase/glucokinase (HK/GK) and pyruvate kinase (PK) were determined, as well as for lactate dehydrogenase (LDH). Although PK and LDH were not detected in our phosphoproteomic analysis (S2 Table), these enzymes were included because significant depletion of pyruvate and lactate were observed in MTy (S5 Table). Further, as pointed out before, the changes in the metabolite levels seem to be independent of the relevant metabolic enzyme expression levels accordingly to the proteomic data (S1 Table).
The enzymes HK and GK catalyze the formation of glucose-6-phosphate, the first reaction of the glycolytic pathway and both lack the regulatory allosteric inhibition by glucose-6-phosphate, common to other organisms. Both are localized inside the glycosomes, of which HK presents the highest activity [53, 54]. A distinct HK, which phosphorylates glucose and fructose, as well as GK activity were also described in the cytosol [53]. The affinity of HK for glucose is higher than that of GK (Km values of 0.06 mM and 0.7 mM, respectively) in addition to the higher amounts of HK over GK in the parasite [55]. However, in the present work we could not separate the activities of both enzymes and, thus, they are collectively represented by HK/GK activities.
The three phosphorylated residues of T. cruzi hexokinase (S161, T153, T159) (Table 2, S1 Table, S2 Fig) were located in the same peptide identified by LC-MS/MS and presented similar changes in the phosphorylation level (MTy/Ty ratio = 0.76; 0.76 and 0.57, respectively). The identified phosphorylated peptide is localized in the catalytic domain or in the substrate-binding site of the enzyme, according to the alignment of T. cruzi-HK (TcCLB.508951.20), T. brucei (Tb927.10.2010) and the three isoforms of human-HK (P52790; P19367; P52789), performed by ClustalW platform (S2B Fig). HK/GK activities from MTy and Ty homogenates were measured spectrophotometrically in the presence of a coupling system containing glucose-6-phosphate dehydrogenase. HK/GK activity is clearly reduced (by approximately 46%) in MTy relative to Ty (Fig 4A,a). Previous treatment of Ty and MTy homogenates with alkaline phosphatase significantly reduced the activity (approximately 45% for Ty and 59% for MTy extracts, Fig 4A,c,d). Also, the activity was drastically reduced (approximately 70%) when the homogenate was previously treated at 56°C for 1 h. Since HK is inhibited by small phosphate molecules, such as PPi present in distinct organelles including glycosomes [53, 56, 57], the experiment was repeated with the parasite extract previously immunoprecipitated with anti-HK antibodies. Similar results have been obtained, confirming the relevance of phosphorylation for HK/GK activity (S4 Fig). Inhibition of the enzymatic activity by dephosphorylation is consistent with the accumulation of glucose detected in MTy. However, one cannot rule out other possibilities, such as changes in glucose transport in MTy, alterations of HK oligomerization, which is usually tetrameric [54] or somehow by contributing to the hysteretic and cooperative behavior of HK at low enzyme concentration described in T. cruzi epimastigotes [58]. The phosphorylation of HK could be by auto-phosphorylation [59] or by protein kinase(s). Of note, the same pattern of HK in MTy and Ty inside the glycosomes was shown by immunofluorescence using specific anti-HK antibodies (S2A Fig). These data also indicate that no significant changes in the number of glycosomes occur in MTy, in accordance with the literature, where approximately the same number was found in the different forms of T. cruzi, in contrast to the variability described during the life cycle of other species [60].
Pyruvate kinase (PK) was not detected in the phosphoproteome/proteome described herein, as pointed out above, but the low amount of pyruvate found in MTy led us to measure a corresponding enzymatic activity. In the cytosol, PK catalyzes the formation of pyruvate and ATP from phosphoenolpyruvate and ADP and, in the case of T. cruzi epimastigotes, PK is inhibited by millimolar concentrations of ATP and Pi and activated by micromolar concentrations of fructose 2,6-bisphosphate (rev. [48, 61]) by the tetrameric stabilization of PK in response to the effector binding [62].
As shown (Fig 4B,a), PK activity is strongly inhibited in trypomastigotes incubated with ECM (approximately 65%). Treatment of the enzyme with alkaline phosphatase increased its activity, mainly in MTy homogenates (25% Ty and 125% MTy), as shown in Fig 4B,c,d. The decrease in pyruvate content observed could be attributed to the inhibition of pyruvate kinase in the cytosol and/or pyruvate phosphate dikinase in the glycosome, which would lead to an increase of dicarboxylic acids from the glycolytic branch (succinate, fumarate and malate) in the glycosome, (cf. Fig 3 and S5 Table). However, higher consumption of pyruvate, for example by its conversion to acetyl-CoA inside the mitochondria or to lactate in the cytosol, cannot be ruled out.
Lactate dehydrogenase-like (LDH) activity was measured in the parasite homogenate due to the decrease observed in lactate content in MTy (S5 Table). In many organisms, a tetrameric form of the enzyme catalyzes the oxidation of lactate to pyruvate in the presence of NAD+ as hydrogen acceptor. LDH activity in T. cruzi epimastigotes was attributed to the isoenzyme I of ∂-hydroxyacid dehydrogenase localized inside the glycosomes and in the cytoplasm [63] and to an unknown protein in T. brucei [51], since a typical lactate dehydrogenase is absent from the genomes of trypanosomatids. The measurement of LDH activity showed approximately 33% reduction in MTy in comparison to trypomastigotes (Fig 4C,a), in agreement with the decrease of lactate detected in MTy. LDH was also inhibited by alkaline phosphatase treatment (approximately 65% for Ty and 41% for MTy extracts), reinforcing the role of phosphorylation in modulating LDH activity (Fig 4C,c,d).
Alanine, rather than lactate, is usually the main product of the reduction of pyruvate in T. cruzi, a reaction catalyzed by alanine aminotransferase inside the glycosomes (rev. [48, 49]). Lactate excretion by the parasite [52] may increase depending on metabolic adaptations, as described for the procyclic forms of T. brucei [51] and may explain the aforementioned results in trypomastigotes. Although no significant differences in alanine content between MTy and Ty were detected by GC/MS analysis, alanine aminotransferase is less phosphorylated in trypomastigotes incubated with ECM and may be responsible for the switch to the LDH reaction for NADH-reoxidation. Since T. cruzi possesses the enzyme repertoire for gluconeogenesis, this pathway may also be activated in MTy, resulting in higher consumption of pyruvate, lactate and glycerol, although no reserve polysaccharide was detected and gluconeogenesis has not been fully established in T. cruzi.
Our data suggest that incubation of trypomastigotes with ECM triggers metabolic adaptations in the parasites, and that phosphorylation, or more specifically protein dephosphorylation, may be involved in these processes. In spite of the relevance of the dephosphorylation and the high representative number of phosphatases in the genome and proteome of T. cruzi [64, 42], only two protein phosphatases with diminished phosphorylation levels in MTy were found: an endonuclease/exonuclease/phosphatase responsible for dephosphorylation of DNA sequences (TcCLB.504073.10) and PP1, a serine/threonine phosphatase (TcCLB.507757.50). Serine/threonine phosphatases are abundant in T. cruzi and constitute more than 50% of the 86 protein phosphatases in the genome. Protein phosphatase PP1 is dephosphorylated in MTy (MTy/Ty = 0.65), which would be expected to increase its enzymatic activity and consequently contribute to the dephosphorylation of proteins [65] (Table 2; S2 Table). Other protein phosphatases, such PP2A, PP2C or dual specific phosphatase were described in the proteome / phosphoproteome of T. cruzi by different groups or in the glycosomes of T. brucei [41] or L. mexicana [50], which might also contribute to the decrease of the phosphorylation level of the proteins.
To understand further the signaling pathways activated in the parasite upon contact with host ECM, the kinases capable of phosphorylate the peptides detected by LC-MS/MS were predicted by the GPS 2.1 software. Only the higher scores for each phospho-residue (S, T and Y) in the peptide were selected, totaling 378 putative kinases for the 303 phosphopeptides identified by LC-MS/MS (S3 and S4 Tables). Most of the identified sites correspond to modifications by serine/threonine kinases and the phosphorylation of tyrosine was attributed to the dual-specificity kinases, in agreement with the absence of conventional tyrosine kinases in the genome of trypanosomatids [44]. Of the 303 phosphopeptides analyzed, 78.5% and 21.1%, respectively, presented one or two phosphorylated sites and only one phosphopeptide showed three phosphorylated-sites (Fig 5A and 5B). The phosphorylated sites identified herein are predicted to be modified mainly by elements of the CMGC kinases superfamily (68), STE kinases (62), TKL kinase (61), AGC kinase (31), CK1 family (31) (Fig 5D, Tables 1 and 2, S3 and S4 Tables). Kinases that do not belong to any characterized family are grouped as “Others”.
Only the catalytic subunit of protein kinase A, a member of the AGC kinase superfamily showed in MTy a slightly increase in its phosphorylation level (MTy/Ty = 1.27). The phosphorylation of T197 of mouse PKA, located in the activation loop, which corresponds to T35 described herein for T. cruzi, increases PKA activity, indicating that upon interaction with ECM, PKA is activated [66, 67]. Proteins responsible for a plethora of biological phenomena in T. cruzi have been described as PKA substrates, such as kinases (type III PI3 kinase-Vps34, PI3 kinase, mitogen-activated extracellular signal-regulated kinase), cAMP-specific phosphodiesterase (PDEC2), putative ATPase, DNA excision repair protein, aquaporin, hexokinase and members of gp85/TS [68, 69]. The role of PKA during metacyclogenesis (differentiation of epimastigotes to metacyclic trypomastigotes [70]) or during the amastigogenesis (differentiation of trypomastigotes into amastigotes [42]) is well established, with the stimulation of adenylyl cyclase and increment in cAMP concentration during the process. Albeit the relevance of PKA in the physiology of the parasite, their specific role in trypomastigotes incubated with ECM was not determined.
In contrast to PKA, reduction in phosphorylation was detected in the majority of the kinases, for example in glycogen synthase kinase 3 (GSK3) and ERK1/2, presumably leading to a decrease of their activities in MTy: Y187 from GSK3 (MTy/Ty = 0.78) corresponds to human Y216 also located in the activation loop and whose phosphorylation is necessary for activity [71]; T190 and Y192 from mitogen-activated protein kinase (MTy/Ty = 0.75), corresponds to human T185/Y187, whose phosphorylation is necessary for ERK1/2 activation [72]. Dephosphorylation of ERK1/ 2 was also observed in the incubation of trypomastigotes with laminin or fibronectin [27].
Decrease in phosphorylation of other kinases was also found: protein kinase ck2 regulatory subunit (MTy/Ty = 0.76), inositol-related signaling kinases; inositol polyphosphate kinase-like protein (MTy/Ty = 0.67) and phosphatidylinositol-4-phosphate 5-kinase type II beta (MTy/Ty = 0.37) (Table 2). Whether this is a reflection of their activation status remains to be determined.
The phosphopeptides assigned to particular kinase families by the GPS 2.1 software (Fig 3D) were used for the construction of sequence logos, which correspond to sequence alignment with the central point as the likely phospho-amino acid residue (Fig 3C). For the AGC kinase families, "other" and CK1, no consensus was observed in the amino acid sequences surrounding the S/T residue. For the "Atypical", TK and STE families some amino acids were identified with a high level of conservation. The analysis of the sequence logos (Fig 3C) indicates some conservation relative to the sites already characterized for humans, such as the group of peptides phosphorylated by CMGC, where a proline next to the phosphorylation site was also identified; lysine and arginine near the phosphorylation site for AGC; residues of aspartate and glutamate for CK1 family substrates and the conservation of phenylalanine and proline residues near the phosphorylated site for the atypical kinases.
Incubation of T. cruzi trypomastigotes with the extracellular matrix results in important and more extensive changes than the ones previously described by the incubation of trypomastigotes with fibronectin or laminin [27]. Reduction in the phosphorylation level of proteins seems to be a general event in trypomastigotes incubated with ECM (Tables 1 and 2, S2 Table). Kinases, except for PKA, PP1 phosphatase and enzymes from the glycolytic pathway and probably the glycolytic branch exemplify these decreases (S2 and S3 Tables). Strikingly, correlating with the observed decrease in phosphorylation level of the enzymes, a significant inhibition of hexokinase, pyruvate kinase and lactate dehydrogenase-like were detected. (S2 and S5 Tables, Fig 4). These results, in association to the slightly increase of glucose and drastic reduction of pyruvate and lactate strongly suggest that ECM triggers important reduction in the glycolytic pathway in T. cruzi trypomastigotes. Although hexokinase is among the many substrates described for PKA [69], a possible correlation between these enzymes has not been explored herein. Interestingly, trans-sialidases, surface glycoproteins belonging to the T. cruzi gp85/trans-sialidase family, are also one of the substrates for PKA [69] and PKA activity and trans-sialidase expression has been associated with differentiation and invasion of host cells by T. cruzi [73]. A similar coincidence of increase in PKA activity and expression of two members of the gp85/trans-sialidase family were observed upon incubation of trypomastigotes with ECM (MTy/Ty = 2.2 and 4.5, S1 Table), perhaps preparing the parasites for an efficient invasion of the host cell. The possibility that members of the large gp85/ trans-sialidase family interact with different components of ECM to trigger all the modifications described here remains to be determined.
Taken together, the data presented herein suggest reprogramming of the metabolism of trypomastigotes triggered by their interaction with the extracellular matrix, an obligatory step before cell invasion and differentiation into amastigotes, the multiplicative stage of T. cruzi in the vertebrate host. The reduction in glycolytic enzyme activity in trypomastigotes by phosphorylation/dephosphorylation events seems to be part of this reprogramming, with the involvement of yet to be identified protein kinases and phosphatases.
|
10.1371/journal.pgen.1004914 | Structured Observations Reveal Slow HIV-1 CTL Escape | The existence of viral variants that escape from the selection pressures imposed by cytotoxic T-lymphocytes (CTLs) in HIV-1 infection is well documented, but it is unclear when they arise, with reported measures of the time to escape in individuals ranging from days to years. A study of participants enrolled in the SPARTAC (Short Pulse Anti-Retroviral Therapy at HIV Seroconversion) clinical trial allowed direct observation of the evolution of CTL escape variants in 125 adults with primary HIV-1 infection observed for up to three years. Patient HLA-type, longitudinal CD8+ T-cell responses measured by IFN-γ ELISpot and longitudinal HIV-1 gag, pol, and nef sequence data were used to study the timing and prevalence of CTL escape in the participants whilst untreated. Results showed that sequence variation within CTL epitopes at the first time point (within six months of the estimated date of seroconversion) was consistent with most mutations being transmitted in the infecting viral strain rather than with escape arising within the first few weeks of infection. Escape arose throughout the first three years of infection, but slowly and steadily. Approximately one third of patients did not drive any new escape in an HLA-restricted epitope in just under two years. Patients driving several escape mutations during these two years were rare and the median and modal numbers of new escape events in each patient were one and zero respectively. Survival analysis of time to escape found that possession of a protective HLA type significantly reduced time to first escape in a patient (p = 0.01), and epitopes escaped faster in the face of a measurable CD8+ ELISpot response (p = 0.001). However, even in an HLA matched host who mounted a measurable, specific, CD8+ response the average time before the targeted epitope evolved an escape mutation was longer than two years.
| The cytotoxic T-lymphocyte (CTL) arm of the immune response is thought to play a significant role in the control of HIV-1 infection. Mutations within the HIV-1 genome allow the virus to escape recognition by CTLs and so evade the immune response. These escape mutations have been well documented but observed waiting times to escape within an individual have ranged from days to years. Many studies describing CTL escape have taken a detailed look at a few patients. Our analysis is based on a cohort of 125 clinical trial participants with immunologic and viral sequence data taken at regular longitudinal time points within the first few years of infection. Results suggested that the majority of CTL-related mutations present early in infection had been transmitted in the infecting viral strain as opposed to arising in the new host due to selection pressure imposed by CTLs. Whilst the prevalence of CTL escape mutations in the dataset was high, the incidence of new escape was relatively low; around one third of patients did not drive an escape within the first two years. Patients possessing a ‘protective’ HLA genotype had a significantly shorter waiting time to first escape than those without.
| The HIV-1-specific cytotoxic T-lymphocyte (CTL) response begins as early as 2 to 3 weeks after infection [1], and there is evidence to suggest it may play an important role in the early control of viraemia [2, 3]. The onset of the response coincides with the decline of viral load. Certain host HLA genotypes (and therefore certain specific responses) have been found to be significantly associated with delayed progression to AIDS [4–6]. In addition to this, some studies based on the depletion of CD8+ T-cells in SIV-infected macaques have shown that the CTL response makes a crucial contribution to viral control in this animal model of AIDS [7]. However, others have questioned the role of CTL in controlling productive infection, both in the SIV-macaque model and in humans [8, 9].
Many studies have demonstrated the capability of the HIV-1-specific CTL response to select for viral variants that escape recognition by CD8+ T-cells or prevent antigen presentation by HLA class I molecules and so evade the immune response [10–14]. Indeed, work done to quantify the effects of the various selective forces acting on HIV has found that 53% of non-env mutations that rise to fixation in the first few years represent viral adaptation to CD8+ T-cell responses [15] whilst a detailed deep sequencing study of one patient found that the majority of early mutations were CTL related [16].
CTL escape has been observed to occur throughout the different stages of HIV-1 infection, from soon after seroconversion [13, 17–20] to many years into chronic infection [11, 21, 22]. Some epitopes consistently escape earlier and more frequently than others [23–26] but time to escape also varies considerably between patients and is likely to be influenced by factors such as strength and quality of the CD8+ T-cell response as well as pre-existence of compensatory mutations in sequences surrounding epitopes [11, 21, 27]. Escape pathways are complex; in many cases low frequency variants arise early, replacing the transmitted sequence, but it is not until some time later that one escape mutant, often containing multiple amino acid mutations, emerges and rises to fixation [2, 28–30]. In the absence of selective pressure from CTLs, reversion of escape mutations back to wild type has been documented [20, 31].
In this study we analysed the timing and extent of CTL escape in 125 patients across the HIV-1 gag, pol, nef and env genes. We used longitudinal viral sequence data covering the first three years of infection to count escape events in each of 46 epitopes in all available patient samples, and only considered periods when patients were off therapy. Approximately one third of patients drove escape in an epitope for which they were HLA-matched within their first year of infection and escape continued to be seen into the second year. However an estimated 33% of patients had still not driven any escape in their HLA-restricted epitopes across gag, pol or nef by the end of their second year off therapy. In looking at whether escape was clustered in patients, we found little evidence to suggest that this was the case. We also observed that time to escape for individual epitopes was faster in HLA-matched patients and in patients who were able to mount a measurable CD8+ T-cell response, and that patients with protective HLA alleles were more likely to have their first escape sooner.
The cohort consisted of 125 treatment-naive, HIV-1 subtype B infected adults who were recruited soon after the estimated date of seroconversion (median 11 weeks; range 1–20 weeks) as part of the SPARTAC clinical trial and followed for up to three years (see Methods, full details in [32], patient characteristics shown in Table 1). Participants were randomised into one of three arms, which determined whether they received no treatment, 12 weeks antiretroviral therapy (ART) or 48 weeks ART. Consensus viral sequences for four immunogenic genes (gag, pol, env and nef), and IFN-γ CD8+ T-cell ELISpot data were collected longitudinally. Participants were followed for up to three years whilst off ART (median 109 weeks, IQR 85–145 weeks); some participants initiated long-term ART before the end of the three years, upon reaching the trial clinical endpoint. The amount of data available at each time point is summarised in S1 Table.
A CTL escape mutation was defined as any amino acid variation in previously identified (and phenotypically confirmed) escape sites within optimal defined epitopes in gag, pol, env or nef (see Methods for more details). All analyses were also run using alternative, more lenient definitions of escape. Those additional analyses, which are shown in S7–S12 Figs., confirm that the results presented here are not dependent on this definition of escape. In addition to drawing on a large body of collated literature to form this definition we also validated it by analysing the strength of patients’ CD8+ T-cell responses to autologous epitope variants (see Methods and S2 Table).
At baseline (the first trial study visit) the data comprised a total of 3831 epitope sequences from 122 patients in which we were able to look for variation within known escape sites. The range of times from seroconversion amongst these patients matched that of the whole cohort. Of these, 1235 contained escape mutations. However, this escape at baseline was, for almost all epitopes, equally prevalent in HLA-matched and HLA-mismatched hosts (Fig. 1A). This pattern is more consistent with variation that has been transmitted than variation that has been driven by within-host HLA-restricted immunity prior to the first sampling. In contrast, after one year, four epitopes derived from gag, pol and nef had markedly higher escape prevalence in HLA-matched hosts than in mismatched hosts (Fig. 1B). Further analysis (see Methods for details) confirmed that the distribution of points in Fig. 1B (at one year) is significantly different from what would be expected under the null hypothesis that escape is randomly distributed between matched and mismatched hosts, where as the distribution at baseline is not.
Of those epitopes sequenced, 2297 were initially wild-type and had at least one later sequence for comparison, and hence were candidates to allow the observation of incident escape (note that in a few cases this initial time point was after baseline due to sequence data being unavailable). Due to the high variability of the env gene, the vast majority of epitopes in env (all 32 HLA-matched patient-epitope pairs and 215/316 HLA-mismatched patient-epitope pairs) were already escaped at baseline, and hence whilst these were included in our analyses we were not able to comprehensively study incident escape in env. In the 2297 candidate epitopes, just 108 ‘escape’ events were observed in patients studied while off ART and only 37 of those were in epitopes in an HLA-matched patient. To see if there was evidence for a separate group of particularly fast escaping patients we compared the distribution of the number of escape events per host (in their HLA-restricted epitopes) with a random distribution, calculated using a Poisson process, and found no significant difference between the distributions (Fig. 2A). In fact this Poisson process in which escape was driven at an average rate of 0.0018 per week across all HLA-matched patient-epitope pairs, described the distribution of incident escape across patients well. Since many patients had missing data, particularly at later time points (S1 Table) we used this average rate to construct a similar expected distribution supposing all patients had had a full set of sequence data for the first two years (whilst untreated) (Fig. 2B). Even after accounting for the missing data in this way, only 12% of patients would have driven more than two escape variants in HLA-restricted epitopes and the median and modal numbers of incident escape events were one and zero respectively.
This analysis is meaningful if patients with missing data would have had similar rates of escape to those observed for a full two years whilst untreated. We believe this to be a valid assumption and assess it in further detail later.
Time to escape can be described either for each patient in a population (how long does it take until a viral variant with a new escape mutation is detectable in a patient?) or for each epitope within each patient (how long does it take until an escape mutation appears in an initially wild-type epitope?). By considering the first such escape event (in patients or in epitopes) the well-developed statistical tools of survival analysis become applicable. Since only 9% of patients have more than one incident escape event in an epitope for which they are HLA-matched (Fig. 2A) very little information is lost in such analysis.
In assessing the suitability of these analyses we also needed to address the question of whether the time at which patients are censored from the survival analysis (which occurs when they have no further sequence data) is independent of the rate at which they drive escape mutations. For some patients the fact that we were unable to obtain sequence data at later time points was a direct consequence of the fact that they had reached the trial primary endpoint (a CD4 cell count < 350 cells /μ l or commencing long-term ART) within the time frame of our observation period. Hence if, as some might hypothesise, escape events can drive clinical failure, it is possible that censored patients would have high rates of escape that we’d fail to observe.
We looked in more detail at those patients who reached the trial endpoint (or clinically ‘failed’) within our three-year observation period. An incident escape was observed in 6/20 patients for whom we had no sequence data after clinical failure compared to 7/24 patients who did have sequence data after clinical failure. In the 7 patients in the latter group who drove at least one escape event, there were just two cases where an escape mutation was first seen at the time point after clinical failure (see S1 Text and S2 Fig. for more details). Therefore the extent to which escape is linked to censoring time is small, and it is fair to assume that the (unobserved) rate of escape in patients who are censored from the analysis is the same as that of those who continue to be observed.
Fig. 3 shows time to first escape for patients whilst they were off ART. Only patients with at least two time points of sequence data for both nef and gag (the genes for which the most data were available) were used to avoid underestimating the occurrence of escape due to missing data. This left 65 patients for this analysis. Note that this number includes both treated and untreated patients (range of previous ART duration 2–12 months). We also considered these patient groups separately (S4–S6 Figs.) and this did not change our conclusions. Multivariate Cox models for time to escape confirmed that the impact of early limited treatment on CTL escape was negligible (S1 Text).
The Kaplan-Meier plot for time to first escape in an HLA-restricted epitope shows that there is large variation between patients in time to first escape, with some patients who received: 12 weeks of therapy escaping before the 24 week time point (so that the midpoint approximation gives the therapy-adjusted time to escape as around 6 weeks, assuming no viral evolution on ART) whilst an estimated 33% of patients did not drive any new escape in their HLA-restricted epitopes in 92 weeks off ART (95% C.I. (16,70)%). Note also that the slope of the curve is close to constant; there is no indication of a propensity for many incident escape events to arise early during the observation period and fewer later.
Although there was no evidence for a separate group of patients with a high frequency of escape (Fig. 2), there was evidence that some patients had a tendency to drive escape mutations sooner than others. Fig. 3B shows a Kaplan-Meier plot for time to first HLA-matched escape event for patients split according to whether or not they have one of the more protective HLA types. First escape is earlier in patients with a ‘protective’ HLA allele than in those without.
Fig. 4 records Kaplan-Meier survival curves for wild-type epitopes. This analysis included all patient epitope pairs which were observed whilst the patient was off ART (i.e. in which it would have been possible to observe escape), giving a total of 46 epitopes and 114 patients. Here ‘failure’ is the appearance of an escape mutation in an epitope. The characteristic time to escape for a wild-type epitope is of the order of months or years. It is faster in HLA-matched hosts (Fig. 4A), and amongst epitopes in HLA-matched hosts it is faster if there is a detectable CD8+ T-cell ELISpot response to that epitope in that host (Fig. 4B). Yet even in a host with the epitope’s restricting HLA and a detectable CD8+ T-cell response more than half of wild-type epitopes still show no escape after two years’ observation (Fig. 4B, survival = 0.69 between 84 and 115 weeks).
The literature on CTL escape in HIV-1 infection is dominated by studies of small numbers of individuals [17, 18, 33–35]. Several publications describe single patients in whom many CTL escape mutations arise and recent papers have described very active, rapid outgrowth of many escape mutations in a small number of patients studied intensively in the first weeks of infection [2, 19]. It has been known for some time that escape mutations that arise in early infection have a faster fixation rate than those that arise during chronic infection [10, 36]. These observations lead to a natural set of questions about the timing of CTL escape in HIV-1 infection. What is a normal number of escape events observable in a patient during the first few years of infection? Is escape very concentrated amongst a subset of patients so that patients fall into two groups—one with lots of escape and one with little? Are escape events also focussed in time so that much escape occurs early in infection and very little later on?
To address questions like this properly it is necessary to follow a large number of patients through time, so that those in whom nothing happens are given proper attention. This study of 125 patients in the first few years of untreated infection from a large randomised controlled trial does just that. It finds that the incidence of escape is very low; after two years of observation an estimated 33% of patients had not seen a first escape incident in gag, pol or nef. The observed frequency of incident escape across all epitopes in HLA-matched hosts translated to an estimated 12% of patients driving more than 2 escape mutations within 2 years (Fig. 2B). The inevitable conclusion is that patients with many escape events are unusual and, further, there was no evidence for a separate group of patients with a very high number of escape events (Fig. 2A). Equally, the distribution of escape through time showed no evidence for much escape early on and little later; the Kaplan Meier survival curves for the proportion of patients without an escape, and the proportion of epitopes without escape both have constant slopes (Figs. 3 & 4).
It has long been known that some HLA types confer a survival advantage upon their host. The relationship between the restricting HLA type and the propensity for an epitope to escape has been contentious. We previously [37] analysed the prevalence of escape in 84 chronically infected patients and concluded that protective T-cell responses were associated with more frequent viral escape, hypothesising that this result was due to a high fitness cost of mutations in epitopes restricted by protective alleles. However, Asquith [38], in a study that involved measuring the entropy of epitopes across sequences in the LANL HIV database, found that epitope entropy was significantly positively correlated with relative hazard of the restricting HLA allele, concluding that protective HLA alleles restricted epitopes that escape less frequently than average. In this cohort patients with protective HLA alleles escape significantly sooner (Fig. 3B), and epitopes restricted by protective HLAs escape faster (although not significantly so) than others (S3 Fig.). Further, all patients with escape in two or more HLA restricted epitopes had at least one of our defined list of protective HLAs. This large cohort study thus supports our previous finding that protective HLAs are associated with more frequent escape.
At first glance it seems counterintuitive that patients with some HLA types escape sooner and more frequently than others but that there was no evidence for a separate group of patients with much escape. However this apparent conflict is easily resolved; it is likely that patients with protective HLAs drive escape more frequently but do not have so much escape as to form a distinct subset of patients. They just lie at the top end of the distribution.
Why is escape in this cohort so slow? Is it because we missed very early escape events that occurred before patients were recruited? Is it that escape is concentrated in a small number of epitopes in which the process of escape is already largely complete? Or is it that escape is rare because responses are rare? Finally, is the definition of escape used here too restrictive?
Liu et al [19] have documented the presence of minor epitope variants within days of seroconversion, rising to fixation quickly in the first few weeks. However our results, presented in Fig. 1A, suggest that there was little escape happening this early on. Since reversion of escape mutations in mismatched hosts is relatively slow [23] we used the proportion of escape in HLA-mismatched hosts at first time point as an estimate of the prevalence of transmitted escape for each epitope, and hence by comparison with the proportion of escape in HLA-matched hosts obtained a quantitative estimate for the frequency of escape before the first time point. As an example, for TST (an epitope that is known to escape particularly early [39]), at the first time point 45% of HLA-mismatched patients had an escaped form of the epitope. We therefore assume that 45% of the 7 HLA-matched patients would have been infected with an escaped epitope. Hence, with reference to S5 Table, the estimated number of escape events in TST that have arisen since baseline is 6–7 × 0.45 = 2.9. Summing over all epitopes, these calculations gave the estimated ‘survival’ of wild-type epitopes in the weeks between seroconversion and first available sequence data (an average of 18.5 weeks) as 98.9% compared to 97.1% in the first 12 weeks of observation in which patients were off ART (see S1 Text for further details). Hence, whilst we have missed a handful of escape events (: 6 in total), these data do not support the idea that the majority of escape happens very early, within the first few weeks or months of infection.
The high prevalence of escape at baseline might suggest that the incidence of escape is low because many of the epitopes that are prone to acquire escape mutations already have them transmitted at baseline. This was true of the env epitopes we studied, which is unsurprising given the various and strong selection pressured on env, and a few others such as Gag TST. However, whilst there were epitopes in which the process of escape was complete even at baseline, this was not the general rule; Fig. 4 shows that more than 70% of HLA-matched patient-epitope pairs remain unescaped throughout the three-year observation period. Nor is it the case that all of the fastest escaping epitopes end up escaping in all HLA-matched patients. In these data only two epitopes (KYK and TAF) escape in all HLA-matched patients within the observation period.
The fastest rate of escape was seen within epitopes against which a positive CD8+ T-cell response had been measured (a group comprising: 30% of HLA-matched host-epitope pairs). Even here the incidence of escape was low (Fig. 4B). Hence we cannot conclude that it was a lack of strong CD8+ T-cell responses in the hosts to the set of epitopes we studied that led to few escape events being observed. However, as only IFN-γ production was measured to indicate the strength of the response we cannot tell whether or not the responses measured were qualitatively good in other ways.
Lastly, the definition of escape that we have used may have prevented some true escape from being classified as such. We were able expand the definition in two ways: by considering amino acid mutations that occur outside of known escape sites (but still within the epitope), and by extending the list of epitopes beyond the set of optimal defined epitopes (see S1 Text and S6 Table). Repeated analyses given these less stringent definitions are presented in S7–S12 Figs. Both of these amendments reduce the time to first escape in a patient so that an estimated: 10% (down from: 30%) of patients have driven no escape in HLA-restricted epitopes within the first two years. Changing the definition does not change the conclusion that the majority of escape at baseline is transmitted, or that escape mutations rise to fixation with a roughly uniform frequency across the observation period. However, extending the list of epitopes diminishes the association between HLA relative hazard and time to escape, suggesting that the strength of this association is sensitive to the set of epitopes considered.
We acknowledge that even this extended epitope list is unlikely to be complete, and that the Los Alamos ‘A-list’ will evolve to include new peptides and HLA-restrictions [40, 41]. Indeed, when using the extended list we observe 144 vs. 69 incident ‘escapes’ in HLA-mismatched vs.-matched patient-epitope pairs (S12 Fig.). Many of these ‘escapes’ in supposedly unrestricted epitopes cannot be accounted for by overlap with another, HLA-restricted epitope. Some of these may be indicative of immune escape driven by other components of the immune response such as NK cells, CD4+ T-cells or the antibody response. But understanding the impact of these alternative selective pressures is not necessary for assessing the concern that our methods are blind to CTL escape occurring in previously unidentified epitopes. We have previously shown that HLA-association studies favour the detection of rapid escape mutations. Hence any escape in unknown or poorly characterised epitopes that we have not looked in is likely to be slower than the time frame of this study and consequently not affect our conclusions [42].
Our definition of escape was also based on consensus viral sequence data; hence we were unable to detect the presence of any low frequency variants unlike with deep sequencing methods [2, 17]. However, variants that never rise to a majority in the population are of limited interest since they lack biological significance in terms of being unlikely to effect viral dynamics or the course of the disease in the long term. Variants that do rise to greater than 50% prevalence would be detectable with our methods. Therefore our use of consensus viral sequence is not problematic.
The final aspect of our definition of escape to address is the use of HXB2 as our comparison sequence. Although there is no ‘gold standard’ comparator for this analysis, there were only four epitopes in which HXB2 differed from the ‘A list’ optimal at a documented escape site. We therefore re-ran the analysis re-classifying these four epitopes and found no significant difference in the distribution of baseline escape prevalence between HLA-matched and-mismatched hosts. We did identify 4 new incident escapes (3 in EVK, 1 in RPN). Although the EVK escape variant is only poorly defined, its inclusion would mean that an estimated 25% of patients had not driven an escape by the end of 2 years off-therapy (as opposed to 33%).
In conclusion, this study of a cohort from a randomised controlled trial, the proper design for assessing rates of escape across a population, reveals that escape is on average both less frequent and less concentrated in the first few months of infection than would be expected from recent intense studies of much smaller numbers of acutely infected individuals.
The SPARTAC trial was approved by the following authorities: Medicines and Healthcare products Regulatory Agency (UK), Ministry of Health (Brazil), Irish Medicines Board (Ireland), Medicines Control Council (South Africa), and The Uganda National Council for Science and Technology (Uganda). It was also approved by the following ethics committees in the participating countries: Central London Research Ethics Committee (UK), Hospital Universitrio Clementino Fraga Filho Ethics in Research Committee (Brazil), Clinical Research and Ethics Committee of Hospital Clinic in the province of Barcelona, Spain, The Adelaide and Meath Hospital Research Ethics Committee (Ireland), University of Witwatersrand Human Research Ethics Committee, University of Kwazulu-Natal Research Ethics Committee and University of Cape Town Research Ethics Committee (South Africa), Uganda Virus Research Institute Science and ethics committee (Uganda), The Prince Charles Hospital Human Research Ethics Committee and St Vincent’s Hospital Human Research Ethics Committee (Australia), and the National Institute for Infectious Diseases Lazzaro Spallanzani, Institute Hospital and the Medical Research Ethics Committee, and the ethical committee Of the Central Foundation of San Raffaele, MonteTabor (Italy). All participants signed a written informed consent.
The design of the SPARTAC trial is reported elsewhere [32]. In brief, SPARTAC was an international open Randomised Controlled Trial enrolling adults with PHI within 6 months of a last negative, equivocal or incident HIV-1 test. All participants gave written informed consent. The trial was approved by research ethics committees in each country. Time of seroconversion was estimated as the midpoint of last negative/equivocal and first positive tests, or date of incident test. Participants were randomised to receive ART for 48 weeks (ART-48), 12 weeks (ART-12) or no therapy (standard of care, SOC). The primary endpoint was a composite of two events: if participants either reached a CD4 count of <350 cells/mm3 (> 3 months after randomisation and confirmed within 4 weeks) or initiated long-term ART. This provided an immunological surrogate of clinical progression, but also allowed inclusion of those participants who commenced ART at CD4 cell counts greater than cells/mm3.
The participants studied here consisted of a sub-group 125 HIV-1 subtype B positive individuals of whom 42, 41 and 42 were randomised to ART-12, ART-48 and SOC, respectively. Peripheral blood mononuclear cells (PBMCs) were collected at regular intervals during the treatment period and follow-up. The sub-group of 125 participants was based on those individuals infected with subtype B HIV-1 for whom both sequence and ELISpot data were available, predominantly determined by specimen availability.
Viral RNA was extracted from patient plasma (Qiagen Viral RNA extraction kit) and the HIV-1 pol, gag, nef and env genes were amplified separately using nested PCR reactions and primers described previously [20, 37, 43]. The PCR products were purified and sequenced using Big Dye dideoxy terminator chemistry (ABI). Sequences are therefore presented as the consensus from a bulk PCR product for each time-point. Sequence data were available for each patient at a subset of the following time points: 0, 16 or 24, 52 or 60, 108 and 156 weeks following recruitment onto the trial (see S1 Table for details).
Quantification of CD8+ T-cell response was performed by interferon gamma ELISpot analysis in all subtype B patients, using methods described elsewhere [44]. Responses were determined to 181 subtype B optimal peptides covering the Gag (n = 67), Pol (n = 47), Nef (n = 30) and Env (n = 37) proteins and 195 autologous variants synthesised followed sequence analysis. The optimal peptides tested were derived from the “best-defined ‘A list’ CTL epitope” database accessed from the Los Alamos HIV Immunology Website. Analyses were performed singly, but in duplicate for the negative control, and results expressed as spot forming units (SFU) per 106 cells. ELISpot responses were measured at 0, 24 and 60 weeks post-recruitment.
Patients’ HLA type was determined to the oligo-allelic level using Dynal RELITM Reverse Sequence-Specific Oligonucleotide kits for the HLA-A, -B and -C loci (Dynal Biotech). To obtain four-digit typing, Dynal Biotech Sequence-Specific priming kits were used, in conjunction with the Sequence-Specific Oligonucleotide type.
A CTL escape mutation was defined as any amino acid variation in previously identified (and phenotypically confirmed) escape sites within an ‘optimal’ defined epitope. Known escape sites were defined as within-epitope sites for which escape (E), calculated escape (CE), inferred escape (IE) or literature escape (LE) had been documented in the ‘CTL/CD8+ Epitope Variants and Escape Mutations’ HIV database table, as downloaded on 05/11/12 [45]. Any amino acid difference between autologous sequence and HXB2 in one of the sites under consideration was classed as escape. The list of ‘optimal’ defined epitopes was initially drawn from the set of epitopes for which ELISpot assays had been conducted (see above) though this was later extended (S1 Text and S6 Table). The 4-digit HLA-type restricting each epitope was determined using the ‘Best-defined’ epitope summary table on the Los Alamos HIV Immunology Database [45] downloaded on 01/03/2013. 2-digit HLA types were used only where 4-digit HLA types were not available. Only epitopes restricted by HLA-A and-B alleles were included since data on patient HLA-C alleles was not initially available. However a subsequent analysis found there were no incident escape mutations observed in any of these patients in epitopes restricted by their HLA-C alleles, so the exclusion of these epitopes is not problematic.
This intentionally stringent definition of escape was chosen to optimise the balance between including all relevant variation available in sequence data whilst excluding variation that is not escape. This definition does not include escape mutations that lie outside the epitope and disrupt proteasomal processing or antigen presentation pathways. It also excludes escape that occurs in poorly studied or unknown epitopes [46]. To complement the published data available and to check that mutations classed as ‘escape’ genuinely did result in diminished IFN-γ production by CD8+ T-cells, a comparison was made between the magnitudes of CD8+ T-cell ELISpot responses for the pre-defined optimal peptides with those for autologous epitope variants observed in the population at baseline. Variants for 7 Gag and 2 Pol epitopes were tested. In all cases except Gag EVD (where the optimal peptide differed from the HXB2 sequence) the major variant(s) elicited a diminished response. In 6/9 cases responses to all ‘escape’ variants were reduced by at least 90% (Table S2).
Firstly, the sequence data was subjected to quality control: A phylogenetic tree was constructed for all sequences of each gene in turn so that patients whose sequences did not cluster with each other could be identified and removed from the dataset if necessary. Due to difficulties in aligning all regions of all genes, epitopes were located using a string-matching search, allowing for 4 amino acid differences between the HXB2 epitope sequence and the patient sequence. Spurious matches were identified by hand and removed. Missing data resulted in escape being classed as ‘not determined’ if the HXB2 and viral sequence matched at all of the defined escape sites in the epitope except those where data was missing.
Permutation test on the distribution of escape between HLA-matched and-mismatched hosts. We wished to test the null hypothesis that escape is equally likely to be present in HLA-matched and-mismatched hosts against the alternative hypothesis that there is more escape in HLA-matched hosts. The statistic used for this was the mean (across all epitopes) of the Fisher’s exact test p-values for whether escape was independent of HLA-matching for that epitope. A permutation test was used to calculate the distribution of this mean under the null hypothesis; 2,000 random data sets were generated by permuting which patients the escape events occurred in for each epitope (i.e. there was no permutation between epitopes). The mean of the p-values for the observed data was then calculated and compared to the null distribution (S1 Fig.).
Poisson process for calculation of expected escape distributions. For each epitope in each patient in which an incident escape could have been observed, the total amount of time for which the patient was observed whilst off ART was calculated. Let E be the total number of escape events observed in the population, T be the total observation time (whilst off ART) across all epitopes in all patients and tp be the total observation time for all epitopes of a particular patient, p. Then, under the assumption that the average rate of incidence of escape in all patients is r = E/T, the probability of patient p driving i escape events is:
Pr
(
p
drives
i
escape events during observation
)
=
e
−
t
p
r
(
t
p
r)
i
i!
The expected number of patients with i incident escape events is then the sum over all patients of these probabilities. The expected distribution of escape across different epitopes is calculated similarly.
To account for missing data (for Fig. 2B), it was supposed that each patient had 108 weeks of data, during which time they were off ART. Where an epitope sequence was entirely missing (so that whether or not the epitope was initially wild-type could not be determined), the probability of it being wild-type at baseline was taken to be equal to the prevalence of escape at the first time point, as in S5 Table (this was necessary for a total of 203 HLA-matched patient-epitope pairs). We could then calculate tp = (#HLA-matched, initially wild-type epitopes)×108.
Note that this model does not impose the condition that only one escape is possible in each patient-epitope pair. We believe this to be a reasonable simplification given that the total number of escape events is small.
A X2 goodness-of-fit test with 4 degrees of freedom (i.e. categories for 0, 1, 2, 3, and > 4 escaped epitopes) was used to test the null hypothesis that the observed data followed the expected distribution.
Survival analysis. For the survival analysis shown in Figs. 3 and 4 only patient-epitope pairs in which it would have been possible to observe an escape were considered. This included all patient-epitope pairs that were wild-type in known escape sites at the first data point and had data available at a second data point. For each patient-epitope pair where an incident escape occurred, escape times were estimated as the midpoint between the last wild-type observation (or if that observation was during therapy, the therapy end time) and the first appearance of the escape mutant in the consensus sequence (or the clinical endpoint, whichever was earlier). Escape events occurring during treatment were discarded. Where no incident escape occurred, right-censoring started at the last sequence data time point or clinical endpoint, whichever was earlier. In Fig. 3, hazard ratios were calculated using a simple Cox Proportional Hazards model with a single predictor—the variable of interest. p-values relate to the Likelihood Ratio test comparing the one-predictor model with the null model. In Fig. 4, where observations from all patients and epitopes were pooled, hazard ratios and associated p-values were calculated using a mixed effects Cox model with a single predictor but also random intercept terms for patient and epitope ID. This was implemented with the package ‘coxme’ in R.
|
10.1371/journal.ppat.1004695 | Spatiotemporal Regulation of a Legionella pneumophila T4SS Substrate by the Metaeffector SidJ | Modulation of host cell function is vital for intracellular pathogens to survive and replicate within host cells. Most commonly, these pathogens utilize specialized secretion systems to inject substrates (also called effector proteins) that function as toxins within host cells. Since it would be detrimental for an intracellular pathogen to immediately kill its host cell, it is essential that secreted toxins be inactivated or degraded after they have served their purpose. The pathogen Legionella pneumophila represents an ideal system to study interactions between toxins as it survives within host cells for approximately a day and its Dot/Icm type IVB secretion system (T4SS) injects a vast number of toxins. Previously we reported that the Dot/Icm substrates SidE, SdeA, SdeB, and SdeC (known as the SidE family of effectors) are secreted into host cells, where they localize to the cytoplasmic face of the Legionella containing vacuole (LCV) in the early stages of infection. SidJ, another effector that is unrelated to the SidE family, is also encoded in the sdeC-sdeA locus. Interestingly, while over-expression of SidE family proteins in a wild type Legionella strain has no effect, we found that their over-expression in a ∆sidJ mutant completely inhibits intracellular growth of the strain. In addition, we found expression of SidE proteins is toxic in both yeast and mammalian HEK293 cells, but this toxicity can be suppressed by co-expression of SidJ, suggesting that SidJ may modulate the function of SidE family proteins. Finally, we were able to demonstrate both in vivo and in vitro that SidJ acts on SidE proteins to mediate their disappearance from the LCV, thereby preventing lethal intoxication of host cells. Based on these findings, we propose that SidJ acts as a metaeffector to control the activity of other Legionella effectors.
| A key attribute of many pathogens is their ability to survive and replicate within eukaryotic host cells. One such pathogen, Legionella pneumophila, is able to grow within macrophages in the lungs, thereby causing a form of pneumonia called Legionnaires’ Disease. L. pneumophila causes disease by translocating several hundred proteins into the host cell. These proteins are typically referred to as ‘‘effectors’’, as they function as toxins to alter normal host cell function. However, since L. pneumophila remains within the host cells for approximately one day, continual poisoning of the eukaryotic cells by the bacterial effectors will result in the premature death of the host cell, thus restricting the growth of the pathogen. Previously the L. pneumophila secreted protein LubX was described as a “metaeffector”, which has been defined as an effector that acts directly on another effector to modulate its function inside the host cell. LubX accomplishes this task by directing the degradation of another effector, SidH. Here we report a second L. pneumophila metaeffector, SidJ, acts in a similar manner to neutralize SidE family effectors by removing them from the intracellular compartment that contains the bacterium. This further establishes the concept of metaeffectors, which are likely to be critical to how Legionella and many other pathogens cause disease.
| Legionella pneumophila, the causative agent of Legionnaires' disease, is a facultative intracellular bacterial pathogen that can replicate within fresh water amoeba and mammalian alveolar macrophages [1–3]. L. pneumophila survives and replicates within host cells by inhibiting the host endocytic pathway and creating a novel replicative compartment designated as the Legionella containing vacuole (LCV) [4–7]. Alteration of host function is mediated by the injection of a large number of proteins into the host cell by the L. pneumophila Dot/Icm type IV secretion system (T4SS) [8–12]. However, inactivation of individual (or even combinations of) Dot/Icm substrates in genetically engineered mutant strains rarely has a strong effect on the intracellular growth of L. pneumophila, consistent with extensive functional redundancy between effectors [13–15].
One notable exception to this generalization is the L. pneumophila SuperΔP170 mutant, which exhibited a substantial growth defect in the amoebae Acanthamoeba castellanii [16]. The SuperΔP170 was constructed while studying a locus that encodes multiple Dot/Icm substrates [16] and consists of two deletions: the first removes five adjacent genes (sdeC, lpg2154, sidJ, sdeB, and sdeA) and the second deletes the unlinked gene sidE. Four of the encoded proteins, SidE, SdeC, SdeB and SdeA, share extensive homology with each other and are all ∼170 kDa in size, thus they have been referred to as “P170s” [16]. In addition, they are called the “SidE family”, as SidE was the founding member of this related group of proteins [17]. The SidE proteins are Dot/Icm substrates that are translocated into the host cell and reside on the cytoplasmic face of the LCV (Legionella containing vacuole) [16], although their molecular function is not known. As the intracellular growth defect of the SuperΔP170 mutant could be complemented by expression of just one SidE family protein, SdeA, it was proposed that the SidE-like proteins were functionally redundant and the other two genes, lpg2154 and sidJ, must be dispensable for growth within host cells [16]. However, subsequently it was shown that inactivation of sidJ alone conferred an intracellular growth defect on L. pneumophila [18], suggesting the situation is more complicated than initially perceived.
Consistent with this observation is the increasingly appreciated paradigm in pathogenesis that secreted effectors are often subjected to spatiotemporal regulation and that there can be a complex interplay between the functions of different effectors. For example, the Salmonella T3SS substrates SopE and SptP, which possess opposing biochemical activities, act at different stage of infection to first induce bacterial uptake and then to down-modulate this effect in order to prevent host cell death [19]. Similarly, the Legionella pneumophila Dot/Icm T4SS effectors SidM/DrrA and LepB exhibit opposing functions. SidM/DrrA recruits and activates Rab1 to mediate fusion of ER microsomes with the LCV (Legionella containing vacuole). At later points, LepB inactivates Rab1 resulting in the removal of the GTPase from the LCV [20–22]. A third example is represented by the L. pneumophila effectors SidH and LubX. SidH is a homolog of the effector SdhA, which is required to maintain the integrity of the LCV [23,24]. LubX is a member of the U-box family of E3 ubiquitin ligases and functions as a metaeffector to inactivate SidH by promoting its proteolysis [25].
Due to their genetic proximity, the surprising phenotype of the ΔsidJ mutant, and the existing precedents of complex interplay between other L. pneumophila secreted substrates, we hypothesized that there may be a connection between the SidE proteins and SidJ. To test this hypothesis, we examined the phenotypes of a strain lacking just the four SidE proteins (ΔsdeC ΔsdeB ΔsdeA ΔsidE) and of an individual ΔsidJ mutant and discovered that overexpression of an individual SidE family protein in the absence of SidJ is toxic to host cells. This result, and the experiments that followed, have led to a model wherein SidJ functions as a metaeffector to regulate the activity of the SidE family of toxins.
The sdeC-sdeA locus encodes SdeC, Lpg2154, SidJ, SdeB, and SdeA (Fig. 1A). A related protein, SidE, is encoded at a separate site on the chromosome. SidE, SdeC, SdeB, and SdeA are each ∼170 kDa in size, share greater than 40% identity to each other, and are substrates of the L. pneumophila Dot/Icm T4SS (S1 Fig.) [16,17]. The sdeC-sdeA locus also encodes another Dot/Icm substrate, SidJ, which has no homology to the SidE family (S1 Fig.). Although the L. pneumophila strain Philadelphia I encodes a homolog to SidJ, SdjA (Lpg2508), it was previously shown to be dispensable for virulence and therefore was not characterized further [18].
To examine the connection between SidJ and the SidE family of proteins, we constructed two additional mutant strains. One mutant lacked only sidJ and the other mutant contained deletions in each of the four sidE-related genes, which we refer to as the “CleanΔP170” mutant. We then compared the intracellular replication properties of the ΔsidJ, the CleanΔP170, and the original SuperΔP170 mutant [16], which does not express SidJ or any member of the SidE family. Growth was assayed by infecting the amoebae A. castellanii for various amounts of time, lysing the cells and determining the fold bacterial growth based on the number of CFUs (colony forming units). In this assay, a wild-type strain of L. pneumophila was able to replicate greater than 1000-fold in forty-four hours whereas a T4SS-deficient strain, Lp03, was unable to grow (Fig. 1B). In contrast, the CleanΔP170 strain exhibited a 100-fold growth defect similar to that previously observed for the SuperΔP170 mutant (Figs. 1B and S2). Likewise, a strain lacking sidJ exhibited a similar intermediate intracellular replication deficiency (Fig. 1C) [18]. The CleanΔP170 growth defect could be complemented by expression of just one SidE family protein, SdeA, consistent with the redundancy previously observed between members of this family (Fig. 1B) [16]. As expected, the replication defect of the CleanΔP170 mutant could not be complemented by expression of SidJ from a plasmid as this strain already makes SidJ (Fig. 1B). Similarly, the strain lacking sidJ could be complemented by expression of sidJ, but not by sdeA (Fig. 1C).
Taken together, these data suggest that expression of SidJ and at least one member of the SidE family is required for virulence of L. pneumophila within the environmental host A. castellanii. But this raised the question of how the original SuperΔP170 mutant strain, which is missing all four members of the SidE family and SidJ, could be fully complemented by expression of only SdeA (S2 Fig.) [16]. We hypothesized that this discordance might be related to expression levels from the sdeA expression plasmid. However, the sdeA complementing clone, pJB3556, did not synthesize aberrant amounts of SdeA and instead made a similar amount to what is normally expressed in Legionella (S3 Fig.).
Although pJB3556 expresses the appropriate level of SdeA in a cell, it is worth noting that this represents only a portion of the total amount of SidE family proteins expressed in a wild type cell. Therefore, we assayed the effect of expressing higher amounts of SdeA from a new complementing clone, SdeA OP (SdeA “over production”) (S3 Fig.). Expression of SdeA from this new construct had no effect on the growth of the wild type strain Lp02 (Fig. 1D) or the CleanΔP170 mutant (Fig. 1E). On the other hand, over-expression of SdeA remarkably completely inhibited the growth of the ΔsidJ mutant to levels similar to that of the T4SS-deficient dotA mutant (Fig. 1F). This extent of inhibition was shared with the SuperΔP170 mutant, which also does not express SidJ (S2 Fig.). Thus, inhibition of replication by SdeA over-production occurs only in strains lacking sidJ. Interestingly, this inhibitory effect appears to be specific to Legionella virulence, as over-expression of SdeA in the ΔsidJ mutant results in increased mis-targeting of Legionella to a late endocytic/lysosomal compartment (Fig. 1G). In summary, the absence of either the SidE family or SidJ results in diminished growth of L. pneumophila within amoebae. Furthermore, the partial growth defect of a ΔsidJ strain, but not of a wild type strain, can be exacerbated by the over-expression of SdeA, thus establishing a link between SidJ and the SidE family.
One possible connection between SidJ and SdeA is that they might modulate each other’s secretion into host cells. To test this hypothesis, we measured the export of the reporter proteins CyaA-SidJ and CyaA-SdeA in different genetic backgrounds. Successful export of Bordetella pertussis CyaA fusions into the host cell cytoplasm is measured by increased cAMP production, since this version of CyaA is activated when it is bound by eukaryotic calmodulin [26,27]. Expression of either CyaA-SidJ or CyaA-SdeA in the wild-type strain Lp02 generated a large increase in host cell cAMP as compared to expression in T4SS-deficient Lp03 cells, mock infected cells, or Lp02 expressing only CyaA (S4 Fig.). As previously observed [16,28], SdeA secretion was strongly dependent on the type IV adaptor IcmS whereas SidJ secretion was only partially dependent. However, export of SdeA was not affected by the absence of sidJ nor was SidJ secretion diminished in a strain that did not express any of the SidE family (S4 Fig.). Therefore, SidJ does not appear to regulate the secretion of SidE family members and SidE proteins do not affect the export of SidJ, suggesting that the molecular connection between SidJ and SidE family members likely occurs within host cells.
Since over-expression of SdeA is toxic to amoebae in the absence of SidJ, we chose to explore this phenomenon in more detail within the model eukaryote Saccharomyces cereviseae [29,30]. We began these yeast studies by expressing SdeA and SidJ under the control of the strong, regulated Pgal promoter. Yeast cells transformed with Pgal vectors carrying sdeA or sidJ were grown in the presence of glucose, diluted, and spotted onto selective media containing glucose (repressing conditions) or galactose (inducing conditions). Galactose-induced expression of either SdeA or SidJ was extremely toxic to yeast cells (Fig. 2A, rows 2–3) consistent with previous results [29,30]. To further evaluate this toxicity, we expressed each protein at lower levels using the constitutively expressed, weak promoter Pcyc. Expression of sidJ under the weaker promoter had only a subtle effect on yeast (Fig. 2A, row 5) whereas it was not possible to construct a yeast strain harboring Pcyc-sdeA, presumably due to SdeA-mediated toxicity. Strikingly, expression of low amounts of SidJ (Pcyc-sidJ) was able to partially suppress the toxicity caused by expression of high amounts of SdeA (Pgal-sdeA) (Fig. 2A, row 7).
We then attempted to recapitulate this result in mammalian cells. Transient transfection of HEK293 cells for 40 hours with mCherry-SdeA resulted in the protein localizing in dense foci in ∼90% of the cells (Fig. 2B and 2C). This result was specific to the SdeA fusion as expression of only mCherry resulted in diffuse, cytoplasmic staining. The longer mCherry-SdeA was expressed in cells, the more toxic it became eventually causing the cells to round up (S5 Fig.). In contrast, expression of YFP-SidJ was not toxic and the protein localized diffusely in the cytoplasm similar to YFP alone (Fig. 2B and 2C). Remarkably, the accumulation and toxicity caused by mCherry-SdeA expression was suppressed by co-transfection with YFP-SidJ resulting in dispersal of mCherry-SdeA into smaller foci (Fig. 2B and 2C).
To further examine SdeA toxicity in mammalian cells, mCherry-SdeA was expressed for a shorter amount of time (20 hours) and the cells were analyzed by fluorescence microscopy using markers to detect the endoplasmic reticulum, mitochondria, Golgi, endosome/lysosome, actin, and tubulin. Shorter expression of mCherry-SdeA only affected the Golgi resulting in its fragmentation (Fig. 2D-2E). In contrast, expression of mCherry alone, YFP alone, or YFP-SidJ had no effect on the Golgi. Similar to the redistribution of SdeA, YFP-SidJ was able to suppress the Golgi fragmentation caused by mCherry-SdeA (Fig. 2D and 2E). In summary, SdeA expression was toxic to both yeast and mammalian cells and SdeA-toxicity could be suppressed by co-expression of SidJ, implying that SidJ may regulate SdeA function.
A number of Dot/Icm substrates localize to the outside of the Legionella containing vacuole (LCV) at various stages of infection, including members of the SidE family [16,17,22,31]. For example, the substrate SidM/DrrA associates with the LCV early during infection but cannot be detected at later time points [22]. In contrast, the substrate LepB has limited association with the LCV at early points but displays increased co-localization over time. Similar to SidM/DrrA, SidE family proteins can be detected in proximity to the LCV early on but then disappear at later points during infection [16].
Based on these results, we hypothesized that SidJ may modulate SidE family association with the LCV as a means of regulating their activity. We began by confirming the observation that SidE family proteins can be detected on the LCV only at the initial stages of infection [16]. Bone marrow macrophages (BMMs) were infected for increasing amounts of time with the wild-type L. pneumophila strain Lp02, the T4SS-deficient strain Lp03, and the ΔsidJ mutant. The cells were fixed and stained with antibodies specific for three Dot/Icm substrates including the SidE family [16], SidM/DrrA, and a protein known to remain on the LCV, LidA [31]. As shown in Fig. 3, SidE staining could be detected on wild-type LCVs adjacent to the bacterial poles at 1-hour post infection. The detected SidE signal at 1 hour represented secreted protein, as it was not detected in Lp03-infected macrophages. As previously observed [16], SidE staining decreased as the infection proceeded resulting in less than 10% co-localization with the LCV after 4 hours (Fig. 3B). SidM/DrrA exhibited a similar pattern of localization over time (Fig. 3C) [22]. In contrast to SidE and SidM/DrrA, the secreted effector LidA was retained on the LCV throughout the infection (Fig. 3D), thus demonstrating that the disappearance of SidE and SidM/DrrA is not a general phenomenon.
To test our hypothesis that SidJ might mediate SidE family localization, we examined the intracellular position of the three Dot/Icm substrates when secreted from a ΔsidJ mutant. The absence of SidJ had no effect on the location of SidM/DrrA or LidA (Fig. 3). However, the localization of SidE was significantly altered in the ΔsidJ strain in a time-dependent manner resulting in retention of SidE on the LCV at later time points of infection (Fig. 3B). The disappearance of SidE proteins from the LCV in a wild-type infection could be due to their degradation and/or their dissociation from the L. pneumophila phagosome. To test these possibilities, we examined the presence of LCV-associated SidE proteins during an infection where the host cells were first treated with the proteasome inhibitor MG132. Interestingly, MG132 treatment resulted in a significant increase in the amount of retained SidE proteins at later time points (Fig. 3B), similar to that seen with the ΔsidJ mutant. In contrast, MG132 had no effect on SidM localization (Fig. 3C), thus demonstrating specificity. In summary, the disappearance of SidE proteins from the LCV at later points of infection is dependent on both SidJ and the proteasome.
Based on the MG132 result, it is possible that SidJ removes SidE family proteins from the LCV via induced proteolysis, similar to the action of the metaeffector LubX on the substrate SidH [25]. To test this theory, we developed a method using the detergent digitonin to measure the total amount of secreted SidE family proteins. Digitonin extracts proteins from mammalian membranes without disrupting bacterial membranes due to the absence of cholesterol in the latter membranes. We then proceeded to infect the human monocytic cell line U937 with wild-type Legionella, gently lysed the cells using a dounce homogenizer and removed unbroken host cells via low speed centrifugation. This generated a post-nuclear supernatant (PNS) fraction that was then separated by a high-speed centrifugation step into a pellet fraction, containing host organelles and the LCV, and a soluble cytoplasmic fraction.
In the absence of digitonin, the majority of SidE-like proteins co-localized in the pellet fraction with DotF, an inner membrane protein of the Dot/Icm T4SS that was used as a marker for the LCV (Fig. 4A). In contrast, SidJ could be detected in the soluble fraction without digitonin, indicating that the majority of the protein was secreted into the host cytoplasm and not retained on the LCV (Fig. 4A). Inclusion of digitonin in the reaction resulted in the solubilization of a large percentage of the SidE proteins, indicating they were secreted into the host cell. DotF could not be extracted and therefore was retained within the digitonin-resistant, bacterial cell wall (Fig. 4A).
Using this PNS/digitonin assay, we measured the levels of secreted SidE and SidJ proteins by Western analysis. Rather than observing decreased quantity of digitonin-soluble SidE proteins consistent with proteolysis, their amounts actually increased over the first 5 hours of the infection (Fig. 4B, WT). The levels of secreted SidJ and LidA were elevated in a similar fashion, although the amount of SidM/DrrA did not significantly increase over time (Fig. 4B and 4C, WT). In order to more carefully examine if any of the initial, secreted protein was degraded, we examined the levels of the proteins after inhibition of protein synthesis using the antibiotic chloramphenicol. The initial wave of secreted SidE, SidJ, and LidA remained fairly stable in the presence of chloramphenicol (Fig. 4, +Cm). In contrast, chloramphenicol-treatment did affect the levels of SidM/DrrA as the infection proceeded, suggesting that continual secretion of SidM/DrrA protein was necessary to maintain the levels of the protein over time (Fig. 4E, +Cm).
Treatment with the proteasome inhibitor MG132 resulted in a slight stabilization of SidM/DrrA consistent with the decreased amounts of SidM in the presence of chloramphenicol being due to proteolysis by the proteasome (Fig. 4E, +MG). In contrast, MG132 surprisingly did not alter the levels of SidE proteins (Fig. 4B, +MG), indicating they were not subject to proteasomal degradation. This result was unexpected based on our previous data showing MG132 treatment prevented the disappearance of SidE from the LCV (Fig. 3). Therefore, the situation must be more complex than initially anticipated. For example, there could be two populations of SidE proteins within the cell, a small portion on the LCV and a distinct population located elsewhere, and SidJ mediates the proteasomal-degradation of only LCV-associated SidE proteins.
To test this hypothesis, we separated the LCV from other cell organelles using discontinuous sucrose gradient analysis. Using this method, we were able to obtain fractions enriched for the cytoplasm (actin), endosomes/lysosomes (LAMP-1), mitochondria (PHB), Golgi (GM130), ER (calnexin), and the LCV (DotF) (Fig. 5A). We then examined the location of SidE proteins and SidJ at 1 hpi (hour post infection) and 3 hpi. Consistent with our immunofluoresence data in Fig. 3, we were able to detect SidE proteins that co-localized with the LCV at 1 hpi (Fig. 5B). However, we also observed a large amount of SidE proteins in fractions that contained host organelles, including endosomes/lysosomes, mitochondria and ER (Fig. 5). Strikingly, at 3-hours post infection there was a significant depletion of SidE proteins in the LCV fraction and instead there was an enrichment of these proteins in the organelle fractions distinct from the peak cytoplasmic fractions containing actin (Fig. 5A). In contrast to SidE proteins, SidJ was found in the cytoplasmic fractions at both 1 and 3-hours post infection (Fig. 5B). Taken together, our data suggests that the SidJ and proteasome-dependent removal of SidE proteins from the LCV is due to localized degradation of protein.
Although the removal of SidE proteins from the LCV requires SidJ, it is possible that this effect is indirect and perhaps due to maturation of the LCV over time. Therefore, we developed an assay to test if purified SidJ protein could remove SidE proteins from LCVs in vitro. This assay involved isolating post nuclear supernatants (PNSs) from bone marrow-derived macrophages (BMMs) infected for 1 hour or 4 hours with the L. pneumophila ΔsidJ mutant (Fig. 6). The PNSs were mock treated or incubated with purified wild-type SidJ, a SidJ mutant (SidJ DD), or BSA as a negative control. The SidJ DD mutant contains mutations in two conserved aspartate residues (D542A D545A) (S6B Fig.), was fortuitously identified based on a fallacious lead using a protein fold recognition server (Phyre), but resulted in a protein that nevertheless failed to complement the intracellular growth defect of a ΔsidJ mutant (S6C Fig.). As shown in Fig. 6A and 6B (mock), SidE proteins remained associated with the polar sites of the Legionella phagosome in the absence of SidJ at both 1-hour and 4-hour post infection. Addition of purified wild-type SidJ to the PNS displaced the SidE proteins from LCVs obtained from either the 1-hour or the 4-hour infection. In contrast, inclusion of an equivalent amount of the non-functional SidJ mutant, SidJ DD, or BSA had no effect on SidE co-localization with the LCV (Fig. 6A and 6B). The disappearance of SidE was specific as addition of SidJ did not cause removal of LidA from the LCV (Fig. 6C and 6D). SidJ-mediated removal of SidE was both concentration (Fig. 6E) and time-dependent (Fig. 6F) consistent with a catalytic mechanism.
The ability of SidJ to remove SidE proteins from the LCV in vitro using lysates prepared from either 1 hour or 4 hours infections indicates that their absence is not simply due to LCV maturation but rather is a direct consequence of SidJ action. In addition, concentration-dependence of the reaction is consistent with SidJ accumulation in the host cell cytoplasm at later time points of infection (Fig. 4). In summary, these data suggest that the SidE proteins function as toxins during early stages of infection and that SidJ inactivates them by mediating their active removal from the Legionella-containing vacuole.
The pathogen L. pneumophila serves as an excellent model system to study the interactions between secreted effector proteins, as it exports between 200–300 substrates via its Dot/Icm T4SS [32]. In this study, we examined the relationship between the L. pneumophila Dot/Icm substrates SidJ and the SidE family. Six lines of evidence support a functional connection between these proteins. First, SidJ and SdeC, SdeB, and SdeA are all expressed from the same locus (Fig. 1). Second, the SuperΔP170 mutant, the CleanΔP170 mutant, and the ΔsidJ mutant each have a similar growth defect within host cells (Figs. 1 and S2). Third, expression of low levels of SdeA was able to complement/suppress the intracellular growth defect of the SuperΔP170 mutant, which lacks both the sidE family genes and sidJ (Fig. 1). Fourth, overexpression of SdeA was detrimental for the growth of only strains lacking sidJ (Fig. 1). Fifth, SdeA toxicity to yeast and mammalian host cells could be suppressed by co-expression of SidJ (Fig. 2). Sixth, SidJ is able to promote the disappearance of SidE proteins from the LCV at later points of infection (Fig. 3) and this could be reproduced in an in vitro assay using purified SidJ (Fig. 6). Based on these results, we propose that the L. pneumophila SidJ protein functions as a metaeffector to regulate the activity of the SidE protein family.
The concept that intracellular pathogens must regulate the activity of their secreted effectors during an infection is not surprising, as unregulated toxin activity would lead to the premature demise of the host cell. One method of regulation might entail the spatiotemporal delivery and/or control of substrates with opposing activities. For example, Salmonella’s SopE and SptP toxins act antagonistically to activate and inactivate Rho-family GTPases CDC42 and Rac1 at different times of infection via a combination of differential activity and temporal stability [19]. Likewise L. pneumophila regulates the activity of Rab1 by using a GEF (SidM/DrrA) and a GAP (LepB) [20–22]. In addition to this general mode of GTPase regulation, L. pneumophila is able to stabilize Rab1 in an active form using ampylation by the effector SidM/DrrA and then reverse the effect via de-ampylation by SidD [33–36]. Legionella also employs two additional effectors with opposing activities, AnkX and Lem3/Lpg0696, to inactivate and then release a separate population of Rab1-GDP via cholination [33–36].
An even more elegant form of effector regulation was recently described, wherein the effector SidH was inactivated by LubX, a L. pneumophila secreted E3 ubiquitin ligase that marks SidH for proteasome-dependent proteolysis by polyubiquitination [25]. The key to effector regulation in this case was the differential translocation of LubX and SidH into host cells, with SidH being rapidly secreted followed by the slower intracellular accumulation of LubX. Based on these results, LubX was described as being a “metaeffector”, which was defined as an effector that regulates another effector protein [25]. Reminiscent of the differential regulation and secretion described for SidH and LubX, the expression of SidE proteins is induced in early stationary phase allowing export to occur immediately upon host cell infection [16]. In contrast, SidJ is expressed constitutively [18] and accumulates within the host cell at later time points of infection (Fig. 4). The gradual accumulation of intracellular SidJ during infection correlates with the decreased level of the SidE proteins on the LCV.
These observations prompted us to propose a model whereby SidJ functions as a metaeffector to modulate the activity of the SidE proteins (Fig. 7). In this model, SidE proteins are translocated into the host cells by the L. pneumophila Dot/Icm T4SS at early points of infection and localize on the cytoplasmic face of the immature LCV. Although the precise molecular function of the SidE proteins is not yet known, their early delivery into the host cell suggests they are involved in avoidance of the endocytic pathway and/or maturation of the LCV. As the infection proceeds, the SidJ protein begins to accumulate in the host cell, eventually reaching a critical threshold when it is competent to mediate the removal of the SidE proteins from the LCV (Fig. 7). Based on the inhibition by MG132, the simplest possibility is that SidJ directly targets SidE proteins for degradation by the proteasome (Fig. 7, top row). Alternatively, it is possible that SidJ mediates the degradation of another component that normally retains SidE proteins on the LCV surface. In the absence of this factor, the SidE proteins would no longer associate with the LCV and thus could redistribute and potentially associate with host organelles (Fig. 7, middle row).
In the absence of SidJ, SidE proteins appear to localize normally to the LCV at early time points of infection (Fig. 7, bottom row). However, as the infection proceeds, the SidE proteins are no longer removed from the LCV, they accumulate to high levels, eventually inhibiting the growth of L. pneumophila. Overproduction of SdeA in the absence of SidJ was toxic to both yeast cells and HEK293 cells and inhibited the growth of L. pneumophila due to delivery of the LCV to the lysosome. The disruption of the Golgi in mCherry-SdeA transfected cells suggests that the target of SdeA is likely to be a component of the secretory pathway.
The failure to eliminate SidE proteins from the LCV in a ΔsidJ mutant does not appear to be due to an indirect effect of the LCV not maturing as we can induce removal of SidE proteins from 4-hour LCVs in vitro by the addition of wild type SidJ. Rather we prefer the idea that SidJ directly mediates removal of SidE proteins from the LCV, perhaps by some form of post-translational modification. Although we have been unable to reproducibly demonstrate a robust interaction between SidJ and SdeA, it is reasonable that the proteins interact based on the SidJ suppression of SdeA-mediated toxicity in yeast and HEK293 cells. It is also possible that SidJ, which is a large protein of ∼90 kDa, possesses multiple biochemical activities, particularly since a partial ER recruitment defect has been reported for a ΔsidJ mutant [18].
In summary, the Dot/Icm substrate SidJ functions as a metaeffector to regulate the activity of the SidE substrates. Similar to the metaeffector LubX, SidJ promotes the removal of Dot/Icm T4SS effectors from the LCV in a proteasome-dependent manner. The presence of dual effectors with opposing activities, and the existence of metaeffectors that modulate the activity of other effectors, may partially explain why L. pneumophila translocates such a vast repertoire of T4SS substrates into host cells. Moreover, the discovery of a second metaeffector in L. pneumophila suggests that the concept of metaeffectors is not unique to LubX and additional pathogens may use similar strategies to highjack host cells.
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Protocol 20120081 was approved by the Institutional Animal Care and Use Committee at the Washington University School of Medicine. All efforts were made to minimize suffering.
Bacterial strains, plasmids, and primers are listed in S1 Table. Detailed plasmid construction is described in S2 Table. All L. pneumophila strains were cultured on ACES [N-(2-acetamido)-2-aminoethanesulfonic acid]-buffered charcoal yeast extract agar (CYE) or in ACES-buffered yeast extract broth (AYE) [37]. Antibiotics and thymidine (100 μg/ml) were added as needed. Strain Lp02 (thyA hsdR rpsL) is a derivative of the clinical isolate L. pneumophila Philadelphia-1 [38]. E. coli strain XL1 Blue was grown in Luria-Bertani (LB) broth or on LB agar with antibiotics as needed. Yeast strains were grown in YPD medium or yeast minimal medium supplemented with amino acids as needed. A/J mice were obtained from Jackson Laboratories. Mouse bone marrow-derived macrophages (BMMs) were differentiated from stem cells isolated from the femurs of female A/J mice and cultured in RPMI-1640 containing 20% FBS, 1.6 mM glutamine, 30% L-cell culture medium, and penicillin (10,000 IU/ml)/streptomycin (10 mg/ml) for one week as previously described [6,39]. Acanthamoeba castellanii cultures were maintained in PYG broth as previously described [40]. HEK-293 cells (obtained from American Type Culture Collection, Manassas, VA) were maintained in DMEM supplemented with 10% heat-inactivated fetal calf serum (FBS) (HyClone, Logan, UT) in a humidified CO2 incubator at 5% CO2 concentration. Human monocytic cell line U937 [41] were cultured in RPMI-1640 supplemented with 10% FBS and 2 mM glutamine. To differentiate the cells, they were treated with phorbol 12-myristate 13-acetate (Sigma, St. Louis, MO) for 36 h before use.
Intracellular growth of L. pneumophila was assayed using A. castellanii as a host cell. A. castellanii was propagated using PYG medium. Cells were grown to near confluency, recovered, counted, and plated into 24-well culture dishes at a density of 6 x 105 per well. The following day, the cells were washed and equilibrated at 37°C for 1 hour in A. castellanii buffer [40,42–44]. The amoebae were infected with stationary phase L. pneumophila cells at a multiplicity of infection (MOI) of 0.2 for 1 hour, washed three times to remove extracellular bacteria, and incubated for two days. L. pneumophila growth was assayed at 0, 20, 32, and 44 hours. At each time point, infected amoebae were lysed with 0.05% saponin (Sigma, St. Louis, MO) in PBS, the lysate was serially diluted and plated on CYE plates to assess bacterial growth. All growth assays were performed in triplicate.
Legionella sidJ and sdeA ORFs were cloned into the yeast expression vectors under control of the Pgal or Pcyc promoters (S1 and S2 Table). Plasmids were transformed into yeast cells (JY221, S1 Table) using the lithium acetate/PEG method [45]. Transformed cells were plated on yeast minimal media (US Biological, Massachusetts, MA), synthetic complete (SC) media lacking uracil (Ura) or leucine (Leu) in order to select for transformants. To determine the effect of SidJ or SdeA protein on yeast cell growth, strains were grown to saturation in SC minus Ura or SC minus Leu media containing 2% glucose. Cells were then adjusted to an A600 of 1.0, serially diluted 10-fold, and 5 μl of each dilution was spotted onto SC minus Ura or SC minus Leu containing either 2% glucose (non-induction) or 2% galactose (induction). Plates were incubated at 30° C for 48–72 hr and growth of recombinant strains was recorded.
For transfection of YFP-SidJ and mCherry-SdeA, HEK293 cells were seeded onto glass coverslips in 24-well dishes, incubated for one day, then transfected with 0.2 μg plasmid DNA using FuGene6 (Invitrogen, Grand Island, NY) as described by the manufacturer. Transfections were allowed to proceed for 20–40 hr in DMEM supplemented with 10% FBS at 37°C with 5% CO2. Cells were then fixed with 4% paraformaldehyde for 20 min at room temperature, and fixed cells were analyzed by fluorescence microscopy. For the Golgi fragmentation assay, cells were permeabilized with 100% methanol for 10 sec, and then blocked for 10 min with 5% goat serum in PBS. Cells were then stained with anti-giantin antibody (1:400, Covance, Princeton, NJ), followed by Alexa blue-conjugated goat anti-rabbit IgG (Invitrogen, Grand Island, NY) as a secondary antibody. Coverslips were mounted using ProLong Gold antifade reagent (Invitrogen, Grand Island, NY) before examined by fluorescence microscopy.
To quantitate effector protein secretion, we measured the adenylate cyclase activity of CyaA fusions (S1 Table). Differentiated U937 cells were plated into 24-well tissue culture plates at 2.5 x 106 per well. Legionella cultures, induced with IPTG at mid-log phase and grown two more hours to reach stationary phase, were harvested, washed, and diluted in RPMI-1640 supplemented with 10% FBS. 5 x 106 bacteria were added to each well for 1 hour, followed by washing three times with cold PBS to remove non-adherent cells, and lysis in 200 μl of lysis buffer (50 mM HCl and 0.1% Triton X-100) on ice. The lysates were transferred to 1.5 ml tubes, boiled for 5 min, and 12 μl of 0.5 M NaOH was added to neutralize the samples. cAMP was extracted using 2 volumes of 95% ethanol and collected after centrifugation at 12,000 g for 5 min to remove cell debris and then lyophilized. Total cAMP concentration was measured using an ELISA kit (GE Healthcare, Pittsburgh, PA).
Protein samples were collected and boiled for five minutes in Laemmli sample buffer and separated by SDS-PAGE gel electrophoresis, followed by transfer to PVDF membranes [46,47]. Membranes were blocked in BLOTTO (PBS containing 5% non-fat dry milk), washed with wash buffer (PBS containing 0.05% Tween 20) and incubated for 1 hour with antibody diluted in BLOTTO. Blots were washed with wash buffer followed by one hour incubation with secondary goat anti-rabbit antibody conjugated to horseradish peroxidase (Sigma, St. Louis, MO) diluted 1:10,000 in BLOTTO. Blots were subsequently washed with wash buffer prior to development using an ECL detection kit (GE Healthcare, Pittsburgh, PA) according to their protocol.
Mouse BMM were seeded on glass coverslips at 1 x 105 cells in 24-well plates and incubated overnight. Legionella cells were grown to stationary phase in AYE, washed in sterile water, then adjusted to OD600 = 1.0. Cells were diluted in warmed RPMI-1640 and 5 x 105 bacteria were added to wells containing BMM attached to coverslip. BMMs were infected for 1 hour. After washing to remove uninfected bacteria, cells were fixed using Periodate-Lysine-Paraformaldehyde (PLP) [48] and then permeabilized with methanol for 10 seconds. For effector localization, cells were stained with the SidE family antibody that was raised against SdeC (1:1,000) [16], LidA (1:1,000) [31], or SidM (1:300) [49] antibodies followed by goat anti-rabbit secondary antibody conjugated to Oregon Green (1:1:1,000) (Molecular Probes, Eugene, OR). DNA (bacteria and host nuclei) was stained with propidium iodide (1 mg/ml, Invitrogen, Grand Island, NY). Coverslips were mounted using ProLong Gold antifade reagent (Invitrogen, Grand Island, NY) before being examined by fluorescence microscopy. Legionella containing vacuoles (LCVs) decorated with effectors were scored positive by the visual presence of foci of Oregon Green adjacent to bacterial-shaped propidium-iodide staining.
To detect the intracellular localization of SidJ and SdeA, differentiated U937 cells were plated at a density of 1 x 107 cells per well in a 6-well plate. The next day, U937 cells were infected for 1 hour with stationary phase cultures of L. pneumophila and washed three times with PBS to remove uninfected bacteria. Cells were harvested using a cell scraper, washed once with cold PBS and pelleted. To fractionate the lysates, harvested cells were dounced in cold PBS without digitonin. Unbroken cells were removed by centrifugation (3 min, 200 g) at 4°C. Pellets and supernatant were separated by centrifugation at 12,000 g for 10 min to collect the pellet and cytosolic fraction. To fractionate secreted effector proteins, cells were resuspended in lysis buffer (PBS containing 0.2% digitonin) and dounced. Unbroken cells were then removed by centrifugation (3 min, 200 g) at 4°C. The secreted effector proteins were collected from supernatant after removing cell pellets by centrifugation (10 min, 12,000 g). Samples were analyzed by Western blot with SidE, SidJ, DotF, LidA, and SidM specific antibodies as above.
Separation of LCV by sucrose density gradient ultracentrifugation was performed as previously described [50]. Postnuclear supernatant (PNS) of infected cells was prepared as follows. Briefly, 1 x 107 differentiated U937 cells plated in 6-well plate were infected with stationary phase L. pneumophila at an MOI of 5. At the indicated time, the infected cells were suspended in 2 ml of homogenization buffer (20 mM HEPES pH 7.2, 250 mM sucrose, 0.5 mM EGTA) and were gently disrupted in a 7-ml dounce homogenizer. Unbroken cells and nuclei were pelleted by centrifugation at 4°C (3 min, 200 g). The PNS containing the L. pneumophila vacuoles were layered onto a 25–65% sucrose gradient and centrifuged at 100,000 g for 1 hour at 4°C. Fractions were collected from the bottom of the gradients and analyzed by SDS-PAGE followed by Western blotting. Separation of LCV from cell organelles was assessed by monitoring for the presence of DotF, a component of the Dot/Icm T4SS.
The sidJ open reading frame was amplified and inserted into pQE-30 to express His-SidJ (S1 and S2 Table). E. coli strain XL1Blue, harboring the resulting plasmid, pJB5331, was used to purify His-tagged SidJ with Ni-NTA columns according to protocols suggested by the manufacturer (Qiagen, Valencia, CA). Dissociation of SidE proteins from PNS was performed as below. Briefly, 2 x 106 BMM cells plated in 6 well-plate were infected with stationary phase L. pneumophila at an MOI of 0.5, incubated for 1 hour, followed by washing with warm RPMI-1640 to remove uninfected bacteria. At the indicated time, the infected cells were suspended in two ml of homogenization buffer containing 250 mM sucrose in PBS and dounced with 3 strokes. PNS was collected by removing unbroken cells by centrifugation at 4°C (3 min, 200 g). Collected PNS were incubated with 0.5 μM of purified SidJ for 30 min at room temperature. The reaction was stopped by addition of equal volume of 4% paraformaldehyde in a stock PLP solution. The fixed cells were attached on lysine-coated glass slides and dissociation of SidE family from LCV was monitored by immunofluorescence detection using SidE antibody (1:1,000 dilution) or LidA (1:1,100 dilution), followed by goat anti-rabbit Oregon Green secondary antibody (Molecular Probes, Eugene, OR). DNA was stained with DAPI and coverslips were mounted using ProLong Gold antifade reagent (Invitrogen) before being examined by fluorescence microscopy. Legionella containing vacuoles (LCVs) decorated with effectors were scored positive by the visual presence of foci of Oregon Green adjacent to bacterial-shaped DAPI staining.
Statistical analysis was performed using the GLIMMIX procedure of SAS 9.4 (9.4 SAS Institute Inc.). Data are presented as means ± SEM from three independent experiments. Statistical significance was declared if P<0.05.
|
10.1371/journal.pcbi.1003457 | Bayesian Reconstruction of Disease Outbreaks by Combining Epidemiologic and Genomic Data | Recent years have seen progress in the development of statistically rigorous frameworks to infer outbreak transmission trees (“who infected whom”) from epidemiological and genetic data. Making use of pathogen genome sequences in such analyses remains a challenge, however, with a variety of heuristic approaches having been explored to date. We introduce a statistical method exploiting both pathogen sequences and collection dates to unravel the dynamics of densely sampled outbreaks. Our approach identifies likely transmission events and infers dates of infections, unobserved cases and separate introductions of the disease. It also proves useful for inferring numbers of secondary infections and identifying heterogeneous infectivity and super-spreaders. After testing our approach using simulations, we illustrate the method with the analysis of the beginning of the 2003 Singaporean outbreak of Severe Acute Respiratory Syndrome (SARS), providing new insights into the early stage of this epidemic. Our approach is the first tool for disease outbreak reconstruction from genetic data widely available as free software, the R package outbreaker. It is applicable to various densely sampled epidemics, and improves previous approaches by detecting unobserved and imported cases, as well as allowing multiple introductions of the pathogen. Because of its generality, we believe this method will become a tool of choice for the analysis of densely sampled disease outbreaks, and will form a rigorous framework for subsequent methodological developments.
| Understanding how infectious diseases are transmitted from one individual to another is essential for designing containment strategies and epidemic prevention. Recently, the reconstruction of transmission trees (“who infected whom”) has been revolutionized by the availability of pathogen genome sequences. Exploiting this information remains a challenge, however, with a variety of heuristic approaches having been explored to date. Here, we introduce a new method which uses both pathogen DNA and collection dates to gain insights into transmission events, and detect unobserved cases and separate introductions of the disease. Our approach is also useful for identifying super-spreaders, i.e., cases which caused many subsequent infections. After testing our method using simulations, we use it to gain new insights into the beginning of the 2003 Singaporean outbreak of Severe Acute Respiratory Syndrome (SARS). Our approach is applicable to a wide range of diseases and available in a free software package called outbreaker.
| Statistical methods for analyzing detailed epidemiological data collected during infectious disease outbreaks have seen rapid development in recent years [1], [2], [3], [4], [5], [6], [7], [8]. These methods probabilistically reconstruct likely transmission links between cases using data on the timing of symptoms and, where available, contact tracing data or other proximity information. The resulting transmission trees allow estimation of the number of secondary infections generated by each case, and thus of the transmission intensity (characterized by the reproduction number, R) over time. Pathogen genetic sequence data provides valuable additional information on potential transmission links between cases in a disease outbreak, particularly when reliable contact tracing data is not available. Indeed, using sequence data alone to estimate transmission rates during epidemics is increasingly frequent [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19]. As genetic sequences can now be obtained nearly in real-time [19], [20], this new source of information opens up exciting perspectives not only for understanding past outbreaks, but also for unraveling the transmission routes of ongoing outbreaks and subsequently adapting public health responses.
Integrated analysis of both epidemiological and sequence data clearly would maximize our ability to reconstruct transmission trees, but there are methodological and computational challenges. These challenges center on constructing and evaluating a unified likelihood for both the genetic and epidemiological data. One of the first attempts at integrated analysis [21] used phylogenetic trees to constrain the set of transmission trees then explored by an epidemiological transmission tree inference algorithm. An alternative approach [22] highlighted limitations of phylogenetic methods for reconstructing densely sampled outbreaks, and proposed an alternative graph theoretic approach for reconstructing ‘genetically parsimonious’ transmission trees, i.e. trees implying the smallest number of genetic changes amongst the sampled isolates. While simple and fast, this method also has a number of limitations: dates of infection are not inferred, the probability of a given transmission event cannot be assessed, and unobserved cases or multiple introductions of the disease cannot be detected. Substantive methodological developments have been made by Ypma et al. [23] and subsequently by Morelli et al. [24], both of which proposed unified likelihoods for genetic and epidemiological data to analyze livestock disease outbreaks (avian influenza H7N7 [23] and foot-and-mouth disease [24]). However, those methods require that the outbreak has a single introduction event and that all cases are observed, which limits their applicability to restricted epidemic contexts.
Here we introduce a novel and generic framework for the reconstruction of disease outbreaks based on pathogen genetic sequences and collection dates. We use the distribution of the generation time (i.e. time interval between a primary and a secondary infection) [7], [8] to define the epidemiological likelihood of a given transmission tree. This is coupled with a simple model of sequence evolution defining the probability of the genetic changes observed between the pathogen genomes along a chain of transmission. Our model is embedded within a Bayesian framework allowing estimation of dates of infections, mutation rates, separate introductions of the pathogen, the presence of unobserved cases, and the transmission tree. Estimate of the effective reproduction number over time, R(t), can also be obtained. As an improvement over previous approaches [23], [24], our method does not require all cases to be observed or there to be a single introduction event which triggers an outbreak. After evaluating the performance of our method using simulated outbreaks, we illustrate our approach by analyzing the 2003 Severe Acute Respiratory Syndrome (SARS) outbreak in Singapore [10], [11], [25]. Our method is implemented in the package ‘outbreaker’ for the R software [26] and represents the first widely available tool for the reconstruction and analysis of disease outbreaks from genomic data.
We analysed simulated outbreaks to assess the performance of our method under a variety of conditions, including different basic reproduction numbers (R0), sampling coverage, rates of evolution, and generation time distributions, with our base scenario resembling an influenza-like illness (Table 1). The outbreak size varied from 10 to nearly 200 infections in a fixed population of 200 susceptible hosts (plus imported cases), with a median sample size of 110 (quartile range: [66–132], Fig. S1). Wherever applicable, reported results refer to the marginal distributions.
Transmission trees were overall very well reconstructed, with 70% to 90% of true ancestries being recovered in most simulation settings (Fig. 1 and Table S1 in Text S1). Better results were achieved when the sampling coverage was high (compare settings ‘Base’ to 75%, 50% and 25% of missing cases). In the absence of genetic information, the transmission tree was very difficult to infer (setting ‘No mutation’). Differences in basic reproduction numbers (settings ‘Low R’ and ‘High R’) and in the shape of the generation time distribution (settings ‘Short generation’ and ‘Long generation’) induced some variation in the proportions of successfully recovered ancestries, although these remained satisfying in every case (Fig. 1 and Table S1 in Text S1). Dates of infections were inferred with accuracy in most settings (Fig. S2 and Table S1 in Text S1). However, this result was mostly driven by the shape of the generation time distribution, with broader distributions leading to greater uncertainty in the dates of infection (Fig. S2). While perfectly inferred in fully sampled outbreaks, the number of generations between ancestor and descendents became ambiguous as the proportion of missing cases increases (Table S1 in Text S1). Mutation rates were also mostly well estimated (Table S1 in Text S1, Fig. S3), albeit with a tendency to over-estimation. This bias was stronger when sampling grew sparser (settings with 75% and 50% missing cases), and to a lesser extent when the number of imported cases grew large (setting ‘Many imports’). Detailed investigation of individual simulations suggested that misdetection of imported cases and increased numbers of erroneous ancestries may be responsible for over-estimating the mutation rates in these settings. The inference of sampling coverage varied largely amongst different simulation settings (Table S1 in Text S1, Fig. S4): well recovered in fully sampled outbreaks, it was largely overestimated in sparse samples (settings with 75%, 50% and 25% missing cases), and slightly underestimated with longer generation time.
The detection of imported cases showed excellent specificity and good sensitivity pooling results across the simulated datasets examined, with a majority of simulations exhibiting perfect results (Fig. 2). However, substantial variations were observed between simulation settings (Fig. S5, Table S1 in Text S1). Unsurprisingly, detection of imported cases was more difficult when imported cases were more frequent and when a higher fraction of cases was unobserved. With longer generation times, the larger numbers of mutations accumulated between ancestors and descendents made the detection of genetic outliers, and thus of imported cases, nearly impossible (Fig. S2).
While our model does not explicitly estimate the effective reproduction number ‘R’ (i.e., the number of secondary cases per infected individual), this quantity can easily be computed from the posterior trees. Our ‘base’ simulations show that reliable estimates of R at an individual level can be obtained when genetic information is available (Fig. 3, left). In contrast, such inference was impossible in the absence of genetic data (Fig. 3, right).
To gain a better understanding of disease outbreak dynamics, identifying systematic heterogeneity in R across cases is also essential. To assess whether our approach could detect such heterogeneity, we implemented two types of simulations in which there were systematic differences in infectivity between groups of hosts. In a first set of simulations, the host population was divided into two groups of equal sizes (e.g. adults and children) with low and high infectivity (infectivity in one group was twice that of the other group, with equal susceptibility). In the second setting, we included rare (5%) super-spreaders, who had the same susceptibility to infections as non super-spreaders, but were 13-fold more infectious. In both sets of simulations, infectivity was fixed for each individual at the beginning of the simulations. The classification of individuals into super-spreaders and regular spreaders was considered as known when comparing estimated reproduction numbers.
Results showed that our method was able to recover contrasted infectivity between different groups (Fig. 4, S6, 7, 8, 9). In the simulations with equally-sized groups, the overall distributions of R for each group were almost perfectly recovered (Fig. 4, top panel), while values of R at an individual level were also well estimated (Fig. S6). Importantly, when ignoring the genetic information, differences between groups were barely detectable (Fig. 4 and S7). Similar results were observed in simulations including super-spreaders (Fig. 4, bottom panel), in which estimates of R values at an individual level were excellent when using genetic information (Fig. S8), and very poor without it (Fig. S9). The reconstruction of average R values over time was not improved by the inclusion of genetic information (Fig. S10, S11), which is unsurprising as this mainly depends on correctly inferring the dates of infections, which was unaffected by the absence of genetic data (Fig. S1, S2).
We analyzed data collected during the beginning of a SARS outbreak which took place in Singapore in 2003 [10], [25]. Previous studies proposed different reconstructions of this outbreak based on indirect contact tracing information and genetic data, and while all agreed on the necessity to combine these two streams of information, a clear consensus on the initial transmission tree has not been reached [10], [11], [25]. Here, we aimed to reconstruct the early stage of this outbreak using 13 full SARS genomes collected from the putative index patient and primary and secondary cases, and previously published estimates of the generation time distribution [27] (Fig. S12).
The genetic diversity amongst isolates was limited, with less than 15 mutations separating any pair of genomes (Fig. S13). For most cases, transmission events could not be readily inferred from the phylogenetic tree (Fig. S14). According to previous estimates of the mutation rate [25], we expect that most direct transmissions (>99%) will exhibit between 0 and 5 mutations. Using this result, we performed a simple graph analysis to derive possible clusters of direct transmissions, which suggested the existence of one main cluster of cases that may be linked directly, the remaining 4 isolates falling into three groups (Fig. S15). However, this crude analysis only relied on genetic diversity, and did not take into account information on the collection dates of the isolates or on the duration of the infectious period.
We used outbreaker to exploit all these data simultaneously. Results of the inferred likely scenarios (Fig. 5 and 6) show that for half of the cases, a well-supported ancestor can be identified from the data (see also Fig. S16). These correspond to all of the first and second generations of infections (Sin2677, Sin2679, Sin2748, Sin2774) and to the last sampled case (Sin850). Ancestries of most cases were compatible with a single generation, although one or two unobserved infections may have taken place between Sin849 and Sin850 (Fig. S17). We found no evidence for separate index cases after Sin2500, in agreement with contact tracing information [10], [11], [25]. However, the small number of cases may impair the detection of outliers and thus the identification of imported cases, so that multiple introductions of the pathogen cannot be ruled out.
The most recent investigation of this outbreak suggested a dual introduction of the pathogen, with a separate index case (Sin2679) nearly 20 days after the initial index case Sin2500 [10], [11], [25]. This may be deemed surprising as this case is genetically close to some preceding cases (Fig. S14, S15). Here, our results suggest that Sin2679 would in fact be part of the second generation of infection, and was infected by Sin2748 (Fig. 5 and 6). Indeed, while the collection dates of Sin2748 and Sin2679 are relatively close, the generation time of SARS (Fig. S12) may have allowed for this transmission to occur. Closer examination of the patterns of mutations between Sin2500, Sin2748 and Sin2679 bring further support to this scenario (Fig. 6, Data S3). Indeed, the four mutations separating Sin2500 from Sin2679 are the simple addition of the mutations accumulated on the chain of transmission, from Sin2500 to Sin2748 (position 26,430: a→g), and from Sin2748 to Sin2679 (18,284: c→a; 19,086: t→c; 23,176: c→t).
Building on past work [23], [24], we have presented a flexible analytical framework for the reconstruction of densely sampled outbreaks from epidemiological and sequence data. We extended previous work by accounting for unobserved cases and proposing a new approach for identifying multiple introductions of the pathogens based on the detection of genetic outliers. Our method is also the first tool for outbreak reconstruction widely available as a free software (the R package ‘outbreaker’) and able to run on standard desktop computers. The analysis of simulated data suggests that our approach will be applicable to a wide range of pathogens with various basic reproduction numbers, generation time distributions, and genetic diversity. We have shown how our approach can be used to infer effective reproduction numbers at an individual level. Importantly, this allows for detecting differences in infectivity of different groups of cases, and for the identification of super-spreaders. Our results suggest that while epidemiological data may suffice for the estimation of mean aggregated quantities such as the mean effective reproduction number, R, genetic data are useful to tease individual heterogeneities apart.
As in other tree reconstruction methods [2], [7], [28], [29], we did not explicitly model the population of susceptible individuals. This is because information on individuals who were not infected during the outbreak (the “denominator” data) is quite often unavailable. Compared with case-only analyses, availability of denominator data also makes it possible to estimate the force of infection and risk factors for infection [4]. We note that our framework could easily be extended to model the uninfected population. This could be done by modifying our likelihood so that the probability of the time of infection of a case would be based on an explicit model of the force of infection; individuals not infected during the outbreak would also contribute to the epidemiological likelihood as is standard in such situations [4]. Integrating and validating these additional features in our approach will be the subject of future research.
Our method relies on several assumptions which can be used to define the scope of its possible applications. The most important element in this respect is the proportion of cases represented in the sampled data, and thus often the scale of the epidemics considered. Our approach aims to reconstruct ancestries in closely related cases. As such, it should be most useful for detailed outbreak investigations. While the reconstruction of transmission tree seems relatively robust to large proportions of unobserved cases (up to 75% of missing cases, Fig. 1), our method is clearly tailored to densely sampled outbreaks, and not meant for the analysis of large-scale, more sparsely sampled epidemics. In such cases, phylogenetic methods are preferred as they explicitly reconstruct unobserved common ancestors of the sampled pathogen genomes, and can be used to infer, if not the transmission tree, the past dynamics of the disease [30], [31], [32].
One of the novelties of our approach is the detection of imported cases, which are identified as genetic outliers. While this method should be useful to detect separate introductions of different pathogenic lineages in an epidemic, it may be sensitive to other events prone to creating genetic outliers, such as sequencing errors or recombination. Care should therefore be devoted to ensuring data quality and filtering out polymorphism due to recombination. Moreover, the assumption that imported cases are genetically distinguishable from other cases may not always be true, especially when multiple introductions take place from a closely related lineage. Such cases cannot be detected by genetic data only, and would require other sources of information (e.g. contact tracing) to be considered. In this respect, an interesting feature of outbreaker is the ability to fix known imported cases (as well as any other known transmissions) before reconstructing the transmission tree.
Another important point is that following a previous, widely-used approach for the analysis of outbreaks [7], we assume the distributions of the generation time and of the time from infection to sample collection to be known. In some situations such as outbreaks of new emerging pathogens, accurate estimates of the generation time may not be readily available. In this case, a conservative approach should allow for a wide range of possible times to infection, at the expense of increased uncertainty in the inferred ancestries. As our method is numerically efficient for the analysis of small outbreaks, we suggest testing different generation time distributions to assess the robustness of the results. As a longer-term alternative, our approach could be extended to include an explicit parameterization and estimation of the generation time distribution.
More fundamentally, the use of a generation time distribution also implies that our method is less appropriate for diseases in which long periods of asymptomatic carriage are frequent. For instance, bacteria such as Staphylococcus aureus can cause infections after months of asymptomatic colonization of the host, but may equally cause outbreaks of cases linked by only a few days [12], [33]. In such cases, the collection dates of isolates effectively carry less information about possible transmissions, which would hamper our current approach. However, our model could be adapted to the analysis of carried pathogens by incorporating specific data on known exposures (e.g. shared occupancy on a hospital ward) [34], [35], [36].
Moreover, carried pathogens are also more likely to cause multiple colonizations of the host, resulting in several lineages coexisting within the same patient. Our model assumes that a single pathogen genome exists within each host, and is therefore not designed to account for multiple infections. A simple workaround would consist in duplicating cases of multiple infections into single infections, assuming that multiple infections are made of independent, single colonization events. However, this would not allow for disentangling multiple infections from mere within-host evolution of a single lineage. A more satisfying approach would consist in modeling explicitly the evolution of isolates within host, but this will likely result in a much more complex model and is beyond the remit of our current approach.
A major simplification made in our model, that could be relaxed in future work, is that we do not consider within host diversity of pathogens. Within-host diversity is particularly prominent in pathogens that infect a host for a long time relative to their within-host replication cycle (e.g. HIV or Hepatitis C Virus), pathogens that can be carried for a long time (e.g. Staphylococcus aureus), pathogens where the infectious inoculum is large (e.g. blood-transmitted HIV), or super-infection is frequent (e.g. Streptococcus pneumoniae in hyper-endemic settings). Limited host diversity leads us to assume that genomes sampled from infectors are effectively ancestral to genomes sampled from secondary cases, allowing us to equate phylogenetic and transmission trees. This substantially reduces the complexity of the inferential problem, and reduces by orders of magnitude the dimensionality of the space of linked augmented variables to be explored. The assumption of no within-host diversity will likely be appropriate for acute infectious pathogens in outbreaks, but will also be relatively appropriate for situations where there is a strong bottleneck on diversity upon transmission and limited opportunities for superinfection, such as sexually transmitted HIV. Inclusion of within-host diversity in the model inference is an important but likely complex task, though efficient approximations may be possible. A related development will be the inclusion of multiple samples per individual, used to sample cross-sectional and longitudinal genetic diversity within infected hosts. Another somewhat simpler extension would be the inclusion of a ‘relaxed’ molecular clock, which would allow accounting for heterogeneities in mutation rates amongst different pathogen lineages.
Finally, we wish to emphasize the importance of including all available prior information in the analysis. Because the estimates of parameters governing an outbreak are often correlated, accurate knowledge of one can be used to refine the estimation of the others. For instance, specifying known transmission chains or imported cases will improve the estimation of the mutation rates, as well as the overall reconstruction of the transmission tree. Conversely, fixing the mutation rate to its ‘true’ value (or a good estimate thereof) is likely to improve the detection of imported cases. As currently implemented, our method allows for fixing any parameter as well as individual ancestries, which are used in the likelihood computations but not changed during the MCMC. This feature should be especially useful for incorporating known transmission events or introductions of the pathogen into the population, based for instance on clinical investigations and contact tracing information. However, results of contact tracing studies should always be considered cautiously, and could be contradicted by the analysis of corresponding sequences, as illustrated by the SARS outbreak in Singapore.
There are other promising avenues for incorporating various streams of information into our approach. The likelihood of our model allows for additional ‘plug-in’ terms for individual transmissions, which could be used to model spatial dispersion processes as well as movement over a contact network. Therefore, we hope that the present method will not only be applied widely, but also motivate further developments for the investigation of infectious disease outbreaks.
Thirteen previously published full SARS genomes [10], [25] (Data S1) were obtained from Genbank and aligned using MUSCLE [39]. The resulting alignment contained 29,731 columns, 39 of which were polymorphic (Data S2). We used a generation time distribution modeled as a discretized gamma distribution with a mean of 8.4 days and a standard deviation of 3.8 days [27], using the function DiscrSI from the R package EpiEstim [29]. The same distribution was used for the the time to collection. Details of the parameters used to run outbreaker are provided in Supporting Methods. The statistical confidence in determining the ancestry of a given case was quantified using the entropy of the frequencies of the posterior ancestors. With different ancestors of posterior frequencies (), the entropy is defined as:(11)The entropy is 0 if one of the , is 1, indicating high confidence in allocation of an ancestry, while larger values of the entropy indicate poorer confidence.
|
10.1371/journal.pcbi.1006134 | Nonmechanistic forecasts of seasonal influenza with iterative one-week-ahead distributions | Accurate and reliable forecasts of seasonal epidemics of infectious disease can assist in the design of countermeasures and increase public awareness and preparedness. This article describes two main contributions we made recently toward this goal: a novel approach to probabilistic modeling of surveillance time series based on “delta densities”, and an optimization scheme for combining output from multiple forecasting methods into an adaptively weighted ensemble. Delta densities describe the probability distribution of the change between one observation and the next, conditioned on available data; chaining together nonparametric estimates of these distributions yields a model for an entire trajectory. Corresponding distributional forecasts cover more observed events than alternatives that treat the whole season as a unit, and improve upon multiple evaluation metrics when extracting key targets of interest to public health officials. Adaptively weighted ensembles integrate the results of multiple forecasting methods, such as delta density, using weights that can change from situation to situation. We treat selection of optimal weightings across forecasting methods as a separate estimation task, and describe an estimation procedure based on optimizing cross-validation performance. We consider some details of the data generation process, including data revisions and holiday effects, both in the construction of these forecasting methods and when performing retrospective evaluation. The delta density method and an adaptively weighted ensemble of other forecasting methods each improve significantly on the next best ensemble component when applied separately, and achieve even better cross-validated performance when used in conjunction. We submitted real-time forecasts based on these contributions as part of CDC’s 2015/2016 FluSight Collaborative Comparison. Among the fourteen submissions that season, this system was ranked by CDC as the most accurate.
| Seasonal influenza is associated with 250 000 to 500 000 deaths worldwide each year (WHO estimates). In the United States and other temperate regions, seasonal influenza epidemics occur annually, but their timing and intensity varies significantly; accurate and reliable forecasts that quantify their uncertainty can assist policymakers when planning countermeasures such as vaccination campaigns, and increase awareness and preparedness of hospitals and the general public. Starting with the 2013/2014 flu season, CDC has solicited, collected, evaluated, and compared weekly forecasts from external research groups. We developed a new method for forecasting flu surveillance data, which stitches together models of changes that happen each week, and a way of combining its output with other forecasts. The resulting forecasting system produced the most accurate forecasts in CDC’s 2015/2016 FluSight comparison of fourteen forecasting systems. We describe our new forecasting methods, analyze their performance in the 2015/2016 comparison and on data from previous seasons, and describe idiosyncrasies of epidemiological data that should be considered when constructing and evaluating forecasting systems.
| Seasonal influenza epidemics cause widespread illness which is associated each year with an estimated 250 000 to 500 000 deaths worldwide [1] and 3000 to 56 000 deaths in the United States alone [2–4]. In contrast to influenza “pandemics”, which are rare global outbreaks of especially novel influenza A viruses [5, 6], seasonal epidemics (i.e., non-pandemics), while still having worldwide reach, occur annually in the United States and other countries with (generally) temperate climates. Time series of influenza prevalence in these areas are typically low and flat for the majority of the season, but trace a single, sharp peak sometime during winter, with significant variability in timing and intensity. Accurate and reliable forecasts of seasonal epidemics can help policymakers plan countermeasures such as vaccination campaigns, and increase awareness and preparedness of hospitals and the general public. The Centers for Disease Control and Prevention (CDC) monitors influenza prevalence with several well-established surveillance systems [7]; the recurring nature of seasonal epidemics and availability of historical data provide promising opportunities for the formation, evaluation, and application of statistical models. Starting with the 2013/2014 “Predict the Influenza Season Challenge” [8] and continuing each season thereafter as the Epidemic Prediction Initiative’s FluSight project [9], CDC has solicited and compiled forecasts of influenza-like illness (ILI) prevalence from external research groups and worked with them to develop standardized forecast formats and quantitative evaluation metrics.
Various approaches to influenza epidemic forecasting are summarized in literature reviews [10–12] and descriptions of the CDC comparisons [8, 9]. Some common approaches are described below, with references to work applicable to the current FluSight project and related seasonal dengue forecasting tasks, emphasizing more recent work that may not be listed in the above three literature reviews:
We present a novel phenomenological approach to epidemiological forecasting using “delta densities”, which assumes an autoregressive dependency structure similar to those of some time series models, but uses a kernel density estimation approach to model these dependencies rather than the common choice of linear relationships plus Gaussian noise. This technique is similar to the method of analogues [19] in that it uses an instance-based, nonparametric estimation procedure, but provides distributional forecasts of entire trajectories rather than point predictions of individual observations. The kernel conditional density estimation (KCDE) forecasting method [22] attacks many of the same issues encountered when applying kernel density estimation methods to seasonal epidemic data, but models the dependency structure of future weeks with a copula, while delta density chains together 1-week-ahead simulations. Compared to approaches that treat the entire season as a unit, such as deterministic, single-strain, fully-mixed compartmental models [13] or our previous empirical Bayes approach based on modifying past seasons’ data [21, 32], this method forms a larger library of possible trajectories by piecing together local models, which appears to help forecast performance, even though the trajectories considered may seem less reasonable on average.
Our second contribution is an adaptively weighted ensemble approach to combining the output of different forecasting methods given their historical and/or cross-validation forecasts. We first implemented this method in preparation for the 2014/2015 FluSight comparison, mixing together our empirical Bayes forecasting method with two baselines (a uniform distribution and an empirical distribution for each target), and later applied it while participating in the Dengue Forecasting project [33] and following FluSight comparisons (adding up to 9 additional components including delta density based methods), and found it improved our forecasts in all cases. Other epidemic forecasting teams have also reported success with concurrently or subsequently developed stacking generalization [34, 35] ensemble approaches to the FluSight forecasting tasks using Bayesian model averaging [36], the fixed weighting scheme that we examine below [37], and alternative adaptive weighting schemes based on gradient tree boosting [37], as well as with earlier ensemble approaches to short-term point predictions [20]. Methodologically, our adaptively weighted ensemble framework differs from these alternatives in that it selects a weighting over components for a particular forecast using “plug-in” statistical estimators for the optimal weights given the context of the forecast being prepared. Like the adaptive approaches presented in [37], component weights for each forecast are selected using regression, but the type of regression used and the manner of incorporating additional information, such as the forecast week, are distinct.
Recording every case of influenza is not practicable; infections are often asymptomatic [38] or symptomatic but not clinically attended [39], laboratory testing may not be performed for clinically attended cases or give false negative results, and reporting of lab-confirmed cases is not mandatory in most instances. Forecast comparisons are instead based on syndromic clinical surveillance data from the U.S. Outpatient Influenza-like Illness Surveillance Network (ILINet) [7, 40], a group of health care providers that voluntarily report statistics regarding ILI, where ILI is defined as a 100°F (37.8°C) fever with a cough and/or sore throat without a known cause other than influenza. CDC aggregates these reports and estimates the weekly percentage of patients seen that have ILI, %ILI, across all health care providers using a measure called weighted %ILI (wILI).
CDC hosts the latest ILINet report and other types of surveillance data through Fluview Interactive, a collection of web modules [41]; we provide current and historical ILINet reports and some other data sources through our delphi-epidata API [42] and epivis visualizer [43].
The FluSight project focuses on in-season distributional forecasts and point predictions of key targets of interest to public health officials:
When making distributional forecasts, wILI values are discretized into CDC-specified bins and a probability assigned to each bin, forming a histogram over possible observations. The width of the bins was set at 0.5 %wILI for the 2015/2016 comparison and 0.1 %wILI for the 2016/2017 comparison; we use a width of 0.5 %wILI for analysis of the 2015/2016 comparison prospective forecasts, and a width of 0.1 %wILI for retrospective evaluation. CDC typically presents wILI values rounded to a resolution of 0.1 %wILI; some targets and evaluations are based on these rounded values.
We focus on three metrics for evaluating performance of a forecast for a given target:
Unibin log score: log p ^ i, where p ^ i is the probability assigned to i, the bin containing the observed value. We use this score for ensemble weight selection and most internal evaluation as it has ties to maximum likelihood estimation, and is “proper score” [44]. A score for a (reported) distributional prediction p ^ is called “proper” if its expected value according to any (internal) distributional prediction q ^ is maximized when reporting p ^ = q ^, i.e., forecasters can maximize their expected scores by reporting their true beliefs. We refer to the “unibin log score” simply as the “log score” except for when comparing it with the multibin log score, which is defined next. The exponentiated mean log score is the (geometric) average probability assigned to events that were actually observed. The exponentiated difference in the mean log scores of method A and method B is an estimate of the (geometric) expected winnings of unit-sized bets of the form “this bin will hold the true value” when bets are placed optimally according to the forecasts of A, and (relative) prices are set optimally according to the forecasts of B. Multibin log score: log ∑ i near observed value p ^ i, where the i’s considered are typically bins within 0.5 %wILI of observed values for a wILI target, or within 1 week for a timing target. The multibin log score was designed by FluSight hosts in consultation with participants, and the judgment “near observed value” was selected as a level of error that would not significantly impact policymakers’ decisions. The exponentiated mean multibin log score is the (geometric) average amount of mass a forecaster placed within this margin for error of observed target values. Absolute error: | y ^ - y |, where y ^ is the point prediction and y is an observed value. (In the case of onset, we consider point predictions for the value of onset conditioned on the fact that an onset actually occurs. We do not consider absolute error for onset in instances where no onset occurred. Some methods considered would sometimes fail to produce such conditional onset point predictions when they were confident that there was no onset, but these methods are not included in any of the figures containing absolute errors.)
The FluSight 2015/2016 forecast comparison evaluations were based solely on the multibin log score [45].
The “flu season” is typically defined as epi week 40 of one year through 20 of the next; we also include data from the rest of the year as part of the season for the purpose of fitting models. In all mathematical notation, we will number the first week of the season as 1 rather than using the corresponding epi week. Let
W 1 . . t t denote the t-th CDC report of the current season, containing wILI values for weeks 1 through t, inclusive, which is normally published on Friday of week t + 1;
T be the number of weeks in the current season (either 52 or 53); we omit all details regarding differing season lengths, presenting forecasting methods and labeling epi week plot axes as if all seasons were of length T;
Y1..T be the ground truth wILI for the current season: the wILI values used for forecast evaluation, specifically the epi week 28 report for the FluSight comparison, or later revisions as they are available for cross-validation analysis;
Y 1 . . T s be the ground truth wILI for past season s; and
Zt be a vector containing the forecasting targets of interest at the t-th wILI report of the current season: Yt+1..t+4 and the seasonal onset, peak week, and peak percentage; for the FluSight comparison, forecasts for these targets were typically due on Monday of week t + 2, and allowed to use ILINet and any other data released before the deadline.
Our goal is to forecast Zt given W 1 . . t t and previous reports. This can be broken down into multiple steps, such as:
“Backcast” updates to the data through time t, producing a distribution over Y1..t based on the value of W 1 . . t t and previous reports.
Connect the backcast for Y1..t with corresponding forecasts for Yt+1..T, yielding a distribution for the entire trajectory Y1..t.
Calculate the distribution for Zt corresponding to this distribution over Y1..t.
We first introduce the delta density method, which forecasts Yt+1..T given Y1..t (step 2). We then discuss a separate procedure for combining multiple forecasts into an adaptively weighted ensemble, forecasting Zt given either Y1..t or W 1 . . t t (steps 2–3 or 1–3). We also outline a method for estimating the distribution of Y1..t given W 1 . . t t (step 1), and analyze its performance when used in conjunction with the delta density method.
Consider the task of estimating the density function f Y t + 1 . . T ∣ Y 1 . . t using an instance-based approach. Kernel density estimation and kernel regression use smoothing kernels to produce flexible estimates of the density of a random variable (e.g., f Y t + 1 . . T ) and the conditional expectation of one random variable given the value of another (e.g., E [ Y t + 1 . . T ∣ Y 1 . . t ]), respectively; we can combine these two methods to obtain estimates of the conditional density of one random variable given another. One possible approach would be to use the straightforward estimate
f ^ Y t + 1 . . T ∣ Y 1 . . t ( y t + 1 . . T ∣ y 1 . . t ) = ∑ s = 1 S I 1 . . t ( y 1 . . t , Y 1 . . t s ) O t + 1 . . T ( y t + 1 . . T , Y t + 1 . . T s ) ∑ s = 1 S I 1 . . t ( y 1 . . t , Y 1 . . t s ) O t + 1 . . t ( y t + 1 . . t , Y t + 1 . . t s ) ,
where {1..S} is the set of fully observed historical training seasons, and I1..t and Ot+1..T are smoothing kernels describing similarity between “input” trajectories and between “output” trajectories, respectively. However, while basic kernel smoothing methods can excel in low-dimensional settings, their performance scales very poorly with growing dimensionality. During most of the season, neither Y1..t nor Yt+1..T is low-dimensional, and the current season’s observations are extremely unlikely to closely match any past Y 1 . . t s or Y t + 1 . . T s. This, in turn, can lead to kernel density estimates for Yt+1..T based almost entirely on the single season s with the closest Y 1 . . t s when conditioning on Y1..t, and excessively narrow density estimates for Yt+1..T even without conditioning on Y1..t. So, instead of applying kernel density estimation directly, we first break the task down into a sequence of low-dimensional sub-tasks. We avoid the high-dimensional output problem by chaining together estimates of f Δ Y u ∣ Y 1 ..u - 1 for each u from t + 1 to T, where ΔYu = Yu − Yu−1; estimating these single-dimensional densities requires relatively little data. However, this reformulation exacerbates the high-dimensional input problem since we are conditioning on Y1..u−1, which can be considerably longer than Y1..t. We address the high-dimensional input problem by approximating f Δ Y u ∣ Y 1 ..u - 1 with f Δ Y u ∣ R u, where Ru is some low-dimensional vector of features derived from Y1..u−1. Smoothing kernel methods are used to approximate the conditional density functions using data from past seasons.
We use two sets of choices for the approximate conditional density function and summary features to form two versions of the method.
This same approach can be applied to estimate the distribution of residuals of a wILI point predictor. Suppose that we have observed
our goal is to estimate the conditional distribution of
given Y 1.. t 1 and X 1.. t 2 , using data from past seasons. This can be achieved by chaining together draws from conditional density estimates of (Y − X)u ∣ Ru for u from t1 + 1 to t2, where Ru is a function of Y1..u−1 and X 1.. t 2 . The delta density method can be seen as a special case where t1 = t; t2 = T; X1..t = Y1..t, past values of Y which are treated as known and are simply duplicated in the simulated trajectories; and Xt+1..T = Yt..T−1, values of Y which begin as unknown but are filled in as needed by previous simulation steps, giving (Y − X)t+1..T = ΔYt+1..T. We use the residual density method to backcast Y1..t from W 1 . . t t and as the basis for another forecaster in the ensemble.
Fig 2 shows sample forecasts over wILI trajectories generated by each of these approaches and compares them to some alternatives described in S1 Appendix.
Forecasting systems that select effective combinations of predictions from multiple models can improve on the performance of the individual components, as demonstrated by their successful application in many domains. For each probability distribution and point prediction in a forecast, we treat the choice of an effective combination as a statistical estimation problem, and base each decision on the models’ behavior in leave-one-out cross-validation forecasts. Additional cross-validation analysis indicates that this approach achieves performance comparable to or better than the best individual component.
Two important features of ILINet data to consider in models and forecast evaluation are 1. timeliness and accuracy of initial wILI values for each week and subsequent updates to these values, and 2. changes in behavior on and around major holidays. We examine these details of the data generation process, describe how they are addressed in the delta density model, and demonstrate the importance of considering the update procedure when performing retrospective evaluation and prospective forecasting.
During the 2015/2016 FluSight comparison, we submitted weekly, prospective forecasts from three forecasting systems:
Delphi-Stat: an adaptively weighted ensemble of instance-based statistical forecasting methods, and the topic of this paper;
Delphi-Archefilter: forms an empirical (rather than mechanistic) process model describing wILI trajectories, and incorporates both wILI and multiple forms of digital surveillance data using statistical filtering techniques [50]; and
Delphi-Epicast: wisdom-of-crowds approach based on combining predictions submitted by several human participants [72].
Our past and ongoing forecasts, as well as Python [73] and R [46] code for components of the systems used to generate them, are publicly available online [74–76]. Changes made to Delphi-Stat throughout the 2015/2016 season are described in S5 Appendix. These three forecasting systems were ranked as the top three in the 2015/2016 comparison in terms of overall multibin score, with Delphi-Stat at the top. Fig 4 shows the performance of the three Delphi forecasting systems, broken down by evaluation metric and forecasting target. S1 and S2 Figs. show the multibin scores broken down by location and by forecasting week. Delphi-Stat had consistently strong aggregate multibin scores across different targets, locations, and forecasting weeks, and the best overall multibin log score of all FluSight 2015/2016 submissions. Delphi-Stat’s unibin log score evaluations relative to the other two Delphi systems seem similar to or better than the corresponding multibin log score evaluations, as Delphi-Stat has the best unibin score of the three for each target rather than just overall; this observation seems natural since Delphi-Stat was developed to optimize unibin log score, and may suggest that optimizing for multibin log score rather than unibin log score when selecting ensemble weights or as a post-processing step could produce multibin log score improvements. However, the system’s point predictions, while optimized for the mean absolute error metric, were less accurate than (but still competitive with) the other two when averaged across all predictions.
Prospective forecast evaluation ensures that performance estimates are truly out-of-sample, not inflated by design decisions or model fits that are influenced by the evaluation data; however, such evaluation data is not readily generated, as it is expensive in terms of physical time: new wILI observations arrive once per week, and performance can vary significantly from season to season and from week to week. The evaluations from the 2015/2016 comparison may be noisy due to these season-to-season fluctuations. To address this issue, we use pseudo-out-of-sample retrospective analysis to provide more stable estimates of performance. Specifically, we use leave-one-season-out cross-validation: for each evaluation season s, we form and evaluate retrospective forecasts for s at every evaluation week using all training seasons except for s as inputs to the forecasting methods as if they were past seasons. (We exclude seasons prior to 2010/2011 from the evaluation set because records of HHS region ILINet data revisions are only available beginning in late 2009. We exclude seasons prior to 2003/2004 from the training set because year-round ILINet observations, which are required by some of the ensemble components, started in 2003. The 2009/2010 season—containing the peak of the 2009 pandemic according to our adjusted definition of “season”—is also removed from the training set. Finally, we do not include the season currently underway (S + 1) in evaluation or training as it has not been completely observed.) Using cross-validation prevents most direct model fitting to evaluation data, and basing design decisions on motivations other than the effects on cross-validation evaluation helps limit fitting through iterative design.
Fig 5 shows the distribution of log scores for several forecasting methods, described earlier in the text and in S1 Appendix, and the three ensemble approaches specified earlier in the text. Except for the uniform distribution and ensembles, all forecasting methods miss some possibilities completely, reporting unreasonable probabilities less than exp(−10) ≈ 0.0000454 for events that actually occurred. In these situations, the log score has been increased to the cap of −10 (as CDC does for multibin log scores). Delta and residual density forecasting methods (Delta density, Markovian; Delta density, extended; and BR, residual density) are less likely to commit these errors than other non-ensemble, non-uniform approaches, and have higher average log scores. Ensemble approaches combine forecasts of multiple components, missing fewer possibilities, and ensuring that a reasonable log score is obtained by incorporating the uniform distribution as a component. For the full Delphi-Stat ensemble, the main advantage of the ensemble over its best component appears to be successfully filling in possibilities missed by the best component with other models to avoid -10 and other low log scores appears, while for ensembles of subsets of the forecasting methods, there are other benefits; S3 Appendix shows the impact of these missed possibilities and the log score cap.
Fig 5 also includes estimates of the mean log score for each method and rough error bars for these estimates. We expect there to be strong statistical dependence across evaluations for the same season and location, and weaker dependencies between different seasons and locations; thus, the most common approaches to calculating standard errors, confidence intervals, and hypothesis test results will be inappropriate. Properly accounting for such dependencies and calibrating intervals and tests is an important but difficult task and is left for future investigation. We use “rough standard error bars” on estimates of mean evaluations: first, the relevant data (e.g., all cross-validation evaluations for a particular method and evaluation metric) is summarized into one value for each season-location pair by taking the mean of all evaluations for that season-location pair; we then calculate the mean and standard error of the mean of these season-location values using standard calculations as if these values were independent. Under some additional assumptions which posit the existence of a single underlying true mean log score for each method, these individual error bars—or rough error bars for the mean difference in log scores between pairs of ethods—suggest that the observed data is unlikely to have been recorded if the true mean log score of the extended delta density method were greater than that of the adaptively weighted ensemble, or if the true mean log score of the “Empirical Bayes A” method were greater than the extended delta density method. The mean and rough standard error estimates in Fig 5 also appear in tabular form in S4 Appendix.
Methods that model wILI trajectories and “pin” past wILI to its observed values have a large number of log scores near 0 because they are often able to confidently “forecast” many onsets and peaks that have already occurred; ensemble methods also have a large number of log scores near 0. Note that these scores are closer to 0 for ensembles that optimize weighting of different methods than for the ensemble with uniform weights. For this particular set of forecasting methods, targets, and evaluation seasons:
the uniformly weighted ensemble has lower average log score than the best individual component (extended delta density),
using the stacking approach to assign weights to ensemble components improves ensemble performance significantly and gives higher average log score than the best individual component,
the adaptive weighting scheme does not provide a major benefit over a fixed-weight scheme using a single set of weights for each evaluation metric.
When given subsets of these forecasting methods as input, with regard to average performance:
the uniformly weighted ensemble often outperforms the best individual, but is sometimes slightly (≈ 0.1 log score) worse;
the stacking approach improves upon the performance of the uniformly weighted ensemble; and
the adaptive weighting scheme’s performance is equal to or better than that of the fixed-weight scheme, sometimes improving on the log score by ≈ 0.1. The adaptive weighting scheme’s relative performance appears to improve with more input seasons, fewer ensemble components, and increased variety in underlying methodologies and component performance. These trends suggest that using wider RelevanceWeight kernels, regularizing the component weights, or considering additional data from 2003/2004 to 2009/2010, for which ground truth wILI but not weekly ILINet reports are available, may improve the performance of the adaptive weighting scheme. In addition to these avenues for possible improvement in ensemble weights for the components presented in Fig 5, the adaptive weighting scheme provides a natural way of incorporating forecasting methods that generate predictions for only a subset of all targets, forecast weeks, or forecast types (distributional forecast or point prediction). For example, in the 2015/2016 season, we incorporated a generalized additive model that provided point predictions (and later, distributional forecasts) for peak week and peak height given at least three weeks of observations from the current season.
Fig 6 shows a subset of the cross-validation data used to form the ensemble and evaluate the effectiveness of the ensemble method, for two sets of components: one using all the components of Delphi-Stat, and the other incorporating three of the lower-performance components and a uniform distribution for distributional forecasts. The Delphi-Stat ensemble near-uniformly dominates the best component, extended delta density, in terms of log score, and has comparable mean absolute error overall. The ensemble approach produces greater gains for the smaller subset of methods, surpassing not only its best components, but all forecasting methods in the wider Delphi-Stat ensemble except for the delta density approaches.
Fig 7 shows cross-validation performance estimates for the extended delta density method based on three evaluation schemes:
Ground truth, no nowcast: the ground truth wILI for the left-out season up to the forecast week is provided as input, resulting in an optimistic performance estimate;
Real-time data, no nowcast: the appropriate wILI report is used for data from the left-out season, but no adjustment is made for possible updates; this performance estimate is valid, but we can improve upon the underlying method;
Backcast, no nowcast: the appropriate wILI report is used for data from the left-out season, but we use a residual density method to “backcast” updates to this report; this performance estimate is valid, and the backcasting procedure significantly improves the log score;
Backcast, Gaussian nowcast: same as “Backcast, no nowcast” but with another week of simulated data added to the forecast, based on a Gaussian-distributed nowcast; and
Backcast, Student t nowcast: same as “Backcast, Gaussian nowcast” but using a Student t-distributed nowcast in place of the Gaussian nowcast.
Backcast, ensemble nowcast: same as the previous two but using the ensemble nowcast (which combines “no nowcast” with “Student t nowcast”).
For every combination of target and forecast week, using ground truth as input rather than the appropriate version of these wILI observations produces either comparable or inflated performance estimates.
Using the “backcasting” method to model the difference between the ground truth and the available report helps close the gap between the update-ignorant method. The magnitude of the performance differences depends on the target and forecast week. Differences in mean scores for the short-term targets are small and may be reasonably explained by random chance alone; the largest potential difference appears to be an improvement in the “1 wk ahead” target by using backcasting. More significant differences appear in each of the seasonal targets following typical times for the corresponding onset or peak events; most of the improvement can be attributed to preventing the method from assigning inappropriately high probabilities (often 1) to events that look like they must or almost certainly will occur based on available wILI observations for past weeks, but which are ultimately not observed due to revisions of these observations. The magnitude of the mean log score improvement depends in part on the resolution of the log score bins; for example, wider bins for “Season peak percentage” may reduce the improvement in mean log score (but would also shrink the scale of all mean log scores). Similarly, the differences in scores may be reduced but not eliminated by use of multibin scores for evaluation or ensembles incorporating uniform components for forecasting.
Using the heavy-tailed Student t nowcasts or nowcast ensemble appears to improve on short-term forecasts without damaging performance on seasonal targets. The performance of the nowcast ensemble is further explored in S5, S6, S7, S8, S9, S10, S11 and S12 Figs. The Gaussian nowcast has a similar effect as the other nowcasters except on the “1 wk ahead” target that it directly predicts: its distribution is too thin-tailed, resulting in lower mean log scores than using the forecaster by itself on this target.
Delphi-Stat forecasts submitted to the 2015/2016 comparison were based solely on wILI observations from the 2015/2016 “pre-season” (EW21–EW39) and season (EW40–EW20) and nonmechanistic models (with a majority of the ensemble weight assigned to the delta and residual density based methods). Additional data and mechanistic models dealing with categories of ILI or type and subtype of influenza, climate, digital surveillance, season-to-season patterns, spatial interaction, etc. were not incorporated. We do think that these types of data are useful, but analyzing their dynamics and effects on wILI is complicated by the fact that the smallest geographical units for which real-time wILI data is readily available for the entire US are HHS regions (on a weekly time scale). We expect that mechanistic components incorporating climate data and separating diseases, types, and subtypes will be more useful when we are able to model, forecast, and validate data at a finer geographical resolution, ideally at the metro area level. Similarly, we believe that digital surveillance data is useful; in fact, we currently use a sensor fusion framework to combine several such data sources and short-term forecasters to produce “nowcasts” for the current week [50, 51], and improve the performance of forecasting methods by incorporating these nowcasts in a manner similar to the ILINet-based backcasts.
During development and throughout this manuscript, we have focused on (thresholded) unibin log score as a (near-)proper, simple-to-implement metric for distributional forecasts. CDC FluSight organizers, on the other hand, selected exponentiated mean thresholded multibin log scores over the entire influenza season as the evaluation metric for forecast comparisons to 1. encourage high-quality distributional predictions rather than point predictions, for better understanding of the risk of certain scenarios, 2. make the scoring metric more accessible to policymakers than unibin and non-exponentiated variants, and 3. avoid −∞ scores due to a single forecaster mistake or unmodeled data revisions. We believe that it is up to policymakers to decide whether these forecasts are ready for use in decision support at the current level of accuracy. For other potential users and forecast comparisons, we provide absolute error evaluations for all targets in 2015/2016 in Fig 4 and S2 Appendix, as well as absolute error and percent absolute error for short-term targets from cross-validation in S5, S6, S7, S8, S9, S10, S11 and S12 Figs.
The delta density forecasting method, stacking-based adaptively weighted ensemble, distributional “backcasts” of wILI updates, and nowcasts from ILI-Nearby provide significant improvements upon other individual forecasting approaches that we considered. Promising avenues for further improvements include refining the methodology to rely less on arbitrary and heuristic feature, kernel, bandwidth, and parameter selections, regularization of ensemble weights, incorporating conditional density estimators from statistical literature, and using additional data sources and finer-resolution data models.
|
10.1371/journal.pcbi.1003402 | A Latent Markov Modelling Approach to the Evaluation of Circulating Cathodic Antigen Strips for Schistosomiasis Diagnosis Pre- and Post-Praziquantel Treatment in Uganda | Regular treatment with praziquantel (PZQ) is the strategy for human schistosomiasis control aiming to prevent morbidity in later life. With the recent resolution on schistosomiasis elimination by the 65th World Health Assembly, appropriate diagnostic tools to inform interventions are keys to their success. We present a discrete Markov chains modelling framework that deals with the longitudinal study design and the measurement error in the diagnostic methods under study. A longitudinal detailed dataset from Uganda, in which one or two doses of PZQ treatment were provided, was analyzed through Latent Markov Models (LMMs). The aim was to evaluate the diagnostic accuracy of Circulating Cathodic Antigen (CCA) and of double Kato-Katz (KK) faecal slides over three consecutive days for Schistosoma mansoni infection simultaneously by age group at baseline and at two follow-up times post treatment. Diagnostic test sensitivities and specificities and the true underlying infection prevalence over time as well as the probabilities of transitions between infected and uninfected states are provided. The estimated transition probability matrices provide parsimonious yet important insights into the re-infection and cure rates in the two age groups. We show that the CCA diagnostic performance remained constant after PZQ treatment and that this test was overall more sensitive but less specific than single-day double KK for the diagnosis of S. mansoni infection. The probability of clearing infection from baseline to 9 weeks was higher among those who received two PZQ doses compared to one PZQ dose for both age groups, with much higher re-infection rates among children compared to adolescents and adults. We recommend LMMs as a useful methodology for monitoring and evaluation and treatment decision research as well as CCA for mapping surveys of S. mansoni infection, although additional diagnostic tools should be incorporated in schistosomiasis elimination programs.
| Schistosomiasis remains one of the most prevalent parasitic diseases in developing countries, with Schistosoma mansoni being the most widespread of the human-infecting schistosomes. For the routine surveillance of human S. mansoni infection more “field-applicable,” sensitive, and cost-effective diagnostics that replicate faecal samples over several consecutive days [the Kato-Katz (KK) method], are needed. We propose a statistical modelling framework in order to evaluate the diagnostic performance of the urine strip test for Circulating Cathodic Antigen (CCA) and single-day double KK measurements over three consecutive days for the diagnosis of S. mansoni infection in two different age groups from Uganda pre- and post- praziquantel (PZQ) treatment. We demonstrate that CCA is an appropriate tool for mapping surveys of S. mansoni infection. Our findings should allow for evaluation of the risk of potential misinterpretation with regards to diagnosis of S. mansoni infection through CCA or KK in this endemic setting pre- and post- PZQ treatment as the numbers and infection intensities are brought down, bridging existing important gaps in schistosomiasis diagnostics research. More generally, the proposed statistical analysis can reveal important biological insights from other diseases without gold standard diagnostic tools whenever longitudinal data are available.
| Schistosomiasis is a debilitating parasitic disease in tropical and sub-tropical areas which has recently attracted increased focus and funding for control through large scale mass drug administration (MDA) of praziquantel (PZQ) [1]. However the ability of current control initiatives to operate cost effectively is reduced by technical limitations of currently available schistosomiasis diagnostics [2], [3]. With the recent resolution on schistosomiasis elimination by the 65th World Health Assembly, schistosomiasis diagnostics research for population-based assessment is critical with careful consideration given to how those tools might be used within disease elimination programmes [4]–[6].
At present the World Health Organization (WHO) recommends the Kato-Katz (KK) method as the standard tool for the qualitative and quantitative diagnosis of Schistosoma mansoni infection because of its assumed high specificity, relative simplicity in field conditions and attractive price. WHO also recommends MDA with PZQ to be delivered to community populations defined where KK surveys show an estimated prevalence of over 50% in school-aged children and to be delivered to children aged 6–16 years where the estimated prevalence is between 10% and 50% in this age group. However, it is well known that KK method from single stool samples, particularly at low infection endemicities and following PZQ MDA, can underestimate Schistosoma infection prevalence (and intensities) and thus overestimate cure rates [7]–[11], whilst even multiple slides over multiple days of stool sampling can influence specificity and overestimate prevalence [10]–[15]. The necessity for more “field-applicable”, sensitive and cost-effective diagnostics than the KK method, at least for the routine surveillance of human S. mansoni infection such as that inherent within mapping of at-risk populations has also been recently highlighted [16], [17]. Even more worrying, in some endemic regions microscopic stool samples examination is considered too logistically difficult in terms of personnel available for routine surveillance and therefore S. mansoni infection remains undetected and untreated in control programmes [18]. A promising diagnostic option is a urine strip test for Circulating Cathodic Antigen (CCA) which is a genus-specific glycan regurgitated by adult schistosome worms into the blood stream [19]. A number of cross-sectional studies evaluating CCA accuracy pre-treatment have stressed the need for further research assessing the potential role of this diagnostic assay at different stages of schistosomiasis control programs [15], [17], [20].
In this study, we assessed for the first time the diagnostic accuracy of CCA and of double KK faecal slides over three consecutive days for S. mansoni infection by age group at baseline and at two follow-up times post treatment which should be viewed as a proxy for low transmission areas. Because no true gold standard diagnoses are available, we developed and fitted latent Markov models (LMMs) to estimate diagnostic test sensitivities and specificities, the true underlying infection prevalence and the probabilities of transitions between infected and uninfected states [21]. LMMs - which are sometimes also referred to as latent transition models or regime-switching models – are used to analyze discrete-time longitudinal data where respondent observations contain measurement error. This approach defines the true states as categories (latent classes) of a dynamic latent (unobservable) variable within a statistical model. The Markov assumption is reflected in the model via transition probabilities which allow for correlation between a respondent's true state at times t−1 and t [22].
We analyzed a detailed dataset from a longitudinal cohort living along the shorelines of Lake Victoria in Uganda who received one or two doses of PZQ treatment at baseline. We demonstrated how the use of LMMs allows estimation of the ‘true’ prevalence of S. mansoni infection over time and the quantification of the additional benefit of a second PZQ dose in reducing re-infection levels by age group.
Ethical clearance was obtained from the Uganda National Council of Science and Technology and the study was also presented to the Danish National Committee on Biomedical Research Ethics in Denmark (Reference Number: UNCST: HS 59). Informed consent was obtained from individual adult participants but for children the parents or guardians consented on their behalf. Thereafter, each individual signed a consent form before any activity started. All information obtained from participants was kept confidential. Because some of the participants might have potentially received PZQ treatment recently through MDA, field survey assistants asked each participant detailed questions about previous treatment in order to exclude such individuals ever treated through MDA, although no such pre-treated individuals were identified in the current study.
We conducted our study in Musoli village, Mayuge district at baseline and nine weeks after treatment during 2005. Participants of this study were randomly allocated to one or two PZQ (Shin Poong Pharmaceuticals, Seoul Republic of Korea) doses at baseline (40 mg PZQ per kg body weight; double treatment group: two times PZQ 40 mg per kg body weight administered two weeks apart). The field survey assistants who delivered these treatments were not aware of the infection status of any participants at any time. First follow-up data collection was performed at nine weeks , and hence at a time aimed to assess cure rates where the risk of any eggs detected occurring as a consequence of reinfection was minimal. Second follow-up data collection was performed two years later during 2007. With regards to treatment after two years, this is part of the National programme to treat everyone living in an endemic area and thus a second MDA was offered after two years. The study location was selected as this is an area of Uganda known for perennial S. mansoni transmission, situated on the shore of Lake Victoria [23] where the community population is not targeted by the National Control Program [24], [25]. The community consists primarily of fishermen and their dependants with the lake being the only source of fresh water for them. Furthermore, infrequent use and/or availability of latrines leads to contamination of the lake especially near the shoreline, where there is underwater vegetation suitable for the aquatic intermediate host snails to thrive. In addition, only individuals >6 years of age were enrolled as very often this is the age from which schistosome-induced morbidity, in general, becomes evident.
Before any data collection took place, trained and experienced demographers conducted in 2005 a census of the village population. During this process the inhabitants in each household registered their relation to the head of the household, year of birth, gender, occupation, duration of residency in the village and tribal membership.
A stratified random sample for age and sex was then selected from the census data. Sample size calculations included detection of a significant difference of cure rates between the two treatment groups with reference to cure rates obtained along Lake Albert. A significance level of α = 0.05, a power of 90% with a drop-out rate of 40% over the two years of studies contributed to the calculation of 552 individuals but at baseline we managed to recruit 446 with full parasitological and CCA in urine data as described elsewhere [23], [26]. A further six cases were then eliminated from the analysis presented in this study due to missing data in number of treatments and their age.
Stool samples were collected on three consecutive days from each member of the cohort and examined for the presence of S. mansoni ova. Two duplicate slides of each stool sample were examined using the KK technique at each day. Each slide was read by two trained microscopists and any discrepancies resolved before results were recorded as eggs per gram (EPG) faeces. The results for all six slides were combined for the descriptive results presented in Table 1 as this is considered the best KK diagnostic performance scenario [13] while results of two slides of each single day are combined and incorporated in the statistical models' derived results (Tables 2A–3B) because in many studies only a single stool sample is analysed [25], [27]–[29].
Single urine samples were kept cold after collection and after return to the laboratory at the end of the day aliquoted and stored frozen. At baseline and nine weeks, CCA urine samples were kept in a freezer in Uganda for nine months before they were transferred frozen to Leiden. At the two year follow-up CCA urine samples were delivered to Leiden within four weeks of collection. In Uganda, all CCA urine samples were kept at −20 Celsius degrees. The freezer in which urine was stored was −18 to −20 degrees C. Results should not be affected as the antigen is very stable and its detection is not influenced by periods of being frozen, freeze thaw cycles, or even storage for weeks at room temperature. They were kept frozen during transport to the Department of Parasitology, in Leiden, The Netherlands, where the CCA urine assays were performed as previously described [19], [30]. Briefly, for the laboratory-based test, 25 µL of completely thawed and vortexed urine was added to a tube containing dried carbon conjugated antibody, along with 75 µL of buffer and mixed well. Test strips were added, and allowed to develop for 40 minutes. Strips were removed, allowed to dry, and read against a set of five Quality Control (QC) standards of 0, 10, 100, 1000 and 10 000 ng of semi-purified worm antigen (containing CCA) per ml negative urine. A score of 0 indicated a negative result; the 0.5 stands for trace, while 1, 2, and 3, indicated that the intensity of the test line was similar to that of the respective 100, 1000, and 10 000 ng/ml spiked QC samples. Strips were scored in a blinded fashion by at least two individuals and in case of discrepancies a third person was consulted to conclude on the score. The classification of the trace result is decided later (see ‘Model selection’ section) based on the use of specific latent variable models. All other positive results (scores 1, 2 and 3) were merged into one positive category.
LMM consists of a structural model for the latent infection states (analogous to the latent classes in Latent Class Analysis) and a measurement model for the observed indicators (these are the four diagnostic tools: the average of two KK measurements on three consecutive days and the CCA), conditional on latent infection state.
Let Y = (yi1t,…, yiPt) be a response pattern for the ith individual at time t on P observed binary indicators (in this study P = 4 binary diagnostic tests) with values ‘0’ and ‘1’ indicating negative and positive diagnostic test results, respectively.
We assume that for each individual the true underlying infection state, at each discrete time point t (where t = 1,…T and T = 3: baseline, nine weeks and two years) is explained by a latent categorical variable denoted by C with two latent infection states (i.e. those with S. mansoni infection and those without).
The responses to the (P×T) y indicators are assumed to be independent conditional on the latent infection state membership which in our analysis implies that the results from the four diagnostics are assumed to be independent conditional on the true underlying infection state both within and across time points. In addition, the latent categorical variable Ci,t depends on Ci,t-1 but not on earlier latent categorical variables, known as the first-order Markov property. Under these assumptions, the probability of observing a particular response pattern Y for a randomly selected individual i is:where represents the baseline infection state prevalence at the first time point (i.e. baseline), and represents the transition probability to latent infection state jt at time t conditional on membership in latent infection state jt-1 at time t-1. represents the diagnostic specificity for infection when the probability of the p diagnostic test is negative conditioned on Ci,t representing the ‘Not Infected’ latent state. Similarly, the diagnostic sensitivity for infection of each of the four diagnostics is obtained when the probability of the p diagnostic test is positive conditioned on Ci,t representing the ‘Infected’ latent state.
Figure 1 presents our LMM in a path diagram.
We fitted several LMMs using MPLUS v. 6.1 (Muthén & Muthén, Los Angeles) [31] with full information maximum likelihood in which we assumed that missing data were missing at random and making maximum use of data from individuals with incomplete data at the time points under study, for two different age groups: a) n = 167 children of age 7–16 years old and b) n = 273 adolescents and adults of age 17–76 years old. It is well known that different contact patterns with infected water bodies and consequently acquired exposure, immunity and susceptibility to infection might be experienced by different age groups [32].
We thus estimated the following sets of parameters for these two age groups:
We selected the LMMs that optimally combined goodness of fit and parsimony as measured using the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC). [33] Both AIC and BIC are penalized-likelihood criteria. AIC is an estimate of a constant plus the relative distance between the unknown true likelihood function of the data and the fitted likelihood function of the model, so that a lower AIC means a model is considered to be better. BIC is an estimate of a function of the posterior probability of a model being true, under a particular Bayesian framework, so that a lower BIC means that a model is considered to be better. BIC penalizes model complexity more heavily and thus, whenever there is an inconsistency AIC indicates a preference for a more complex model than BIC [34]. Both criteria are based on assumptions and asymptotic approximations. Each, despite its heuristic usefulness, has therefore been criticized as having questionable validity for real-world data [34], [35]. In the current study, we examined AIC and BIC as well sample-size-adjusted BIC among models that were considered biologically plausible for the epidemiological settings under study (given the effect of one and two PZQ doses on schistosomiasis cure rates within the studied age groups).
We first tested whether the interpretation of a ‘trace’ CCA test line should be classified as positive or negative as several studies have reported on the ambiguity of infection status among those classified as ‘trace’ [16], [17], [36], [37]. For this problem we used a one factor analysis model [38] on the baseline data, to explore the interrelationships among the four diagnostic tests for both age groups, treating the three results of CCA (‘negative’, ‘trace’ and ‘positive’) as nominal (unordered). We plotted the posterior distribution of the latent variable given the three possible responses to CCA for children (Figure 2) and adolescents and adults (Figure 3), after estimating the one factor analysis model. Having observed the similarities in the posterior distributions for the ‘trace’ and ‘negative’ categories in Figures 2 and 3, we decided to treat ‘negative’ and ‘trace’ CCA as a single ‘negative’ category for the rest of this analysis. Similar statistical analysis might be useful for data related to point of care-CCA in order to further validate results and proceed with generalization of recommendations about trace results for the CCA in the field.
We subsequently fitted LMMs where we initially tested whether the item response probabilities (ρ's, i.e. the sensitivities and specificities) assumed to be the same at baseline and two years but allowed to vary at nine weeks for both KK and CCA tests. In other words we tested the hypothesis of measurement invariance which assumes the equality of the parameters of the measurement model i.e. the conditional item response probabilities for the latent infection states at the different time points. We did not test for different ρ's at each of the three time points (baseline, nine weeks and two years) because if they did differ at each time point the meaning of the latent infection states would have changed over time making the transition probabilities τ's uninterpretable. The reason for this is that along with interpreting quantitative change over time in latent infection state membership (i.e. through the τ's), it also becomes necessary to interpret change over time in the meaning of the latent infection state (i.e. which latent infection state represents the ‘true’ negatives and ‘true’ positives at each time point).
We also tested whether the provision of one or two PZQ doses affected the transition probabilities τ's in the non-treatment intervals (nine weeks and two years) or only in the treatment interval (baseline to nine weeks) since there might be long-term benefits of two PZQ doses in the absence of snail control as in this community. Equal transition probability τ's across each of the treatment intervals under study was not tested because this was not consistent with prior knowledge of the MDA impacts.
Although there was information identifying the household of each individual in the study, there were not enough data within each of the age groups studied to take into account between-household variability in the considered models.
The observed numbers of “true” and “false positives” and “true” and “false negatives”, and corresponding observed sensitivity and specificity, for the CCA diagnostic test are presented assuming, for illustration, the combination of three duplicate KK measures over the three consecutive days as being both 100% sensitive and 100% specific (Table 1).
When the sensitivities and specificities of single-day double KK results were estimated by fitting LMMs, the information criteria indicated that, for both age groups, the transition probability matrices [for the baseline-to-nine weeks follow-up (treatment interval) and for the nine weeks-to-two years follow-up (non-treatment interval)] depended on the provision of one or two PZQ doses. Similarly, the measurement invariance hypothesis for CCA was accepted based on the information criteria; however, the measurement invariance hypothesis was rejected for KK (i.e. the diagnostic performance of KK was found to vary over time but that of CCA was not). For further details on final model selection, see Table S1. The estimated sensitivities and specificities for CCA and KK tests, from the best LMM for each age group are presented in Table 2. Based on these estimates and if one took the crude average of the estimated sensitivities and specificities of the two diagnostic measures and time points under examination, CCA overall was found to be more sensitive but less specific than double KK from a single faecal sample for both age groups. The sensitivity of double KK from a single faecal sample was found to be lower at nine weeks than at baseline and two years for both age groups in all three days for adolescents and adults and for two of the three days for children.
The estimated transition probabilities between the latent infection states over time are presented in Table 3. Diagonal elements represent the probabilities of remaining in the same latent infection state at time t as at the previous time (t-1). The majority of adolescents and adults apparently remained, over the non-treatment interval, in the same latent infection state. Among those infected at baseline in both age groups, two PZQ doses produced a higher probability of clearance of infection at nine weeks than just one PZQ dose; for children these probabilities were 0.717 (2 doses) and 0.458 (1 dose) (see Table 3A), while for adolescents and adults these probabilities were 0.846 (2 doses) and 0.561 (1 dose) (see Table 3B).
The S. mansoni prevalence estimate for each time point and each age group is presented in Figure 4 with 95% confidence intervals. Following treatment the prevalence of S. mansoni infection decreased dramatically between baseline to nine weeks for both age groups. Among children there is a substantial rebound by two years. These patterns are also reflected in the transition probabilities in Table 3.
Although visual examination of Table 4 superficially suggests that the LMM estimate conflicts with that obtained based on the 6 KK measurements (assumed positive if one or more of the six measurements was positive), further calculations are required to underpin full interpretation of the results.
For instance, for children at baseline, the estimated prevalence based on the LMM is 92.6%.
One would expect the following percentage of true positives:where 0.990, 0.938 and 0.960 are the estimated sensitivities of two KK measurements from days one, two and three, respectively
One would expect the following percentage of false positives:where 0.751, 0.641 and 0.709 are the estimated specificities of two KK measurements from days one, two and three, respectively.
Thus, based on the estimated sensitivities, specificities and the assumption that results from days one, two and three, were independent, conditional on the true status, the LMM predicts that the estimated prevalence based on 6 KK measurements would be 97.6% (92.6% true positive and 5.0% false positives). This is highly consistent with the estimate obtained from the 6 KK measurements: 97.6%.
For the remaining time points and age groups please see in Supporting information.
This study analysed a longitudinal detailed dataset from Uganda in which one or two doses of PZQ treatment were provided at baseline using LMM that accounted for the longitudinal study design and the measurement error in the diagnostic methods under study. Our primary objective was to assess the CCA diagnostic accuracy at baseline and at two follow-up times after treatment but we also evaluate double KK faecal slides over three consecutive days for S. mansoni infection. To our knowledge, this is the first study which provides rigorous model-based diagnostic performance of CCA and single-day double KK measurements over three consecutive days for the diagnosis of S. mansoni infection in two different age groups pre- and post- PZQ treatment.
CCA's diagnostic performance was found to be constant over time and overall approximately 90% sensitive but less specific than single-day double KK faecal slides for S. mansoni infection in both age groups. Day-to-day variation in faecal egg output among Schistosoma parasites [7]–[9] has been shown to be greater [11] and with lower sensitivity of KK after PZQ treatment [8]. The single-day double KK sensitivity is likely to depend strongly on the observed prevalence [8], [15], [39]. Our study confirmed these findings and arguments and showed clearly that sensitivity of single-day double KK was much lower at nine weeks than at baseline and two years for both age groups in all three days for adolescents and adults and for two of the three days for children while its specificity increased after PZQ treatment (Table 2).
These findings bridge existing gaps in schistosomiasis diagnostics research such as for instance the lack of CCA evaluation in adolescents and adults and the lack for evidence for its capacity to determine if a person has been cured after treatment, as previously highlighted [4], [17]. The current analysis provides model-based estimates of sensitivity and specificity and their uncertainties (through the provision of 95% confidence intervals) without assuming any gold standard diagnostic test in the statistical analysis. The exact numbers of false positive and false negative results are almost always unknown and thus in the current study we estimated rather than assumed values for the parameters displayed in Tables 2 and 3 (sensitivities, specificities and transition probabilities) [40], [41]. Without quantification of the uncertainties regarding the performance of the key diagnostic tests, generalization of epidemiological results and development of useful recommendations for which diagnostics to use and at which stages of schistosomiasis control are hampered.
This approach evaluates the risk of potential misinterpretation with regards to diagnosis of S. mansoni infection through CCA or KK in this endemic setting pre- and post- PZQ treatment as the numbers and infection intensities are brought down [12]. For instance, results in Table 1 demonstrate that by using 6 KK measurements over three days as the gold standard (i.e. assuming 100% sensitivity and 100% specificity), the resulting empirical estimates of CCA sensitivity and specificity are mistakenly shown to vary over time. We do not expect that the clearance of the antigen is influenced by treatment. Furthermore, the hypothesis of measurement invariance was not rejected based on information criteria for the fitted LMMs (see Text S1 and Table S1 in Supporting Information).
Glinz et al. 2010 [42] discussed possible reasons for false positive diagnoses from KK tests. Results from model in Table 2 clearly indicate that the specificity of single-day double KK measurements is lower than 100%. This means that despite highly qualified and skilled co-workers, contamination of stool sieves in the field and data entry errors cannot completely be avoided. Because our estimated specificities of single-day double KK measurements were less than 100%, the estimated ‘true’ S. mansoni prevalence (Figure 4) is lower at each time points than the estimated prevalence obtained assuming that any individual with positive results on one or more of the six KK tests conducted over three consecutive days was infected (Tables 1 and 4 ). This is in accordance with work on diagnostic performance of KK for animal schistosomiasis infection [14]. Previous work using stochastic models have demonstrated that the sensitivity of the KK would vary according to the number of stool samples provided [7]–[9] and the ‘true’ S. mansoni prevalence at baseline can be calculated using the De Vlas pocket chart [8]. For the children group, the chart is not applicable since the observed prevalence in this study is beyond the limits where the De Vlas model is valid and thus we cannot compare it with the estimates of ‘true’ prevalence from our model (Figure 4). For the adolescents and adults group however estimates of the ‘true’ prevalence (Figure 4) were not consistent with this chart because it was based on an assumption of 100% KK specificity.
The estimated transition probability matrices (Table 3) provide parsimonious yet important insights into the re-infection and cure rates in the two age groups. The cure rate was higher in adolescents and adults than in children following treatment. This can be explained by the fact that those infected in the older age group had lower burdens than the infected children and would be therefore more likely to become negative. From an immunological perspective view, it can be argued that the older age group are more likely to have developed protective immune response and are therefore more efficient in affecting and killing the worms [43], [44]. The quantification of the additional benefit of a second PZQ dose in reducing infection levels for both age groups was demonstrated by higher transition probabilities from infected to non-infected among those who received two PZQ doses compared to those who received one PZQ dose within the treatment interval. For the non treatment interval (i.e. between nine weeks and two years) there were no differences in reinfection or cure rates among those who received one or two PZQ doses. As the LMMs are estimated using an iterative algorithm, at each step of the algorithm, estimated values very close to 0 or 1 can create estimation instability and therefore are automatically from MPLUS fixed to 0 and 1 respectively in order to avoid non-convergence of the estimation algorithm. Consequently, such values should be treated with caution due to computational limitations in these categories during the model estimation. Finally, we recognize that the results of this study depend upon the assumption of conditional independence assumed by the models fitted here. Once one conditions on the latent infection state, we believe though that there are good reasons to assume that KK and CCA are independent as Schistosoma eggs and antigens are excreted through different routes in the human body, for instance.
To conclude, in the absence of a diagnostic gold standard this study has demonstrated that LMMs can be useful for the evaluation of available diagnostic tools for S. mansoni infection. More generally, we recommend LMMs to be used for the evaluation of diagnostic tests of other diseases without gold standard diagnostic tools whenever longitudinal data are available as such modelling permits questions about changes in true infection states and test the measurement invariance hypothesis of the diagnostic tests of interest over time - making them very useful tools indeed for control program M&E research [21]. Further work in evaluating the trace result and the ability of CCA to quantify intensity of infection is also warranted in the M&E of schistosomiasis control programs and dynamic latent factor models (such models would assume continuous hypothetical constructs or typologies-i.e. intensity of infection) might be appropriate statistical methods for the analysis of relevant data. Similar studies should be considered at other sites in order to build on our results. We found that the CCA diagnostic performance remained constant after provision of PZQ treatment and that the test is overall more sensitive but less specific than single-day double KK for the diagnosis of S. mansoni infection. In line with the results from our study and those of a recent multi-country cross-sectional study which showed that for lower S. mansoni intensity settings the CCA sensitivity was demonstrated to be higher than KK [17], we recommend that CCA to be used for mapping surveys of S. mansoni infection. As public health measures are aimed at the elimination of residual foci of schistosomes, data generated using diagnostics with high specificity will be required to avoid unnecessarily prolonging MDA and wasting scarce resources [5]. Detection of parasite-specific DNA [45], [46] or circulating anodic antigen in serum or urine [47] might present alternative opportunities in schistosomiasis elimination programs and further evaluations of these diagnostics merit attention.
|
10.1371/journal.ppat.1004499 | Restriction of Francisella novicida Genetic Diversity during Infection of the Vector Midgut | The genetic diversity of pathogens, and interactions between genotypes, can strongly influence pathogen phenotypes such as transmissibility and virulence. For vector-borne pathogens, both mammalian hosts and arthropod vectors may limit pathogen genotypic diversity (number of unique genotypes circulating in an area) by preventing infection or transmission of particular genotypes. Mammalian hosts often act as “ecological filters” for pathogen diversity, where novel variants are frequently eliminated because of stochastic events or fitness costs. However, whether vectors can serve a similar role in limiting pathogen diversity is less clear. Here we show using Francisella novicida and a natural tick vector of Francisella spp. (Dermacentor andersoni), that the tick vector acted as a stronger ecological filter for pathogen diversity compared to the mammalian host. When both mice and ticks were exposed to mixtures of F. novicida genotypes, significantly fewer genotypes co-colonized ticks compared to mice. In both ticks and mice, increased genotypic diversity negatively affected the recovery of available genotypes. Competition among genotypes contributed to the reduction of diversity during infection of the tick midgut, as genotypes not recovered from tick midguts during mixed genotype infections were recovered from tick midguts during individual genotype infection. Mediated by stochastic and selective forces, pathogen genotype diversity was markedly reduced in the tick. We incorporated our experimental results into a model to demonstrate how vector population dynamics, especially vector-to-host ratio, strongly affected pathogen genotypic diversity in a population over time. Understanding pathogen genotypic population dynamics will aid in identification of the variables that most strongly affect pathogen transmission and disease ecology.
| Co-infection, the presence of multiple genotypes of the same pathogen species within an infected individual, is common. Genotype diversity, defined as the number of unique genotypes, and the interaction between genotypes, can strongly influence virulence and pathogen transmission. Understanding how genotypic diversity affects transmission of pathogens that naturally cycle among disparate hosts, such as vector-borne pathogens, is especially important as the capacity of the host and vector to sustain genotypic diversity may differ. To address this, we exposed Dermacentor andersoni ticks, via infected mice, to variably diverse populations of Francisella novicida genotypes. Interestingly, we found that ticks served as greater ecological filters for genotypic diversity compared to mice. This loss in genotypic diversity was due to both stochastic and selective forces. Based on these data and a model, we determined that high numbers of ticks in an environment support high genotypic diversity, while genotypic diversity will be lost rapidly in environments with low tick numbers. Together, these results provide evidence that vector population dynamics, vector-to-host ratios, and competition among pathogen genotypes play critical roles in the maintenance of pathogen genotypic diversity.
| Genetic diversity within a single microbial species can lead to infection of hosts with mixtures of pathogen genotypes. Remarkably, studies across numerous systems have demonstrated that mixed-genotype infections are more common than infections with a single clonal variant [1]–[5]. The degree of genotypic diversity, defined here as the number of unique genotypes within a population, has been associated with pathogen transmission rates and virulence [6]–[9]. For example, greater numbers of circulating Plasmodium faliciparum genotypes were positively correlated with increased virulence or a greater probability of transmission [6],[8]. Competition experiments among Dengue virus serotypes resulted in the more virulent serotype being selected at the expense of less virulent serotypes during both human and mosquito infection [7]. Additionally, during the early years of West Nile virus circulation in New York, transmission intensity was associated with increases in viral genetic diversity [9].
The capacity of hosts to sustain multiple pathogen genotypes, and the within-host interactions among co-infecting genotypes, can impact pathogen transmission, virulence, and immune evasion. However, for pathogens that cycle among multiple host species, especially vector-borne pathogens that cycle between disparate species (mammals and arthropods), the impact of genotypic diversity and genotypic interactions on individual genotype transmission and infection success is largely unknown. Vector-borne pathogens, which cause diseases of importance for human and animal health, therefore provide a platform to study how genotypic diversity and interactions among genotypes affect colonization of the vector and resulting pathogen transmission.
Genetic diversity is a hallmark of vector-borne pathogens. Numerous studies have described the circulation and infection of individual hosts or vectors with multiple genotypes of bacterial (e.g., Anaplasma sp., Borrelia sp.), viral (e.g., West Nile virus, Dengue virus) or protozoal (e.g., Trypanosoma sp., Plasmodium sp.) vector-borne pathogens [2],[5],[10]–[17]. Competition among vector-borne pathogen genotypes within the mammalian host is common, with competitive success frequently achieved by the more virulent genotype [1],[4],[18]–[22]. For example, in experiments with P. falciparum and B. burgdorferi, the more virulent genotype replicated to greater levels compared to the competitor, resulting in numerical dominance and preferential transmission. Whether similar genotypic diversity-limiting competition occurs within the arthropod vector is unknown. Further, most studies examine the interactions of only two genotypes at a time; therefore, whether the degree of pathogen genotypic diversity influences the number of genotypes able to infect individual hosts and particularly individual vectors is similarly unknown.
Similar to other tick-borne bacterial pathogens, natural genetic variation within Francisella tularensis, including subspecies, is well described [23]–[28]. For example, using multiple loci variable-number tandem repeat analysis on only two loci, 10 unique F. tularensis genotypes were recovered from ticks; with the most genotypic diversity found in areas with the greatest prevalence of F. tularensis in ticks [23]. The large degree of circulating genotypic diversity observed in that study was indicative of long-standing enzootic transmission of multiple genotypes [23]. Additionally, unlike the majority of tick-borne bacterial pathogens which are refractory to genetic manipulation, F. tularensis subsp. novicida (herein referred to as F. novicida) can be genetically manipulated with relative ease, and thus can serve as a powerful model to address broader questions concerning tick-borne bacterial pathogens. Here, we used a set of differentiable Francisella novicida transposon mutants and Dermacentor andersoni ticks, which are a natural vector of Francisella sp. [29], to investigate how genotypic diversity affects the success of individual genotypes in colonizing the tick vector as compared to the mammalian host. Specifically, we determined (i) if similar numbers of genotypes were able to co-infect mice and ticks, (ii) whether exposure of hosts and vectors to differing numbers of genotypes affected the proportion of genotypes able to be recovered from the host or vector, and (iii) if competition limits the ability of certain genotypes to colonize the vector. To address these questions, pools of F. novicida genotypes of varying diversity were inoculated into mice. The genotypes able to infect mice, be acquired by feeding D. andersoni nymphs, and persist in the tick midgut through the molt to the adult stage at population and individual host and vector levels were identified. As the tick midgut is the primary site of colonization for most tick-borne pathogens, it serves as a relevant location to examine the effects that varying genotypic diversity has on individual genotype transmission success between host and vector [30],[31]. Finally, we designed a population model to demonstrate how variations in pathogen genotypic diversity, vector and host abundance, and vector-to-host ratios could influence the retention of genotypic diversity in a pathogen population over time.
We first determined whether the breadth of pathogen genotypic diversity is similarly sustained among mice and ticks at a population level. In all experiments ‘genotypic diversity’ refers to the number of different genotypes, the ‘vector’ refers to the tick and the ‘host’ refers to the mouse. The genotypes ‘available’ to colonize mice and ticks will refer to those genotypes that were inoculated into mice and those genotypes that were detected in terminal mouse blood during peak bacteremia, respectively. Our experiments were initiated by infection of mice, instead of ticks, because of the difficulty and more importantly the variability of artificially infecting ticks.
To simulate diverse genotype populations we used differentiable F. novicida transposon-containing genotypes in two large pools (Pool A = 93 genotypes, Pool B = 94 genotypes) each comprised of a different set of F. novicida transposon-containing genotypes (Table S1). Genotypes were identified in mouse blood at peak bacteremia (concurrent with completion of nymph feeding) and in adult tick midguts. Ticks fed as nymphs on infected mice over the entire duration of mouse bacteremia and genotypes were identified from the midgut of ticks after the infected nymphs molted to adults. This time point was specifically chosen to avoid detection of genotypes present in the undigested blood meal and confirm that any detected genotype(s) were able to infect and be transstadially maintained in the tick midgut. One limitation of this approach is that we were unable to determine if genotypic diversity was lost prior to or during early infection of the midgut or during transstadial transmission. Our readout of genotype success is colonization of the adult tick midgut, a time point which reflects the cumulative loss of genotypic diversity at any prior point during tick infection. Of the genotypes present in the large-pools, 84% and 81% of Pool A and Pool B genotypes were recovered from their respective mouse cohorts (Table 1). As these large pools encompassed genotypes with variable fitness, it was expected that some genotypes would not be recovered. Of the genotypes that successfully colonized mice, 76% and 54% of genotypes from Pool A and Pool B, respectively, were also acquired by the feeding nymph cohort and transstadially maintained in tick midguts (Table 1). The percentage of genotypes recovered from large-pools was significantly lower for ticks compared to mice (χ2 = 13.5, P = 0.0002). These results demonstrate that at a population level, despite simultaneous exposure to a large number of genotypes, not all available genotypes colonize mice and ticks. The inability of some in vitro generated genotypes to colonize mice was expected given the presence of the introduced transposon; however, the results also suggested additional loss of genotype diversity upon infection of the tick cohort.
To determine whether reducing genotypic diversity affected the recovery of genotypes from ticks during mixed-genotype infections, genotypes from pools A and B that had successfully infected mice but were not recovered from ticks were divided into three smaller pools (Pool C = 16 genotypes, Pool D = 17 genotypes, Pool E = 16 genotypes) and the experiment was repeated (Table S1). As expected, all of the genotypes in the small pools (Pools C–E) were recovered from their respective mouse cohorts (Table 1). Interestingly, 81, 88, and 94% of genotypes the from small-genotype pools C, D, and E, respectively, were recovered from their respective tick cohorts despite not being recovered from ticks during the large-genotype pool experiments (Table 1). Similar to the large-pools, the percentage of genotypes recovered from ticks was significantly lower compared to mice (χ2 = 6.39, P = 0.012). In summary, at a population level, a smaller proportion of available genotypes were recovered from ticks as compared to the mammalian host irrespective of the size of the genotype pool. Further, a greater proportion of available genotypes were recovered from ticks when genotypic diversity was reduced (χ2 = 9.30, P = 0.0023). These results support that at a population level, F. novicida genotype diversity is not equally sustained by mammalian hosts and tick vectors, and suggests that the latter serve as greater ecological filters for F. novicida diversity.
To determine if the observation that the greater reduction in genotypic diversity in the vector population compared to the mammalian host population was also reflected at the level of an individual, we identified the F. novicida genotype(s) that colonized individual mice and ticks. For example, if 59 genotypes were recovered from the population of ticks that fed upon mice inoculated with 93 genotypes in Pool A, we determined whether an individual tick was colonized by all or subsets of those 59 genotypes. In the large-genotype pool experiments, individual mice were colonized by a significantly greater percentage of the available genotypes (78 and 53% of the available genotypes in pools A and B, respectively, colonized individual mice) compared with individual ticks (12 and 10% of the available genotypes in pools A and B, respectively, colonized individual ticks) (χ2 = 707.4, P<0.001) (Figure 1). With regard to ticks in the large-genotype pool experiments, ticks were exposed to a mean of 62 genotypes while feeding on infected mice, and individual ticks were colonized with a mean of 8.5 genotypes (range = 1 to 25, median = 6.5) (Figure S1). These results indicate that the observed genotype diversity sustained by ticks at a population level was the cumulative product of individual ticks infected with subsets of the available genotypes.
To determine if reducing genotypic diversity affected the overall number or proportion of genotypes recovered we identified the genotypes that colonized individual mice and ticks from the small-genotype pool experiments. Similar to the large-genotype pool experiments, a significantly smaller proportion of the available genotypes colonized individual ticks (23, 29, and 21% from Pools C–E, respectively) compared to individual mice (100, 82, and 100% from Pools C–E, respectively) (χ2 = 227.5, P<0.0001) in the small-genotype pool experiments (Figure 1). In the small-genotype pool experiments overall, ticks were exposed to a mean of 14.3 genotypes and individual ticks were colonized by a mean of 4 genotypes (range = 1 to 11, median = 3.5) (Figure S1).
Examining genotype recovery from individual mice and ticks supported the population level genotype recovery results, and demonstrate that genotype diversity is most severely constrained in the tick. Further, the degree of genotypic diversity influenced both the mean number and proportion of genotypes that colonized ticks. Ticks exposed to more diverse F. novicida populations were colonized by a greater total number of genotypes (Z = 2.14, P = 0.033), but a smaller proportion of the available genotypes (χ2 = 44.8, P<0.0001) as compared to ticks exposed to less diverse genotype populations (Figure 1, S1).
To determine if the low number of genotypes colonizing ticks compared to mice was the result of a few dominating genotypes, the number of times each genotype was recovered from each tick and mouse was quantified. In general, individual genotypes were recovered from a greater proportion of mice than ticks (Figure S2, S3). On average an individual genotype was recovered from significantly fewer ticks in large pools (11%) compared to small pools (24%) (χ2 = 871.1, P<0.001). Thus the reduction in genotype diversity during tick infection was not the result of a small subset of genotypes infecting ticks at a greater frequency. Importantly, identification of different genotype combinations from individual ticks that fed upon similarly infected mice indicated that ticks were exposed to a wider array of genotypes then those that were recovered from an individual tick. Further, since ticks fed on mice during their entire duration of bacteremia (approximately 3 days), ticks were likely exposed to all or most or the genotypes identified in the terminal mouse blood. Therefore, the decreased genotype diversity observed in ticks is unlikely to be due to limited sampling opportunities or exposure to a limited number of genotypes.
The reduction in F. novicida genotypic diversity upon infection of ticks at both the population and individual level may reflect competition among genotypes. Alternatively, this reduction in diversity may be due to the inability of specific genotypes to infect the tick. To test these hypotheses, the only six genotypes (Genotype 1–6, Table S4) that were consistently recovered from mice but absent from ticks in pooled genotype experiments were further explored. First, we determined if each of these six genotypes, when inoculated individually into mice, were able to colonize feeding ticks. All six genotypes colonized both mice and ticks at infection levels (CFU/ml mouse blood or tick midgut) similar to wild-type with the exception of Genotype 3 that failed to colonize infect ticks (Figure 2A, B) (F5,40 = 0.88, P = 0.50). Moreover, with the exception of Genotypes 3, the other genotypes were recovered from a similar proportion of ticks as wild-type (P>0.30 for all comparisons) (Figure 2C).
As all of these genotypes, except Genotype 3, were competent to infect ticks, each was examined in 1∶1 competition experiments with wild-type to determine if a single additional genotype [wild-type] produced sufficient competition to result in competitive exclusion or suppression of the genotype of interest. In addition to wild-type, the competing genotype in all competition experiments was recovered from the terminal mouse blood (1.1×106, 2.1×107, 2.3×103, 1.9×105, 4.0×106, and 1.6×105 CFU/ml blood for Genotypes 1–6, respectively), thus confirming that ticks were exposed to the genotype of interest during feeding. The mean wild-type bacterial level recovered from terminal blood during competition with individual genotypes was 1.6×107 cfu/ml blood. During competition with wild-type, Genotypes 3, 4 and 6, which had the lowest bacteremia in mice, failed to colonize ticks (Figure 3A). The absence of Genotype 3 in ticks during competition was expected as, when alone, it resulted in a low bacteremia in mice and was not recovered from ticks (Figure 2). The absence of Genotypes 4 and 6 during competition with wild-type is indicative of competitive exclusion as these genotypes, when alone, had similar infection levels in mice and ticks compared to wild-type. When examined individually, both Genotype 4 and 6 were similar to wild-type in terms of both percent infected ticks (χ2 = 1.05, P = 0.30 for both comparisons) and midgut infection level (F2,26 = 0.18, P = 0.83) (Figure 2B, 2C).
Genotypes 1, 2, and 5 were able to colonize ticks during competition with wild-type (Figure 3); however, a smaller percentage of ticks were colonized by these genotypes compared with wild-type (χ2 = 3.60, P = 0.058, χ2 = 3.81, P = 0.051, χ2 = 3.53, P = 0.060 for genotypes 1, 2, and 5, respectively). In ticks colonized by Genotypes 1, 2, or 5, colonization by wild-type was also observed. Although wild-type could exclude Genotypes 1, 2, or 5 in individual ticks, none of these three genotypes excluded wild-type. Moreover, Genotypes 1 and 2 established significantly lower infection levels in the tick midgut compared to wild-type indicating that these two genotypes were competitively suppressed by wild-type (Figure 3B) for Genotype 1 and wild-type, t13 = 2.59, P = 0.023; for Genotypes 2 and wild-type, (t12 = 3.87, P = 0.0022). Interestingly and despite a lower colonization prevalence compared to wild-type, Genotype 5 achieved infection levels in the tick midgut similar to wild-type (Figure 3) (t15 = 0.42, P = 0.68).
As a control to demonstrate that wild-type specifically out-competed Genotypes 1–6, we performed a 1∶1 competition assay with wild-type and Genotype 7, which has a transposon in a non-coding region (isftu-2), and behaves similarly to wild-type in both mice and ticks [29]. In the terminal mouse blood the bacterial levels for wild-type and Genotype 7 were 9.3×106 and 7.0×106 CFU/ml blood, respectively, confirming ticks were exposed to both genotypes. Equal proportions of ticks were colonized by Genotype 7 and wild-type together, Genotype 7 alone, and wild-type alone. In ticks that were co-infected, both Genotype 7 and wild-type achieved similar infection levels in the tick midgut (Figure 3) (t6 = 0.25, P = 0.81). The equal success of Genotype 7 and wild-type in colonizing ticks during competition with one another demonstrated that Genotypes 1–6 were diminished or excluded due to competition rather than random effects. In summary, these results indicate that Genotypes 1–6 have a fitness disadvantage in the vector as compared to wild-type as co-infection of any of these genotypes with wild-type results in their competitive exclusion (e.g., Genotype 3, 4, and 6) or competitive suppression (e.g., Genotypes 1, 2, and 5). This demonstrates that co-infection with a single, more fit genotype is sufficient to alter the success of the competing genotype even if the less fit competitor is competent upon single-infection. Further, both competitive suppression and competitive exclusion offer explanations for the loss of genotypic diversity observed during pathogen infection of ticks.
Our experiments suggest that pathogen genotypic diversity is restricted within the tick vector at both population and individual levels. This restriction in diversity is most pronounced within individual ticks, suggesting that the abundance of ticks will strongly affect pathogen genotypic diversity within an environment. To further explore how variations in vector and host populations influence pathogen genotypic diversity, we developed a simple population model that incorporated data from our experiments. The model contained separate functions for vectors (ticks), hosts (mice), and pathogens (Francisella genotypes) (Figure S4). We used this model to investigate how vector-to-host ratios, vector and host abundance, and the initial number of pathogen genotypes within a population influenced the overall maintenance of genotype diversity in the population.
With all model conditions, individual mice harbored greater pathogen genotypic diversity than ticks (Figure 4, S5, S6). Thus, rare pathogen genotypes were more likely to be lost from the vector population than from the mammalian host population. At the population level, vector-to-host ratios strongly influenced the retention of pathogen genotypic diversity (Figure 4). When vector densities declined and vector-to-host ratios approached 1, pathogen genotypic diversity rapidly declined as individual genotypes were lost from the system. In contrast, high vector-to-host ratios increased the retention of genotypic diversity because the filtering effects of individual ticks were reduced due to large population sizes (Figure 4). Variation in vector or host abundance did not influence pathogen genotypic diversity as strongly as vector-to-host ratios; however, in general, larger vector and host populations led to greater maintenance of pathogen genotypic diversity (Figure S5). Initial pathogen genotypic diversity also influenced the number of pathogen genotypes maintained in the vector and host populations (Figure S6). Not surprisingly, both vectors and hosts individually harbored more pathogen genotypes when the number of initial genotypes was greater. However, the proportion of genotypes in the population infecting individual vectors and hosts declined with greater initial pathogen genotypic diversity (Figure S6) as observed in our experiments with large- and small-genotype pools (Figure 2). Thus, the model showed a trade-off between the raw number of pathogen genotypes that infected individual vectors and hosts and the proportion of the pathogen genotype population they represented.
Our results demonstrate that pathogen genotypic diversity is restricted to a greater degree in the tick vector as compared to the mammalian host. Moreover, the extent to which hosts and vectors contribute to the maintenance of pathogen genotypic diversity is influenced by the initial degree of genotypic diversity in the pathogen population and competition among genotypes during infection of the vector. Within a host or vector, competitive interactions among genotypes can result in the reduction or elimination of one or more genotypes. Studies on co-infecting Plasmodium genotypes illustrate how a more virulent genotype can competitively suppress or prevent a less virulent genotype from being transmitted between the mammalian host and mosquito vector [32]–[35]. Additionally, studies on arboviruses such as West Nile virus and Dengue virus have demonstrated that genotypic diversity can be correlated with transmissibility or virulence [7],[9]. Similar to our results, intrahost examination of West Nile virus revealed that viral genetic diversity was restricted in mosquito midguts compared to the input pool [36],[37]. Interestingly, however, despite a reduction of viral diversity in the mosquito midgut, corresponding salivary samples were similar in diversity to the input pool, perhaps contributed to accumulation of mutations as a result of relaxed purifying selection during infection of the mosquito [36].
In this study, F. novicida genotype diversity was not equally sustained by mice and ticks, and the greatest restriction in genotypic diversity occurred in individual ticks. This reduction in diversity was mediated by a combination of both stochastic and selective forces, and was unlikely to be an artifact of tick feeding. Based on our results, despite exposure to a large array of mutants, individual ticks were not able to support the same number of mutants as mice. One possible reason for genotypic restriction is that resources for bacterial colonization, such as nutrient availability or receptors for cell entry, are more limited in ticks than in mice, which could lead to competition among genotypes for limited resources. Several lines of evidence suggest that strong competition among genotypes occurred in ticks. First, individual ticks that fed upon the same mouse infected with up to 94 genotypes were colonized by different combinations of genotypes. Second, five genotypes not recovered from ticks during pooled-genotype experiments were competent to colonize ticks, in most cases to wild-type levels, in the absence of a second genotype. Third, in competition assays with wild-type and a wild-type-like genotype (Genotype 7), both were equally able to compete and colonize ticks, which further indicated that the absence of Genotypes 1–6 from the pooled-genotype experiments was not random. Our experimental design allowed us to examine the genotypic diversity that was sustained by ticks from the genotypes present during the nymphal blood meals to recovery of genotypes from adult tick midguts. This period of time encompassed several points where genotypic diversity could have been lost in the tick midgut including during initial entry into the nymph midgut, early replication and colonization events in the nymph midgut, transstadial transmission from nymph to adult, or continued colonization in the adult midgut. Although our results clearly demonstrate that competition is occurring among F. novicida genotypes during infection of the tick vector, it is interesting to note that a previous study speculated that facilitative interactions among genotypes in mixed B. burgdorferi genotype infections conferred an advantage for the bacteria to establish and maintain infection in ticks [15],[38]. It is possible that such interactions may occur in this system.
Additional variables that could further influence competition among genotypes and contribute to the observed reduction in genotype diversity in ticks include the infection level for an individual genotype, transmission priority (the order in which genotypes are transmitted), and genotype fitness. With regard to the latter two variables, our results indicated that reduction in fitness can in some instances overwhelm the stochastic forces that dictate tick infection by pathogen genotypes [29],[39],[40]. Overall, F. novicida bacterial levels did not vary based on genotype diversity and were similar to previously reported single-genotype infection levels [29]. This suggests that ticks have an infection threshold limit for F. novicida, such that as the number of genotypes a tick is exposed to increases, the maximum infection level of any individual genotype is proportionately reduced [29]. Therefore, genotypes that are able to replicate first will have a greater opportunity to colonize the tick while reducing the amount of available resources for incoming genotypes (founder effect) [12]. Additionally, greater numerical success in one host or vector will confer a greater probability of subsequent transmission.
The transmission priority of genotypes between mice and ticks was stochastic, such that ticks had an opportunity to acquire the genotypes that colonized mice relative to the genotype-specific infection level in mice. Transmission priority is potentially important if resources are more limited within the tick and monopolized by genotypes on a “first come, first serve” basis. In most pooled-genotype experiments, genotype recovery was random and ticks were colonized by small subsets of the available genotypes in different combinations. Although we strived to initiate our pooled-genotype experiments with equal ratios of genotypes, four genotypes in Pool B were recovered from a greater percentage of mice and ticks implying that they had a numerical advantage in the initial inoculum, maintained that advantage while colonizing mice and were available at a greater frequency for feeding ticks to acquire (Figure S2). These four genotypes, which were recovered from a greater percentage of ticks than the other genotypes comprising Pool B, provide evidence for genotypes with an initial advantage having greater transmission and colonization success. Importantly, although these four genotypes were identified in a greater percentage of ticks, they were not the sole genotypes observed and were commonly identified in individual ticks with less frequently occurring genotypes. These results are similar to those of a Trypanosoma brucei study and recently a B. burgdorferi study where vector acquisition of genotypes from mice, infected with multiple, similarly fit pathogen genotypes, was noted as random and the first genotype able to infect an individual vector had an advantage during dissemination to other tissues and in subsequent transmission [12],[41]. Stochastic forces also play a prominent role in shaping arboviral transmission, and has been demonstrated for West Nile Virus and Venezuelan equine encephalitis virus [42],[43].
Genotype fitness can influence competitive ability as well as virulence as demonstrated by co-infection studies using genotypes with known fitness differences [33]–[35]. A range of fitness among the F. novicida genotypes examined was expected, depending on the location of the transposon. The overall genetic similarly of genotype populations suggests that the majority likely shared similar abilities to infect mice and ticks (Table 1). We surmised that the six genotypes absent from ticks in the pooled-genotype experiments were out-competed. This postulation was supported by the results of the 1∶1 competition assays between wild-type and Genotypes 1–6, where wild-type succeeded disproportionately in terms of infection prevalence and infection load compared to the competing genotype. The competition assay between wild-type and Genotype 7 confirmed that if Genotypes 1–6 had been similarly fit as wild-type, they would have succeeded to a similar extent as Genotype 7 did during competition with wild-type. The finding that Genotypes 1–6 were able to colonize ticks during single-genotype experiments but not in during competition with more fit genotypes supported the notion that the location of the transposon in these genotypes exacts some fitness cost, although the exact mechanism by which this is occurring remains unknown.
In the field, genotypic diversity is likely to be dynamic and heavily influenced by environmental variables. Genotypic diversity, when measured, generally occurs as insertions, deletions, and polymorphisms in individual and small numbers of nucleotides [44]–[46]. Additionally, gene duplications and deletions do occur [47]. While insertions are over-represented in our population, the use of naturally occurring genotypes is not possible, as a collection of greater than 150 different genotypes that can easily be distinguished one from another do not exist for any tick borne bacterial pathogen. Importantly, the alterations in phenotype in our population are likely highly variable and represent a broad spectrum, from complete knock-out of gene function to no alteration in gene function. Thus, while the type of genetic mutation represented in our population is limited as compared to a natural population, a broad spectrum of alterations in phenotype is likely to be represented. Further, in our experiments more mutational robustness was observed in the vertebrate, however, within a host infected with naturally occurring genotypes those genotypes could possess very different fitness abilities, thus altering the outcome of within-host interactions and ongoing transmission.
To extrapolate our results to a broader range of field scenarios we created a model to explore how variations in vector-to-host ratio, vector and host abundance, and initial pathogen genotypic diversity affected the retention of pathogen genotypic diversity in a population over time. These variables were selected because our experimental data indicated that the greatest restriction in F. novicida genotype diversity occurred during colonization of ticks compared to mice. We assumed that there was no mortality of vectors and hosts, and thus the model likely over-estimated the conservation of diversity (as pathogen genotypes might be lost from dying vectors and hosts). Our modeling results suggested that local extinction of pathogen genotypes, and genotypic diversity overall, is more likely to be affected during pathogen infection of ticks. Vector-to-host ratio was the most important variable in the maintenance of pathogen genotypic diversity over time in a population; however, abundance of vectors and hosts, and initial pathogen genotypic diversity also contributed. Finally, our model was conservative in design in that it assumed equal fitness among genotypes, that all ticks fed, and does not incorporate the addition of new genotypes beyond those initially present. If additional values are known, derivations of this model could be used to examine these variables which could result in accelerated specific genotype extinction or retention. Although our model featured a tick-borne pathogen, our experimental results and model predictions are in line with epidemiological data of other vector-borne pathogens, including Plasmodium spp., where areas of high transmission are associated with abundant vector populations that collectively support a great diversity of pathogen genotypes [6].
Most vector-borne pathogen studies examining genotype co-infection to date either survey the circulating pathogen genotypes in an area or conduct competition assays among pairs of genotypes, frequently differing drastically in fitness (e.g., attenuated versus virulent, transmissible versus not transmissible) [2],[5],[11],[15],[19],[48]–[50]. Knowledge gaps exist regarding the role of the vector in supporting or restricting pathogen genotype diversity in a population. In this study, both the experimental data and population modeling data revealed that the tick vector acted as a greater ecological filter for pathogen genotypic diversity compared to the mammalian host. This restriction of F. novicida genotypic diversity in ticks was further affected by the initial amount of genotypic diversity and competition among genotypes. Extrapolation of our results in a model revealed variables, including vector-to-host ratio, which over many generations played important roles in the maintenance of pathogen genotypic diversity.
The marked reduction in genotypic diversity within the tick indicates that intervention strategies targeting the pathogen within the tick, such as introduction of highly competitive genotypes, are likely to be effective in disrupting disease transmission. Further, understanding how pathogen genotypic diversity and genotype interactions within the host and vector affect colonization success is essential to understanding pathogen transmission, selection and disease ecology.
This study was carried out in accordance with the following: Animal Welfare Act (9 CFR Ch. 1 Subpart C 2.31 (c) (1–8)), Guide for the care and use of Agricultural Animals in Agricultural Research and Training (Chap.1), and the Public Health Service Policy on Humane Care and Use of Laboratory Animals (Section IV.B. (1–8)). All protocols involving the use of animals were approved by the Washington State University Institutional Animal Care and Use Committee (IACUC) (ASAF Number: 3686 and 4430).
Dermacentor andersoni (Reynold's Creek) nymphs were obtained from a colony maintained by USDA-ARS-ADRU (Pullman, WA). All nymphs were fed on C57BL/6 mice [29]. After inoculation with F. novicida, mice were monitored twice daily for signs of illness. At the onset of severe illness (ruffled fur, hunched posture, ocular or nasal discharge, ataxia, etc), mice were euthanized and blood was cultured to determine bacteremia as described in the following section. If mice did not develop disease, they were euthanized upon completion of nymph feeding and bacteremia similarly determined via culture. In some experiments, adult D. andersoni were fed to repletion on male New Zealand white rabbits (Western Oregon Rabbit Company, Philomath, OR) [29]. Rabbits remained asymptomatic and were culture negative at the end of tick feeding.
Wild-type F. novicida (U112) or transposon mutants containing a kanamycin resistance cassette [51] (Table S1) were used in all experiments. All F. novicida mutant genotypes were cultured in tryptic soy broth (TSB) or on tryptic soy agar (TSA) containing 0.1% L-cysteine and kanamycin (15 µg/ml) (kanamycin was omitted when culturing wild-type F. novicida) [29]. Briefly, F. novicida broth cultures were incubated at 37°C and 225 rpm either overnight or for 3 hr, depending on the experiment. F. novicida agar cultures were incubated at 37°C for 48 hrs and the resulting colony forming units (CFU) enumerated. To recover F. novicida from blood, whole blood was plated from individual mice. To recover F. novicida from ticks, tick midguts were individually dissected and homogenized in Lysing Matrix H tubes (MP Biomedical, Solon, OH) containing 500 µl of 1× PBS for 13 seconds at 3 M/s and plated.
For all samples, genomic DNA (gDNA) was isolated from lawns (>10,000 CFU) of F. novicida culture using a DNeasy kit (Qiagen, Valencia, CA). Individual F. novicida genotypes were identified by PCR amplification of a 350–700 bp fragment using a universal primer located within the kanamycin cassette and a genotype-specific primer in the adjacent sequence (Table S2). Individual reactions included 2× GoTaq Mastermix (Promega, Madison, WI), 2.5-µM of each primer, and 50-ng of gDNA template. Thermocycler conditions were as follows: Step 1 (×1), 94°C for 2 min; Step 2 (×35), 94°C for 45 sec, 54°C for 45 sec, 72°C for 45 sec; Step 3 (×1), 72°C for 5 min. Following electrophoresis, PCR products were visualized on a 2% agarose gel containing SYBR Safe (Invitrogen, Carlsbad, CA).
To simulate diverse genotype infections, clones from two randomly chosen plates (NR-8058 and NR-8065, BEI Resources, Manassas, VA) from a F. novicida transposon mutant library were used to assemble the large pools (Pool A, n = 93; B, n = 94) for infection assays (Table S1). The populations of genotypes that comprised the small pools (Pool C, n = 16; D, n = 17; E, n = 16) (Table S1), were those that were recovered from the mouse blood but not the tick midgut in the large pool infection assays.
To generate diverse inocula, glycerol stocks of individual F. novicida genotypes were each inoculated into a single well in a 96-well plate containing 1 ml of TSB and grown overnight. Overnight cultures were sub-inoculated into fresh TSB for a starting concentration of 1∶1500 (1 µl overnight culture into 1.5 ml of TSB). Cultures were incubated for 3 hr and 50 µl of individual genotype cultures were combined to generate the mixed genotype inocula. An OD600 measurement was obtained for the combined culture and the appropriate dilutions were made in 1× PBS for a final concentration of 4000 CFU (∼40 CFU/genotype) or 1000 CFU (∼60 CFU/genotype) in 100 µl for the large- and small-pool infection assays, respectively.
Mice infested with D. andersoni nymphs were intraperitoneally inoculated (6 mice/pool and 3 mice/pool for the large and small pool infection assays, respectively). To verify that all genotypes were present in the inoculum, an aliquot was plated, allowed to grow to a lawn (>10,000 CFU) and re-suspended in 5 ml 1× PBS, from which 100 µl was used for gDNA extraction, and examined by genotype-specific PCR. Terminal mouse blood was plated and the resulting bacterial cultures examined to determine the bacterial load and identify the genotypes that successfully infected the mice and thus were available for the feeding nymphs to acquire. After feeding, nymphs were incubated at 25°C and allowed to molt to adults. For the large pool infection assays, the infected adult ticks were fed on a naïve rabbit to expand the F. novicida infection load in the tick midgut; however, we later determined this extra feeding was not necessary to detect the population of genotypes in the tick midgut and was omitted in subsequent infection assays. Once molting to adults was complete, midguts were dissected from individual ticks, homogenized and plated, and the resulting bacterial lawns were examined to determine the bacterial level and identify the genotypes that had colonized the tick midgut and had been transstadially maintained. Bacterial lawns derived from blood (n = 3) or midgut cultures (n = 10 to 12) were processed as described above from individual mice or ticks and the F. novicida genotype population determined from individual or pooled (combine aliquots of re-suspended culture from like inoculated/exposed mice or ticks) samples.
In the infection assays using multiple genotypes, described above, six transposon-containing genotypes were consistently recovered from the mouse blood but not the tick midgut. These genotypes were then tested in individual infection assays and competition assays with wild-type. As a control for the competition assays, a transposon-containing genotype that has a phenotype similar to that of wild-type [29] was used. For individual genotype infection assays, the inoculum was prepared as previously described with each mouse receiving 1000 CFU of a single genotype. Detection of the individual genotype in the inoculum, mouse blood, and tick midgut was accomplished by culture and the identity of the genotype was verified by PCR.
For competition assays, a 1∶1 ratio (500∶500 CFU) of two different genotypes were injected into mice in the same manner as described above. To enumerate each genotype within blood or midgut, CFUs were calculated by dual plating samples on antibiotic-free and kanamycin-containing TSA plates. This allowed enumeration of the transposon-containing genotype (CFU on kanamycin-containing TSA plates) and wild-type (CFU on antibiotic-free plates minus CFU enumerated on the reciprocal kanamycin-containing TSA plates). The ratio of wild-type to transposon-containing genotype was determined for each competition assay in the inoculum, mouse blood, and adult tick midgut. Six to twelve ticks were assessed for each competition pairing.
We developed a population model that incorporated data from the experiments to explore how variation in vector and host populations would influence pathogen genotypic diversity over a range of vector and host conditions. The model contained functions for vectors (ticks), hosts (mice), and pathogens (Francisella genotypes) (Figure S5). The model was initiated by allocating pathogen genotypes to a population of mice, with each mouse receiving all pathogen genotypes. These pathogen genotypes were then tracked over time in both tick and mice populations. The model had a generational time step, and at each time step uninfected ticks attached to mice, fed, and acquired pathogens. Infected ticks then molted and fed on uninfected mice (i.e., the next generation), transmitting pathogens in the process (Figure S5).
The model was individual-based, such that each tick only acquired pathogens from the mouse it fed on; similarly, mice only acquired pathogens from ticks that fed upon them. The probability that an uninfected tick acquired pathogen genotype g from mouse m was:(1)where is the number of pathogen genotypes harbored by mouse m. Thus, the maximum probability of a tick acquiring any genotype was 28%. The model was stochastic, and a random number was drawn from a uniform distribution between 0–1 and compared with to determine whether ticks acquired each pathogen genotype. In turn, the probability that an uninfected mouse acquired a pathogen genotype g from ticks was as follows :(2)where is the number of pathogen genotypes harbored by tick t. The summation adds up the total number of genotypes for all ticks that fed on each particular mouse (total = n). Thus, the maximum probability of a mouse acquiring any genotype was 100%. Like ticks, mice were modeled individually and the pathogen genotypes they acquired were stochastic. Equations for Pt and Pm were generated by fitting linear model to data from the experiments.
One limitation of the model is that using linear functions sets a maximum acquisition value that may be lower than the probability for “fit” genotypes. However, such functions were used to approximate the average genotype. This was done because it is difficult to assume the proportion of genotypes that would be “fit” (i.e., have a higher acquisition probability) and “unfit” (i.e., have a lower acquisition probability) in natural populations; therefore, we only modeled the “average genotype”. We did explore alternative forms of the acquisition function with a greater maximum acquisition value. However, with any form of the model our qualitative results on the role of vector-to-host density, initial vector and host abundance, and initial pathogen diversity did not change. Thus, we only present results of this simple model that did not distinguish between genotypes in terms of fitness. While simple, results with this model were used to demonstrate how diversity might be maintained in natural population with varying conditions.
In the baseline set of simulations, there were 100 mice, 1,000 ticks and 100 pathogen genotypes. However, these values were varied in sensitivity analyses to investigate the effects of different vector-to-host ratios, differences in vector and host abundance, and different initial pathogen genotypic diversity on the maintenance of pathogen genotypic diversity over time (Table S3). For each set of initial conditions, the model was run for 100 generations to examine the maintenance of pathogen genotypic diversity over time in ticks and mice. For each set of initial model conditions (Table S3), we ran the model 1,000 times to account for the stochastic nature of the model. Results presented represent the average values from these 1,000 simulations.
All statistical analyses were conducted using JMP Statistical Discovery Software Version 11 (Cary, NC). We used logistic regression to explore effects of genotype diversity (i.e., pool size), genotype group nested within pool size, and host (mouse vs. tick), and all two-way interactions on the recovery of genotypes from hosts and vectors. Genotype group was not significant in these analyses, (χ2 = 3.36, P = 0.34), and so final analyses were only run with the factors genotype diversity, host, and their interaction. In these analyses, the number of genotypes recovered or not recovered from hosts and vectors were binomial count data. To look at the number of genotypes recovered from individual ticks in differing pool sizes (large vs small) we used non-parametric Wilcoxon tests, as data on the number of genotypes recovered were non-normal. For single wild-type or genotype infection assays, we used an ANOVA to compare F. novicida genotype bacterial levels to wild-type bacterial levels during single-genotype infection experiments. For 1∶1 competition experiments between wild-type and a select genotype, we used two-sample t-tests to compare wild-type and genotype bacterial levels in tick midguts. Moreover, we used Fisher's exact tests to determine the proportion of ticks that were infected with each genotype compared to wild-type in these 1∶1 competition assays. For all analyzes, an α value of 0.05 was used to determine statistical significance.
|
10.1371/journal.ppat.1005357 | Characterization of RyDEN (C19orf66) as an Interferon-Stimulated Cellular Inhibitor against Dengue Virus Replication | Dengue virus (DENV) is one of the most important arthropod-borne pathogens that cause life-threatening diseases in humans. However, no vaccine or specific antiviral is available for dengue. As seen in other RNA viruses, the innate immune system plays a key role in controlling DENV infection and disease outcome. Although the interferon (IFN) response, which is central to host protective immunity, has been reported to limit DENV replication, the molecular details of how DENV infection is modulated by IFN treatment are elusive. In this study, by employing a gain-of-function screen using a type I IFN-treated cell-derived cDNA library, we identified a previously uncharacterized gene, C19orf66, as an IFN-stimulated gene (ISG) that inhibits DENV replication, which we named Repressor of yield of DENV (RyDEN). Overexpression and gene knockdown experiments revealed that expression of RyDEN confers resistance to all serotypes of DENV in human cells. RyDEN expression also limited the replication of hepatitis C virus, Kunjin virus, Chikungunya virus, herpes simplex virus type 1, and human adenovirus. Importantly, RyDEN was considered to be a crucial effector molecule in the IFN-mediated anti-DENV response. When affinity purification-mass spectrometry analysis was performed, RyDEN was revealed to form a complex with cellular mRNA-binding proteins, poly(A)-binding protein cytoplasmic 1 (PABPC1), and La motif-related protein 1 (LARP1). Interestingly, PABPC1 and LARP1 were found to be positive modulators of DENV replication. Since RyDEN influenced intracellular events on DENV replication and, suppression of protein synthesis from DENV-based reporter construct RNA was also observed in RyDEN-expressing cells, our data suggest that RyDEN is likely to interfere with the translation of DENV via interaction with viral RNA and cellular mRNA-binding proteins, resulting in the inhibition of virus replication in infected cells.
| Dengue is the most common arthropod-borne viral infection and is spreading to new areas every year. Its causative agent, dengue virus (DENV), has immunologically distinct serotypes that increase the risk of life-threatening diseases such as dengue hemorrhagic fever. However, an effective medication for dengue has not yet been established. There is, therefore, an urgent need to develop new antivirals and vaccines against DENV. Here, we have characterized C19orf66, named Repressor of yield of DENV (RyDEN), as a cellular gene that inhibits the replication of all DENV serotypes. The expression of RyDEN was found to be upregulated by interferon (IFN) treatment and played a critical role in the IFN-mediated anti-DENV response. We also found that RyDEN was likely to block the protein translation of DENV RNA through its association with cellular mRNA-binding proteins and viral RNA. Intriguingly, replication of several other viruses, such as hepatitis C virus, Kunjin virus, and Chikungunya virus, was also limited by RyDEN expression. Thus, this study describes a novel mechanism of an IFN-inducible inhibitory factor for DENV and provides the basis for future development of broad-spectrum antivirals against infectious viral diseases, including dengue.
| Dengue virus (DENV) is a mosquito-borne virus belonging to the genus Flavivirus, which is a large family of enveloped, positive-stranded RNA viruses. DENV has four antigenically distinct serotypes (DENV-1 to -4); all serotypes are able to cause dengue fever (DF) and dengue hemorrhagic fever (DHF) in humans. While primary infection with one of the four DENV serotypes is often asymptomatic or causes self-limiting DF, due to the presence of non- or sub-neutralizing antibodies produced during the primary infection, a secondary infection with a different serotype increases the risk of a more severe form of dengue infection, such as life-threatening DHF and dengue shock syndrome (DSS). However, there is currently no effective vaccine or specific antiviral treatment available for dengue prevention and control [1].
At the cellular level, DENV infection begins with entry via receptor-mediated endocytosis, followed by particle disassembly to release an ~11-kb single-stranded RNA genome into the cytoplasm. The viral genomic RNA contains an open reading frame (ORF) encoding a single polyprotein, which is flanked by a capped 5’ untranslated region (UTR) and a non-polyadenylated 3’UTR, and serves as a template for the translation of a viral precursor protein. The single polypeptide is then cleaved co- and post-translationally into three structural (C, prM, and E) and seven non-structural (NS) proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5). The structural proteins are used for the assembly of virus particles, while the NS proteins are mainly involved in synthesis of the viral RNA genome and the further translation process during DENV infection [2].
Many host factors have been reportedly implicated in the replication of DENV; however, the biological relevance of those factors in in vivo infection and the pathogenesis of DENV has not been fully addressed [3,4]. Meanwhile, it has also become apparent that host cells may harbor factors whose expressions potentially restrict DENV replication. In this regard, the induction of the interferon (IFN) response is considered to be the first line of defense against an invading DENV [5]. DENV infection is able to induce the IFN response, probably through the recognition of viral genomic RNA by intracellular receptors such as TLR-3, RIG-I, and MDA-5 [6–8], which in turn triggers a cellular antiviral state that suppresses the early replication and subsequent spread of DENV. Several in vitro studies have reported that the establishment of a DENV infection is capable of antagonizing IFN signaling cascades by employing viral NS proteins [9–14]. However, pretreatment of human cells with type I (IFN-α and IFN-β) or type II (IFN-γ) IFNs has been shown to limit the replication of DENV [15]. Also, mice deficient in IFN receptors [16] or an IFN signaling component, signal transducer, and activator of transcription 2 (STAT2) [17,18] are reported to be highly susceptible to DENV infection. Given additional evidence that DHF/DSS patients have higher levels of circulating IFN-α and IFN-γ as compared to DF patients [19–21], IFN response is likely to play a key role in controlling DENV replication in vivo [22].
The antiviral effect of IFN is known to be mediated by interferon-stimulated genes (ISGs), which disrupt various steps of virus replication [23]. So far, hundreds of genes have been classified as ISGs; among them, a number of ISGs have been demonstrated to restrict divergent families of viruses, including flaviviruses [23–26]. As for DENV, gene overexpression and knockdown studies have reported that several human ISGs, including interferon-inducible transmembrane proteins (IFITMs), ISG15, ISG20, Viperin, and BST2, have suppressive effects against in vitro virus infection [27–32]. Additionally, a recent large-scale screening study using a library of ISG comprising more than 350 genes revealed that at least 10 ISGs were potent cellular inhibitors of DENV replication that modulate DENV infection in the early or late stage of virus replication [33]. Although the precise mechanisms of action of the anti-DENV ISGs are not yet clear, many of them are likely to function as effector molecules that directly interfere with viral components during infection [23]. Therefore, we believe that understanding how IFN-inducible effector molecules restrict virus infection will be the molecular basis for developing new antiviral agents and vaccines against DENV.
This study aimed to identify new cellular suppressive factors against DENV infection by a gain-of-function screen using a cDNA library derived from type I IFN-treated human cells. We then found that a previously uncharacterized cellular gene, C19orf66, named RyDEN (Repressor of yield of DENV), conferred resistance to all serotypes of DENV in human cells. RyDEN was considered to be an ISG whose expression was essential for the full activity of the type I IFN-mediated suppression of DENV replication. Other than its impact on DENV, overexpression of RyDEN in human cells limited the replication of several RNA and DNA viruses. Interestingly, RyDEN was found to form a complex with cellular mRNA binding proteins, PABPC1 and LARP1, which are required for the efficient replication of DENV. Moreover, RyDEN was likely to interact with DENV RNA and impair the protein translation of viral RNA. Our data demonstrate a novel mechanism of ISG in the inhibition of DENV infection.
It has been demonstrated that pretreatment with type I IFN protects human cells from DENV infection in vitro [15,34]. In order to identify anti-DENV effector molecule(s) in the IFN response, a pool of cDNA was generated from the mRNA of HeLa cells that had been treated with type I IFN (a mixture of human IFNα and ω [Sigma]) and transferred to a lentiviral vector, pYK005C [35], by the Gateway recombination system (Fig 1A). The mean sizes of the IFN-derived cDNA in the Gateway entry (i.e., pDONR22) and destination (pYK005C) vectors were 1.43±0.74 and 1.29±0.63 kbp, respectively (Fig 1B). Infectious lentiviral vectors carrying the cDNA library were produced as a vesicular stomatitis virus G protein (VSV-G) pseudotype and used to transduce a human hepatoma cell line, Huh7.5, which exhibited a massive cytopathic effect with DENV-2 infection (S1 Fig). Transduced cells were then challenged with DENV-2 (Singapore isolate EDEN2 3295 [36]) at a multiplicity of infection (MOI) of 1, and surviving cells were selected (Fig 1A). From the initial screen, a total of 52 surviving cell clones were collected and further verified for their resistance against DENV infection. Plaque assay revealed that, among the 52 cell clones, inhibition of DENV-2 replication was still observed in 32 clones (Fig 2). Reduced replication of DENV in the clones was also confirmed by immunofluorescent analysis (IFA) using anti-double-stranded (ds) RNA antibodies (Fig 1C). PCR amplification and subsequent sequencing analysis using a BLAST search revealed that cDNA from 19 of 32 DENV-resistant clones (59.3%) contained an ORF of a previously uncharacterized gene on chromosome 19, C19orf66, in the integrated pYK005C vector genome (Fig 1D and 1E). Because the inhibitory effect of C19orf66 on DENV replication was confirmed by the following experiments, we named this gene Repressor of yield of DENV (RyDEN).
RyDEN/C19orf66 is an eight-exon gene located on genomic region 19p13.2. This gene spans approximately 7.1 kb in the human genome and encodes a 291 amino acid protein in its ORF (Fig 1E). However, the functional characteristic of the protein product of the RyDEN gene is unknown. A secondary structure prediction by the JPred program (http://www.compbio.dundee.ac.uk/jpred/) represented RyDEN as consisting of eight α-helixes and seven β-strands (Fig 1F). The RyDEN amino acid sequence was also predicted to contain a nuclear localization signal (NLS, 121–137, by cNLS Mapper [http://nls-mapper.iab.keio.ac.jp/cgi-bin/NLS_Mapper_form.cgi]), a nuclear export signal (NES, 261–269, by NetNES [http://www.cbs.dtu.dk/services/NetNES/]), a zinc-ribbon domain (112–135) that is defined by CXXC(H)-15/17-CXXC [37] and a coiled-coil motif (261–286) (Fig 1F). In addition, a characteristic glutamic acid (E)-rich domain was found in the C-terminal region (274–286, Fig 1F).
In order to verify the inhibitory action of RyDEN against DENV, the ORF of RyDEN was cloned back into a lentiviral vector as an N-terminal V5-tagged gene and used to create human cell lines (Huh7.5 and HepG2) that stably expressed V5-RyDEN (Fig 3A). When the cell lines were infected with three doses of DENV-2 (MOIs of 0.1, 1, and 10), virus replication was significantly suppressed, reducing the virus titer by ~43-fold, as compared to the control protein (V5-tagged bacterial dihydrofolate reductase [DHFR])-expressing cells (Fig 3B). Although a more potent inhibitory effect of V5-RyDEN expression was observed in HepG2 cells than in Huh7.5 cells (Fig 3B), it was presumed that this difference occurred due to a higher susceptibility of Huh7.5 cells to DENV infection or a higher expression of V5-RyDEN in HepG2 cells (Fig 3A). DENV inhibition by RyDEN expression was also observed in HEK293T cells (S2 Fig). The ability of RyDEN to inhibit three other serotypes of DENV (Singapore isolates [36]) and another strain of DENV-2 (New Guinea C [NGC]) was also examined. Results showed that the replication of DENV-1, -2, -3, and -4 were inhibited 12.3-, 72.7-, 20.0-, and 92.3-fold, respectively (Fig 3C).
Using an RNA interference experiment, we next investigated whether endogenous expression of RyDEN acts as a suppressor against DENV. To create RyDEN knockdown cells, lentiviral vectors expressing three different small hairpin RNA (shRNA) sequences against RyDEN mRNA (sh1425, sh3151, and sh5890) or a non-targeting control shRNA (shCtrl) were constructed and used to transduce HeLa cells. Quantitative reverse transcription-PCR (qRT-PCR) analysis showed that the expression levels of endogenous RyDEN mRNA in shRNA1425-, sh3151-, and sh5890-expressing cells were 33.3, 67.8, and 99.2%, respectively, as compared with those of shCtrl-expressing cells (Fig 3D). Following infection of the knockdown cell lines with DENV-2 at an MOI of 1 revealed that virus replication was significantly stimulated by RyDEN silencing, which was in accordance with the depletion efficiency of RyDEN mRNA in the three shRNA cell lines (Fig 3E). This enhancement of DENV replication by the knockdown of endogenous RyDEN was also observed in HepG2 and Huh7.5 cells (S3 Fig). In order to test the specificity and reproducibility of the shRNA experiment, we also created sh1425- and shCtrl-expressing cell lines using HepG2 cells. Then, sh1425-susceptible wild-type (WT) or sh1425-resistant mutant (1425R) V5-RyDEN was expressed in the cell lines by the transduction of the lentiviral vector system (Fig 3F). The created cells were challenged with DENV-2. When compared to untransduced (parental) shRNA cells, both V5-RyDEN (i.e., WT and 1425R) suppressed DENV replication at a similar levels in shCtrl cells, whereas, in sh1425 cells, a significant reduction of virus replication was observed only with 1425R RyDEN expression but not with WT RyDEN expression (Fig 3G). Taken together, these data conclude that the expression of RyDEN confers resistance to DENV infection in human cells.
Since RyDEN was first identified by a gain-of-function screen using a type I IFN-treated HeLa cell-derived cDNA library (Fig 1), we examined whether RyDEN was upregulated by IFN treatment in human cells. Immunoblotting analysis using a commercially available anti-RyDEN antibodies (anti-C19orf66 rabbit IgG purchased from Abcam) revealed that RyDEN expression was indeed enhanced in HeLa cells in response to the increasing concentration of IFN-α/ω (Fig 4A). In addition to the type I IFN treatment, the expression of RyDEN was also upregulated by treatment with type II (IFN-γ) and type III (IFN-λ) IFNs (Fig 4B). It is notable that the specificity of the anti-RyDEN antibody on treatment with IFNs was confirmed by knockdown of RyDEN mRNA in sh1425-expressing HepG2 cells (Fig 4B). Quantitative analysis of RyDEN mRNA by qRT-PCR showed that upregulation of RyDEN expression by type I IFN treatment was also observed in all of the human cell lines tested; however, the induction level varied among cells (S4 Fig).
Next, we evaluated how RyDEN expression could contribute to IFN-mediated anti-DENV functions using the RyDEN knockdown cell line. As was observed in a previous report [15], pretreatment with IFN-α/ω suppressed the replication of DENV-2 104-fold in control shRNA (shCtrl)-expressing HepG2 cells (Fig 4C, white bars). However, in HeLa cells in which endogenous RyDEN had been depleted by sh1425, type I IFN treatment inhibited DENV infection by only 29% (Fig 4C, gray bars). Thus, these results indicate that RyDEN is an ISG that plays a critical role in the IFN-mediated anti-DENV response in human cells.
In order to test the possibility that RyDEN may be a key component in the IFN signaling pathway, we compared gene expressions of a variety of ISGs between RyDEN and control protein-expressing cells. HepG2 cells were transfected with V5-tagged RyDEN or control BAP -expressing plasmid DNA, and 48 h after transfection, the level of mRNA expression of the ISGs (LY6E, ISG15, ISG54, and RIG-I) and IFN-β were measured by qRT-PCR analysis. The results showed that no significant activation in the mRNA expression of these genes was observed with the V5-RyDEN transfection, whereas the transfection of a stimulator of the interferon gene (STING), an endoplasmic reticulum-associated adaptor molecule regulating the IFN production [38], upregulated the ISGs and the IFN-β gene (Fig 4D). Additionally, a parallel experiment using RyDEN knockdown (sh1425) and control (shCtrl) HeLa cells revealed that gene expressions of ISG15, ISG54, RIG-I, and IFN-β upon treatment with IFN-α/ω were not reduced by the depletion of endogenous RyDEN (Fig 4E).
A recent study reported a unique regulation of ISG expression, in which some host RNA-binding proteins activated the translational process of ISG mRNA [39]. Hence, we also tested whether RyDEN was involved in the translational regulation of ISGs. An immunoblotting analysis against ISG15, which has been reported to restrict DENV replication [30], showed that a comparable level of ISG15 protein expression following type I IFN treatment was detected in RyDEN knockdown (sh1425-expressing) and control (shCtrl-expressing) HeLa cells (Fig 4F). In summary, these results indicate that RyDEN is not a regulator of the IFN response.
To gain insight into the process of DENV replication that is affected by RyDEN, we first assessed the efficiency of virus entry using previously reported entry assays [34,40]. V5-tagged RyDEN or control protein (Renilla luciferase [RLuc])-expressing Huh7.5 cells were exposed to DENV-2 at an MOI of 5 at 37°C for 2 h, which allowed binding and internalization of virions, and then treated with a high-salt concentration alkaline solution on ice to remove uninternalized viruses, followed by additional washing with PBS. qRT-PCR analysis targeted against the DENV 3’UTR to measure the amount of internalized viruses showed that RyDEN expression did not influence the virus binding/entry process (Fig 5A).
To confirm the observation that the post-entry process is affected by RyDEN, an indirect assay, in which viral genomic RNA was transfected to cells to bypass the binding, entry, and uncoating steps of DENV replication, was performed [34,41]. Naked viral RNA was purified from a culture supernatant that contained infectious DENV-2 and was transfected to V5-RyDEN or control protein-expressing Huh7.5 cells. When infectious titers of DENV produced from transfected cells were analyzed by plaque assay 3 days after transfection, the production and subsequent replication of DENV was significantly inhibited by the expression RyDEN (Fig 5B).
Next, we examined the inhibitory effect of RyDEN on intracellular events in DENV replication using a reporter luciferase-expressing DENV-2 subgenomic RNA replicon system (DENrepPAC2A-Rluc [42]). The transient transfection of V5-RyDEN-expressing plasmid DNA into A549 cells harboring the DENV replicon exhibited a significant and dose-dependent suppression of the luciferase activity as compared with the V5- BAP control-expressing plasmid transfection (Fig 5C). As indicated in previous studies [42,43], treatment with two antiviral components, small interference RNA (siRNA) against DENV-2 NS3 (siNS3, Fig 5C) and mycophenolic acid (MPA) that has been demonstrated to prevent viral RNA replication (S5 Fig), resulted in drastic reductions in the replicon signal. Although the inhibition of DENV replicon activity by a transfection of RyDEN was less effective when compared to the NS3 siRNA or MPA treatment, it was still comparable to the level of inhibition by type I IFN treatment (S5 Fig).
We further monitored the kinetics of DENV RNA accumulation in virus-infected cells. When total DENV RNA (T) was measured by qRT-PCR analysis using a random primer for RT, a slight and insignificant decrease in the amount of viral RNA was detected in RyDEN-expressing cells 6 h after infection (2.3 times lower than in control protein-expressing cells, Fig 5D). Nevertheless, further and significant reductions in the level of total viral RNA were observed 18 and 24 h after infection (Fig 5D). When the level of negative-strand DENV RNA (N) as measured by qRT-PCR analysis using 3’UTR-specfic forward primer for RT was compared, a significant decrease in the amount of negative-strand RNA was also detected 18 and 24 h after infection in RyDEN-expressing cells (Fig 5D). However, the kinetics of accumulating total and negative-strand DENV RNA in the V5-RyDEN-expressing cells appeared to be similar to those in control cells (Fig 5D). Taken together, these data suggest that RyDEN somehow inhibits intracellular events of DENV replication independent of the entry, uncoating, assembly, or negative-strand RNA synthesis of a virus.
In order to search for additional clues regarding the function of RyDEN, we attempted to identify the interacting partners of RyDEN using an affinity purification-mass spectrometry approach. For this purpose, HepG2 cell lines stably expressing RyDEN or control protein (BAP) were fused by lentiviral vector transduction with an N-terminal tandem affinity purification (TAP) tag that contained two IgG binding units [44]. TAP-fused RyDEN and its associated proteins were recovered from the extract of the HepG2 cell lines using IgG Sepharose beads under physiological conditions [44]. SDS-PAGE and subsequent silver staining analysis showed that TAP-RyDEN, but not the TAP-BAP, were specifically co-purified with a >70 kDa band (Fig 6A). Mass spectrometry analysis then identified the >70 kDa band as a poly(A)-binding protein cytoplasmic 1 (PABPC1). Likewise, as we shall see below, mass spectrometry analysis revealed that the additional protein band with a molecular weight mass of around 150 kDa was the La motif-related protein 1 (LARP1). The >40-kDa protein band was confirmed to be TAP-RyDEN (Fig 6A and S6 Fig).
PABPC1 belongs to the evolutionally conserved PABP family of proteins that bind the 3’ poly(A) tail of the mRNA and have multiple roles in translation and mRNA stability [45]. PABPC1 is expressed widely in a variety of human tissues [46] and is reported to have multiple roles in cytoplasmic mRNA function [47,48]. Interestingly, a previous biochemical study by Polacek et al. demonstrated that PABP binds to the DENV 3’UTR RNA in vitro, despite the lack of a poly(A) tail in the viral genome, suggesting the modulatory activity of PABP in DENV mRNA translation [49]. LARP1 is one of the La-motif related proteins that is a superfamily of RNA-binding factor, which is conserved in eukaryotes [50]. This RNA-binding protein was first identified in Drosophila, and is reported to be involved in spermatogenesis, embryogenesis, and cell cycle progression [51–53]. In mammalian cells, it has been demonstrated that LARP1 regulates cell division, apoptosis, and cell migration [54]. It should be noted that LARP1 is found in a complex of certain poly(A)-binding proteins and also interacts with PABPC1 in Drosophila and human cells [53–55].
To examine the role of these cellular mRNA-binding proteins in DENV infection, HepG2 cells were subjected to gene silencing by siRNA against PABPC1 and LARP1, resulting in 10.0- and 2.9-fold reductions in mRNA expression, respectively, as measured by qRT-PCR analysis (Fig 6B and 6C). When siRNA-transfected cells were infected with DENV-2 at an MOI of 1, we found that the level of virus replication was significantly decreased in both knockdown cells as compared with the replication in non-targeting control siRNA (siCtrl, Fig 6D and 6E), implying that PABPC1 and LARP1 positively impact DENV replication. Note that, at least in the PABPC1 siRNA-transfected cells, severe growth defects, which might lead to limited DENV replication, was not likely to be caused by the depletion of PABPC1 (S7 Fig).
We sought to determine the RyDEN domain that is required for interaction with PABPC1. To map the binding domain, a series of V5-tagged RyDEN containing N- and C-terminally truncated mutants was constructed (Fig 7A) and stably expressed in Huh7.5 cells. In this experiment, V5-tagged RLuc was used as a control protein to confirm the specificity of RyDEN-PABPC1 interaction. Co-immunoprecipitation using anti-V5 antibodies followed by immunoblotting using anti-PABPC1 antibodies confirmed that full-length (WT) RyDEN interacted with PABPC1 (Fig 7B, middle panel, lane 2). Importantly, interaction with PABPC1 was also detected with RyDEN-truncated mutants 1–250, 51–291, and 101–291 (lanes 3–5), whereas RyDEN mutant 151–291 was not co-precipitated with PABPC1 (lane 6). Therefore, this result indicates that the domain of interaction with PABPC1 is located in RyDEN’s middle region, which is between amino acid positions 102–150.
As described above, RyDEN was predicted to possesses a sequence resembling a bipartite NLS (121RRVPQRKEVSRCRKCRK137, Fig 1F), which was called an NLS-like (NLS-L) sequence in this study. An IFA using V5-RyDEN-exppressing HepG2 cells and anti-V5 antibodies showed that a higher concentration of ectopically expressed RyDEN was found in cytoplasm (Fig 7C). When intracellular distribution of RyDEN was analyzed by IFA using a newly generated anti-RyDEN rabbit serum, endogenous RyDEN that had been induced by type I IFN localized mainly in the cytoplasm of HepG2 cells (Fig 7D). This cytoplasmic localization of RyDEN was observed in several other human cell lines as well (S8 Fig). By contrast, a parallel IFA using V5-RyDEN truncation mutant (Fig 7A)-expressing cells revealed that the localization of RyDEN to the nucleus was only observed when the C-terminal domain encompassing the putative NES sequence was deleted (V5-RyDEN 1–250, Fig 7E). These data indicate that, in the presence of C-terminal NES, the NLS-L sequence may not function as an active NLS to accumulate RyDEN in the nucleus.
Meanwhile, since the NLS-L sequence is located in the domain of RyDEN’s interaction with PABPC1 (102–150, Fig 7A), we examined whether the NLS-L mutations influenced the binding of RyDEN to PABPC1. To this end, we constructed a site-directed mutant of RyDEN, in which positively charged arginine (R121, R122, R126, R131, R133, and R136) and lysine (K127, K134, and K137) residues in NLS-L were changed to alanine (121AAVPQAAEVSACAACAA137, Fig 7A). Intriguingly, immunoprecipitation analysis using lysates of V5-tagged WT or NLS-L mutant RyDEN-expressing HepG2 cells revealed that the binding efficiency of RyDEN to PABPC1 was decreased by the mutation of NLS-L (Fig 7F). More importantly, when the replication of DENV-2 in each cell line was compared, although some inhibition of virus replication was still observed in the NLS-L mutant-expressing cells, its inhibitory effect was 25.4-fold lower than that obtained in WT RyDEN-expressing cells (Fig 7G). These results suggest that interaction with PABPC1 participates in RyDEN’s anti-DENV activity.
The functional interaction of RyDEN with cellular mRNA-binding proteins PABPC1 and LARP1 (Figs 6 and 7) prompted us to test whether RyDEN was recruited to DENV RNA during infection. To analyze the association of RyDEN with DENV RNA, RNA immunoprecipitation (RIP) assay was performed. HepG2 cells stably expressing V5-tagged protein were infected with DENV-2 at an MOI of 5, and cell lysates were subjected to immunoprecipitation 6 h after infection. In this experiment, we needed to harvest infected cells at an early time point to recover sufficient amounts of DENV RNA because, at a later time point, the amount of viral RNA synthesis had been shown to be dramatically inhibited by the overexpression of V5-RyDEN (Fig 5D). In fact, a significant reduction in the amount of DENV RNA was already detected in the input fraction of V5-RyDEN-expressing cells 6 h after infection (Fig 8A). Note that a significant reduction of viral RNA was not observed in NLS-L mutant RyDEN-expressing cells (a 27% reduction relative to V5-DHFR-expressing cells, Fig 8A). Cell lysates from V5-tagged WT RyDEN-, NLS-L mutant RyDEN-, and control DHFR-expressing cells were used for immunoprecipitation using anti-V5 antibodies, and the total RNA extracted from immunoprecipitates was detected by qRT-PCR analysis against the DENV-2 3’UTR. When the input fraction and the immunoprecipitates were subjected to an immunoblotting analysis, comparable levels of V5-tagged proteins were found to be pulled down by the immunoprecipitation (Fig 8B). However, as anticipated, immunoprecipitation with V5-RyDEN significantly enriched DENV RNA as compared to the level of viral RNA detected in V5-DHFR immunoprecipitates (Fig 8C). In contrast, immunoprecipitation with the NLS-L mutant of RyDEN, which was not able to impair DENV RNA synthesis 6 h after infection (Fig 8A) exhibited only slight enrichment of viral RNA (not statistically significantly different from the V5-DHFR sample, Fig 8C).
To further examine the association of RyDEN with DENV RNA, we performed an in vitro RNA-binding assay based on AlphaScreen technology (PerkinElmer). For this experiment, recombinant proteins (RyDEN and PABPC1) were obtained by the wheat germ cell-free protein production system, a eukaryotic cell-based in vitro translation method that allows the generation of properly folded high-quality proteins [56], because the expression of RyDEN in E. coli was found to be toxic to the bacterial cells. N-terminal FLAG-tagged (RyDEN WT, RyDEN NLS-L mutant, and control DHFR) and glutathione S-transferase (GST)-tagged (PABPC1 and control DHFR) proteins were produced, affinity purified (S9 Fig), and mixed with biotin-labeled DENV-2 3’UTR RNA (450 base), followed by incubation with streptavidin-coated donor beads, anti-FLAG antibodies, and protein A-conjugated acceptor beads. If FLAG-RyDEN interacts with biotinylated 3’UTR RNA, the reaction bridges the donor and acceptor beads by recognizing the biotin of RNA and FLAG-tagged proteins, respectively, which in turn enables the generation of singlet oxygen (O2(1Dg)) from donor beads upon the illumination and the chemical energy transfer to acceptor beads, resulting in a luminescent AlphaScreen signal (Fig 8D) [56]. As shown in Fig 8E, a reaction containing FLAG-RyDEN WT and unlabeled (i.e. non-biotinylated) DENV 3’UTR RNA (Rxn 1) or FLAG-DHFR and biotinylated 3’UTR RNA (Rxn 4) gave a negligible background signal in the AlphaScreen assay. When FLAG-RyDEN WT was incubated with biotinylated 3’UTR RNA (Rxn 5), a significantly increased luminescent signal was detected, while this was also observed in the incubation with biotinylated nonspecific control RNA (Rxn 2), indicating the RNA-binding property of RyDEN. However, the binding signal between WT RyDEN and biotinylated 3'UTR was significantly enhanced by the presence of GST-PABPC1 (Rxn 6). The specific interaction between RyDEN and DENV 3'UTR in this reaction was shown by a competition assay using unlabeled 3'UTR RNA as a competitor (S10 Fig). In contrast, the addition of GST-PABPC1 did not change the luminescent signal of FLAG-RyDEN WT and the biotinylated control RNA incubation (Rxn 3). More importantly, even in the presence of GST-PABPC1, the RyDEN NLS-L mutant (Rxn 7), which was shown to have reduced binding activity to PABPC1 (Fig 7F), did not generate the higher interaction signals with biotinylated 3’UTR RNA that were observed in the incubation with RyDEN WT (Rxn 6). These data, therefore, demonstrate that RyDEN is an RNA-binding protein, and binding specificity to DENV RNA is provided through a complex formation with PABPC1.
The interaction of RyDEN with PABPC1, an important molecule involved in cellular mRNA translation, led us to hypothesize that RyDEN might interfere with the translation process of DENV RNA. First, in order to investigate the effect of RyDEN expression on global cellular translation, puromycin labeling of newly synthesized proteins was performed using RyDEN-expressing cells [57]. HepG2 cells expressing V5-RyDEN or V5-DHFR were pulsed with puromycin, and cell lysates after the 40 min pulse were subjected to immunoblotting using anti-puromycin antibodies to compare the total protein synthesis of these two cell lines. As evident in control treatments in which cells had been treated with a protein synthesis inhibitor, cycloheximide, before puromycin pulse (CHX, Fig 9A), proteins detected by immunoblotting indicated de novo synthesized proteins that incorporated puromycin during mRNA translation in cells [57]. When the level of puromycin-labeled proteins was compared, there was no obvious difference in the protein synthesis of V5-RyDEN and V5-DHFR-expressing cells (Fig 9A), indicating that the global translation rate was not reduced by the expression of RyDEN.
We next examined the ability of RyDEN to interfere with protein synthesis from DENV RNA by employing a DENV-2-based luciferase reporter construct, DENrepPAC2A-Rluc [42]. DENV reporter RNA was transcribed in vitro transcribed using linearized construct DNA in the presence of an m7GpppA cap analogue and transfected to V5-RyDEN- or V5-DHFR-expressing HepG2 cells. As shown in Fig 9B, the RNA transfection of WT DENV reporter replicon (DENrepPAC2A-Rluc WT) exhibited reduced luciferase activity in V5-RyDEN-expressing cells when compared to V5-DHFR-expressing control cells 4 and 8 h after transfection. Importantly, diminished luciferase activity in the RyDEN-expressing cells at the early time points were also observed by RNA transfection of a mutant DENV reporter construct, DENrepPAC2A-Rluc GVD, in which the GDD motif in the active site of the RNA-dependent RNA polymerase (RdRp) had been changed to GVD (Fig 9B) [58]. Since the GVD mutation in the NS5 RdRp is reported to abolish viral RNA replication [58], the luciferase activity was considered to reflect the level of protein production from mRNA of the transfected construct. Although the inhibitory effect of RyDEN on the reporter protein production was relatively modest as compared to the inhibition levels observed in DENV replication (Fig 1) and viral RNA accumulation (Fig 5D), these data suggest that the expression of RyDEN is likely to be suppressive to the translation process of DENV RNA.
Since RyDEN was found to be involved in establishing an antiviral state against DENV in human cells, we also investigated whether the expression of RyDEN influences the replication of other viruses. To this end, V5-RyDEN-expressing cells were further created using human cell lines including HeLa, Jurkat, and A549 cells by lentiviral vector-mediated transduction. Cell lines were then infected with several RNA (hepatitis C virus [HCV, Flaviviridae], West Nile virus Kunjin strain [WNVKUN, Flaviviridae], Chikungunya virus [CHIKV, Togaviridae], poliovirus [Picornaviridae], human enterovirus 71 [EV71, Picornaviridae], and human immunodeficiency virus type-1 [HIV-1, Retroviridae]) and DNA (herpes simplex virus 1 [HSV-1, Herpesviridae], HSV-2, and human adenovirus type 3 [HAdV-3, Adenoviridae]) viruses. Measurements of the virus titer in the supernatants of infected cells indicated that significant inhibition by the overexpression of RyDEN was observed in HCV, WNVKUN, and CHIKV, but not in poliovirus, EV71, or HIV-1 infections, as compared to that in control protein-expressing cells (Fig 10A). A preliminary result showed that replication of the Sindbis virus, a Togaviridae family virus, was also suppressed in V5-RyDEN-expressing HeLa cells (S11 Fig), suggesting that Flaviviridae and Togaviridae family members are broadly susceptible to RyDEN. Intriguingly, the replication of some DNA viruses, including HSV-1 and HAdV-3, were negatively affected by V5-RyDEN expression, whereas it had no influence on HSV-2 infection (Fig 10B).
In terms of morbidity and mortality, dengue has emerged as one of the most important arthropod-borne diseases in the world, with cases predominantly documented in tropical and sub-tropical urban centers. Currently, the development of new antiviral medications and vaccinations against DENV is an urgently needed. In this regard, understanding the host innate immune response that restricts DENV replication, such as the IFN response, will be important for the development of antiviral agents and effective vaccines. In this study, we present RyDEN (C19orf66) as an ISG that limits all serotypes of DENV. Our findings suggest that RyDEN may target the translation of DENV RNA through interaction with other cellular RNA-binding proteins.
Expression cloning of the cDNA library is a powerful approach to the functional and comprehensive analysis of cellular genes; such a gain-of-function screen has been applied to identify host factors involved in DENV replication [33,59]. In this study, a library of cDNA was generated from mRNA of type I IFN-treated HeLa cells and lentivirally expressed in Huh7.5 cells that exhibited massive cell death with DENV infection (S1 Fig), which was expected to confer extensive resistance to DENV-induced cell death (Fig 1A). Indeed, one round of a DENV-2 challenge resulted in more than 50 surviving cell clones on a 150-mm dish. An additional infection assay showed that 32 clones remained more or less resistant to DENV infection (Fig 2). Sequencing analysis of cDNA recovered from DENV-resistant Huh7.5 cells revealed that 19 cells harbored the RyDEN gene. Although some of the cells also contained all or parts of other genes or non-ORF sequences, the full ORF of RyDEN was isolated from all cells (Fig 2), indicating that RyDEN should be a major determinant of resistance to DENV in a cDNA library screening assay. Intriguingly, almost the same mutant (amino acid position 304–702) of DNAJC14, an Hsp40 family member that has been identified as an anti-flavivirus factor by a cDNA library screen [60], was also recovered in this study (Fig 2, clone 31), demonstrating the integrity of our screening. The previous report by Yi et al. showed that despite screening using cDNA from IFN-α-treated cells, DNAJC14 mRNA levels were not upregulated by interferon treatment, although the DNAJC14 mutant was again identified with a cDNA library of IFN-treated HeLa cells in our study. Thus, it still would be interesting to investigate how the DNAJC14 function is associated with the IFN-mediated antiviral response. In addition, the future investigation of other genes identified in our IFN cDNA library screen (e.g., IFN-α-inducible protein 27, C19orf53) in flavivirus replication including DENV will provide fascinating insights into the interaction between virus and host.
RyDEN is expressed from chromosome 19 as an eight-exon gene that encodes a 291 amino acid protein (Fig 1E). A BLAST search analysis using RyDEN’s amino acid sequence did not show any overt similarities with other proteins in mammals; however, this protein was predicted to contain a zinc-ribbon domain in the central region and a coiled-coil motif in the C-terminal region (Fig 1F). The zinc-ribbon motif, which is basically defined by CXXC(H)-15/17-CXXC, is a general architectural motif initially found in some eukaryotic transcription factors and RNA polymerase subunits that currently form largest group of zinc fingers [37]. Although the zinc ribbons seem to display limited sequence similarities, structural analysis revealed that a variety of cellular and viral proteins possess this motif as a binding domain for zinc [37]. Of interest is the fact that cyclic GMP-AMP synthase (cGAS), a cytosolic DNA-recognition receptor for the induction of IFN responses, has recently been shown to contain zinc ribbon, which is likely to be required for DNA recognition [61]. Since zinc ribbon is found in many DNA- and RNA-binding proteins [37], RyDEN may harbor nucleic acid-binding activity, as discussed below. Also, in amino acid sequence-based protein motif prediction programs, putative NLS (referred to as NLS-L) and NES sequences were found in the zinc-ribbon (121–137) and C-terminal domains (261–269), respectively (Fig 1F). Since IFA experiments showed that RyDEN was mainly dispersed throughout the cytoplasm (Fig 7C and S8A Fig), at least in a normally dividing cell, the NLS-L sequence does not function to accumulate RyDEN in the nucleus. However, deletion of the C-terminal domain containing the putative NES sequence led to an exclusively nuclear location (Fig 7E), suggesting that RyDEN is a potential nucleocytoplasmic shuttling protein, which is mostly retained in the cytoplasm. Note that no obvious changes in the subcellular localization of overexpressed RyDEN were observed with IFN treatment or DENV infection (S8B Fig).
In this study, RyDEN was shown to be an antiviral ISG. The overexpression of RyDEN in human cells suppressed all serotypes of DENV (Fig 3C) and, importantly, the endogenous expression of RyDEN was upregulated with the treatment of types I, II, and III IFNs (Fig 4B). Although the level of artificially expressed RyDEN (i.e. V5-RyDEN) was 38.7±2.1 times higher than that of IFN-induced endogenous RyDEN in HepG2 cells as measured by qRT-PCR analysis, we believe that the expression level of IFN-induced RyDEN sufficiently participates in the inhibition of DENV replication in human cells. Supporting this, in the RyDEN knockdown cell line, the inhibitory effect of type I IFN against DENV-2 was reduced by more than 70% (Fig 4C), indicating a major contribution of RyDEN to the IFN-mediated anti-DENV response. It should also be noted that even without IFN treatment, knockdown of the endogenous expression of RyDEN significantly enhanced DENV replication in several cell lines (Fig 3E and S3 Fig), indicating that a steady-state level of RyDEN acts as a DENV inhibitor. In addition, expression levels of RyDEN as measured by qRT-PCR varied among different human cell lines (S4 Fig), RyDEN expression may be one intracellular factor that determines the cellular tropism of DENV.
One question to ponder is, how does RyDEN suppress the replication of DENV? When the efficiency of virus entry was assessed by qRT-PCR, the level of viral RNA internalized in RyDEN-expressing cells was comparable to that in the control cells (Fig 5A). In contrast, a significant decrease in the level of intracellular DENV RNA was observed in RyDEN-expressing cells 18–24 h after infection (Fig 5D). RyDEN was, therefore, suggested to inhibit the post entry process during DENV replication. Consistent with this, the use of a cell line that harbored the RLuc reporter gene-carrying DENV subgenomic RNA replicon showed that the suppression of luciferase activity occurred with the transient expression of RyDEN (Fig 5C) at a level similar to IFN treatment (S5 Fig). Importantly, transfection of a replication-defective mutant of the DENV reporter construct RNA (DENrepPAC2A-Rluc GVD [58]) showed that luciferase activity of the reporter construct RNA was diminished by the expression of RyDEN (Fig 9B). Since RyDEN was not a mediator of the IFN response (Fig 4), these results suggest that RyDEN is a downstream effector molecule in the anti-DENV IFN response, which may target the translation process of viral RNA. Nevertheless, when compared to more pronounced effect on DENV titers (Fig 1) and viral RNA levels (Fig 5D), the inhibitory effect of RyDEN on the protein translation was modest (Fig 9B). Therefore, we cannot rule out the possibility that RyDEN may also interfere with other step(s) of DENV replication such as RNA transcription or protein processing.
Affinity purification-mass spectrometry analysis using TAP-tagged RyDEN then provided an important clue about RyDEN’s mechanism-of-action: RyDEN was likely to form a complex with the cellular RNA-binding protein PABPC1 (Fig 6A). PABPC1 is one of the major PABP-family proteins in eukaryotes and is ubiquitously expressed in cytoplasm [45]. Although PABPC1 is reported to play multiple roles in the translation, deadenylation, and stability of mRNA through binding to a 3’ poly(A) tail, the typical function of this protein is to form the closed-loop structure of mRNA by interaction with eIF4G, a subunit of the 5’ cap-binding eIF4E complex, to initiate protein translation [47,48]. Of particular interest, a previous study by Polacek et al. has shown that PABP is able to bind the 3’UTR of DENV in vitro [49]. Although the DENV RNA genome lacks a terminal Poly(A) tail, Polacek et al. reported that A-rich stretches upstream of the stem-loop in the 3’UTR appeared to be involved in PABP binding [49]. In our study, the interaction domain of RyDEN with PABPC1 was mapped to the central region between amino acid positions 102–150 (Fig 7A and 7B). Importantly, alanine substitution of positively charged arginine and lysine residues in the NLS-L sequence (121–137) of RyDEN resulted in decreased efficiency in the interaction with PABPC1 and reduced inhibitory activity against DENV replication (Fig 7G).
Additionally, the affinity purification-mass spectrometry analysis identified LARP1 as another interactor with RyDEN (Fig 6A). LARP1 is also an RNA-binding protein that contains two RNA-binding motifs called the La motif and the RNA recognition motif [50]. While it has been documented that the La motif-related protein family is involved in a broad range of activities in cellular RNA, including tRNA processing and mRNA metabolism, LARPs are also reported to affect the translation process of mRNA [50]. In fact, it has been shown that LARP1 associates with PABPC1 and eIF4E in human cells and has a positive role at an early stage of translation initiation [54]. In our study, PABPC1 and LARP1 were found to be positive regulators of DENV, since the siRNA-mediated knockdown of these genes significantly reduced the level of virus replication in HepG2 cells (Fig 6). Given the fact that both PABPC1 and LARP1 have RNA-binding activity [45,50], one could envisage that RyDEN may associate with DENV RNA through its interaction with these proteins during infection. As expected, our data of RIP assay showed that DENV RNA was significantly enriched by V5-tagged RyDEN (Fig 8C). Moreover, AlphaScreen technology-based in vitro RNA-binding assay revealed that RyDEN possessed binding activity to DENV 3’UTR RNA, and the association of RyDEN with 3’UTR RNA was enhanced by the presence of PABPC1 (Fig 8E).
Therefore, based on our findings and the reported functions of PABPC1/LARP1, the following possibility could be proposed regarding the mechanism of RyDEN-mediated antiviral activity in DENV-infected cells: RyDEN forms a complex with PABPC1 (and LARP1) on DENV RNA, and then somehow interferes with the translation machinery of circularized viral RNA (Fig 11). This scenario would be consistent with the previous report by Diamond and Harris, in which IFN treatment was shown to inhibit the translation of DENV RNA rather than by preventing the association of DENV RNA with ribosomes [34]. In light of the data obtained by in vitro RNA-binding assay (Fig 8E), one could envisage that RyDEN’s RNA-binding activity is RNA sequence-nonspecific, but that it gains specificity to positive-strand DENV RNA via interaction with PABPC1 that has been suggested to recognize A-rich stretches in the 3’UTR [49]. Intriguingly, Paip2, a suppressor of PABPC1, has been reported to be such a cellular inhibitor in the viral translation machineries [49,62]. The above-mentioned study of Polacek et al. also presented fascinating evidence that Paip2 is able to block the interaction of PABPC1 with DENV 3’UTR RNA in vitro [49]. Furthermore, a recent work has revealed that Paip2, whose expression is stimulated by human cytomegalovirus (HCMV) infection, limits HCMV protein synthesis and replication [62]. It is noteworthy that a characteristic glutamic acid (E)-rich domain that has been characterized as a binding domain of Paip2 to PABPC1 [63] was also found in the C-terminal region of RyDEN (Fig 1F). Although the C-terminal region surrounding the E-rich domain of RyDEN appeared not to be critical to its interaction with PABPC1 (Fig 7B), RyDEN and Paip2 may have evolutionarily gained a similar regulatory function controlling PABPC1 activity. One concern would be that the translational suppression by RyDEN through interaction with PABPC1 might lead to a translation arrest of the host cell, which would result in the suppression of DENV replication. However, global cellular protein synthesis was not inhibited by the overexpression of RyDEN (Fig 9A). It is therefore conceivable that there is specificity to RyDEN’s recognition of the viral RNA translation complex. In addition, RyDEN may stimulate the degradation of DENV RNA in cytoplasmic P-bodies or stress granules (SGs) in collaboration with PABPC1 and/or LARP1, since the role of PABPC1 and LARP1 in eukaryotic mRNA decay as P-body and SG components has also been demonstrated [64,65]. These should be interesting topics to address in the future.
Our study has also shown that multiple viruses are susceptible to the inhibitory action of RyDEN to a greater or lesser extent, including HCV, WNVKUN, and CHIKV, whereas the replication of other RNA viruses tested (poliovirus, EV71, and HIV-1) was not suppressed by RyDEN overexpression (Fig 10A). Interestingly, some DNA virus replications (HSV-1 and HAdV-3) were also affected by RyDEN (Fig 10B). Our preliminary data showed that the replication of the Sindbis virus was impaired in V5-RyDEN-expressing cells (S11 Fig), suggesting that RyDEN acts as a broad-ranging inhibitory factor, at least against the Flaviviridae and Togaviridae families. Given the proposed model of RyDEN’s inhibitory mode of action against DENV (Fig 11), viruses whose replication is influenced by RyDEN may utilize PABPC1/LARP1 in their replication, particularly in the viral protein translation process. It should be emphasized that PABPs are well-known targets of several viruses, and it has been demonstrated that enteroviruses and lentiviruses cleave PABP by their protease to shut off cellular translation; in contrast, an HSV-1 protein binds PABP to stimulate viral mRNA translation [66]. Therefore, we hypothesize that the antiviral activity of RyDEN depends on whether the virus requires PABPC1 (and LARP1) function in its replication cycle. Indeed, PABPC1 is shown to promote HCV infection [67], which was inhibited by RyDEN (Fig 10A). In agreement with our data, a recent comprehensive study by Schoggins et al. using an overexpression screening of an ISG library has also reported the anti-HCV activity of RyDEN (shown as FLJ11286 gene [24]). Thus, further understanding of the molecular detail of RyDEN will contribute to the development of broadly active antiviral inhibitors.
HEK293T (human embryonic kidney, American Type Culture Collection [ATCC] CRL-11268), Huh7.5 (human hepatocellular carcinoma [68], obtained from Apath, LLC), HepG2 (human hepatoma, ATCC HB-8065), and HeLa (human cervical carcinoma, ATCC CCL-2) cells were cultured in DMEM supplemented with 10% fetal calf serum (FCS, Life Technologies) and antibiotics (100 units/ml penicillin and 100 μg/ml streptomycin). A549 (human lung adenocarcinoma, ATCC CCL-185) and Vero (green monkey kidney, ATCC CCL-81) cells were maintained in F-12K and Eagle's Minimum Essential Medium, respectively, which were supplemented with 10% FCS and antibiotics. BHK-21 (baby hamster kidney, ATCC CCL-10) and Jurkat (human lymphoblastoid T, ATCC TIB-152) cells were grown in RPMI 1640 supplemented with 10% FCS and antibiotics. C6/36 (Aedes albopictus mosquito, ATCC CRL-1660) cells were maintained at 28°C in HEPES-modified RPMI 1640 containing 10% FCS and antibiotics.
The four serotypes of DENV, which was isolated from isolated from patients recruited into the EDEN (early dengue infection and outcome) study in Singapore (DENV-1: Singapore isolate S144; DENV-2: Singapore isolate EDEN2 3295; DENV-3: Singapore isolate EDEN 130/05; and DENV-4: Singapore isolate S8976 [36,41]), DENV-2 (New Guinea C strain), CHIKV (Ross strain), and WNVKUN were propagated in the C6/36 mosquito cells, and viral infectivity was titrated by plaque assays using BHK-21 cells as described previously [30]. HCV J6/JFH1-P47 (genotype 2) was produced using Huh7.5 cells and the virus titer was determined as focus forming units (FFU)/ml by previously reported IFA [69] on Huh7.5 cells using mouse anti-HCV core monoclonal antibodies (MA1-080, Pierce). Poliovirus (Sabin strain) and human enterovirus 71 (Singapore isolate) were propagated in RD cells, and viral infectivity was titrated by plaque assays using RD cells. HIV-1 (NL4-3) was produced by a transfection of HEK293T cells with pNL4-3, and the virus titer of the culture supernatants collected was determined as previously described [35]. Production and titration of HSV-1/2 and HAdV-3 were carried out using Vero and A549 cells, respectively. Virus titer was calculated as plaque-forming units (PFU)/ml (except for HCV and HIV-1).
A Gateway-compatible cDNA library was generated from mRNA isolated from HeLa cells that had been treated with 1,000 U/ml type I IFN (a mixture of human interferon α and ω, Sigma) for 24 h. Briefly, total RNA was extracted using the RNeasy Mini Kit (Qiagen), and mRNA was then isolated using a PolyATtract mRNA Isolation System II (Promega) according to the manufacturer’s recommendations. The cDNA was synthesized using the CloneMiner cDNA Library Construction Kit (Life Technologies) from 3 μg of mRNA and fractionated with cDNA Size Fractionation Columns (Life Technologies). After BP recombination reaction (Life Technologies) using 100 ng of cDNA and 300 ng of an entry vector, pDONR221, the entry library containing approximately 2.5 x 107 clones, was amplified as a pool of transformants in One Shot TOP10 Electrocomp E. coli cells (Life Technologies). The entry vector plasmid DNA was purified using the QIAGEN Plasmid Midi Kit (Qiagen). To generate the lentiviral vector cDNA library, LR recombination reaction (Life Technologies) was performed using 300 ng of the entry cDNA library and 300 ng of an EcoRI-digested destination vector, pYK005C [35]. The resultant vector library was amplified as a pool of recombinants in One Shot TOP10 Electrocomp E. coli cells and purified using the QIAGEN Plasmid Maxi Kit (Qiagen).
A VSV-G-pseudotyped lentiviral vector expressing the IFN cDNA library was produced by the calcium phosphate-mediated transfection method using HEK293T cells as described previously [35]. Concentrated lentiviral vectors were titrated with HEK293T cells by evaluating the percentage of humanized Renilla green fluorescence protein positive cells 48 h after infection using a CyAn ADP flow cytometer (Beckman Coulter).
In a 150-mm dish, 1 x 107 of Huh7.5 cells were seeded 1 day before transduction and infected with 5 x 106 infectious dose of the IFN cDNA carrying lentiviral vectors for 24 h. After 48 h post-transduction, the cells were challenged with DENV-2 (EDEN2 3295) at an MOI of 1. The culture medium was changed every 2–3 days, and after 2 weeks, cell colonies that survived the DENV challenge were transferred to 48-well plates and expanded for further analysis.
Genomic DNA was isolated from the resistant clones using the Wizard Genomic DNA Purification Kit (Promega) from cells that displayed low infectivity of DENV in immunofluorescence and plaque assay. The cDNA was then amplified by PCR using KOD-Plus 2 DNA polymerase (Toyobo) and primers (5’-CTT CCA TTT CAG GTG TCG TGA ACA CGC TAC CGG TCT CGA G-3’ and 5’-CAA ACG CAC ACC GGC CTT ATT CCA AGC GGC TTC GGC CAG-3’) flanking the Gateway cassette in the pYK005c lentiviral vector. cDNA was further amplified by nested PCR using primers (5’-ACC GGT CTC GAG AAT TAT CAA CAA-3’ and 5’-GCT GCA GAA TTA TCA ACC ACT TTG-3’) and cloned into the pCR-Blunt II-TOPO vector (Life Technologies). The sequence of cDNA in the pCR-Blunt II-TOPO vector was analyzed by an automated DNA sequencer, and the data was compared with the DNA database at the National Center for Biotechnology Information using a BLAST search.
To stain for DENV dsRNA in surviving clones, 3 x 104 of cells preseeded in Lab-Tek II 8-well chamber slides (Thermo Scientific) were infected with DENV-2 at an MOI of 5. Two days after infection, cells were fixed with 4% PFA for 30 min, permeabilized with 0.1% Triton X-100 in PBS for 10 min, and blocked with 5% goat serum and 0.5% BSA in PBS for 30 min at room temperature. The cells were stained with anti-dsRNA mouse monoclonal antibody (J2, English & Scientific Consulting Bt.), followed by a secondary antibody, Alexa Fluor 488-conjugated goat anti-rabbit IgG (Life Technologies). A slide was mounted with a ProLong Gold antifade reagent containing DAPI (Life Technologies) and observed under an Olympus IX81 fluorescence microscope. Images were captured with the CellSens Dimension software (Olympus). Staining of V5-tagged proteins was performed using a primary antibody, anti-V5 mouse monoclonal (Life Technologies), followed by Alexa Fluor 488-conjugated anti-mouse secondary antibody (Life Technologies).
To detect endogenous RyDEN, a rabbit serum was generated by Sigma using synthesized 4 peptides derived from RyDEN (amino acid positions 1–19, 51–69, 186–205, and 223–242). Cells preseeded in 8-well chamber slides (3 x 104 of cells per well) were incubated with 1,000 U/ml type I IFN for 24 h, fixed, permeabilized with 1% Triton X-100, and blocked with Blocker Casein (Thermo Scientific). Immunostaining was carried out by an incubation with anti-RyDEN rabbit serum (1:5,000 in blocking buffer) and subsequent incubation with FITC-conjugated donkey anti-rabbit IgG (Rockland).
To create stable cell lines expressing V5-tagged proteins, the ORF of RyDEN and the control proteins (DHFR and RLuc) were amplified by PCR and cloned into pDONR221 through a Gateway BP reaction. The individual ORF was then transferred to a Gateway-compatible lentiviral vector, pYK-nV5-Bla, in which a V5 epitope tag sequence had been added to the upstream of the Gateway unit in pYK005C-Bla [70] by LR reaction. A VSV-G-pseudotyped lentiviral vector was produced as described above and used to transduce human cells, including Huh7.5, HepG2, HeLa, Jurkat, and A549 cells. Transduced cells were selected in the presence of 10 μg/ml of blasticidin (InvivoGen). Expressions of V5-tagged proteins in the stable cell lines were confirmed by immunoblotting using anti-V5 antibodies as described below.
To construct a lentiviral vector that expressed shRNA, synthesized oligonucleotides that contained shRNA sequences against RyDEN ORF (sh1425: 5’-GAA CTA AGT AAC GAT CTG GAT-3’; sh3151: 5’-GAG AAG TTT CAT GGG AAG GTA-3’; sh5890: 5’- GAA GCC AAC CTA CGC ATG TTT-3’) were designed by using the RNAi Consortium web portal (http://www.broadinstitute.org/rnai/public/) and inserted into AgeI-EcoRI sites of a lentiviral vector pLKO.1 puro (Addgene). VSV-G-pseudotyped lentiviral vector particles were produced by the transfection of lentiviral vector DNA encoding sh1425, sh3151, sh5890, or non-targeting control shRNA (SHC002, Sigma) and used to transduce HeLa cells. Transduced cells were selected over 2 weeks with 2 μg/ml of puromycin (InvivoGen). The knockdown efficiency of RyDEN mRNA in each cell line was analyzed by qRT-PCR as described below. The shRNA-resistant RyDEN expression vector was constructed using pYK005C-Bla by replacing the sh1425-targeting sequence of 5’-GAG CTG AGC AAT GAC CTC GAC-3’, which introduced seven silent mutations without changing the amino acid sequences of RyDEN.
Total RNA was isolated from cells using the RNeasy Mini Kit (Qiagen) and was treated with DNase using the TURBO DNA-free Kit (Ambion). cDNA was synthesized using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems), and subjected to real-time qPCR using SsoAdvanced SYBR Green Supermix and CFX96 Real-Time PCR detection system (Bio-Rad). The expression levels of target RNA were calculated by the comparative cycle threshold (CT) method and normalized with GAPDH mRNA levels. In some experiments for the detection of DENV-2 RNA, qRT-PCR was performed by High-Capacity cDNA Reverse Transcription Kit and SsoFast Probes Supermix (Bio-Rad) using previously described primers and fluorescent probe targeting 3’UTR of the DENV genome [71]. For qRT-PCR analysis of DENV-2 minus-strand RNA, cDNA synthesis was carried out using forward primer of 3'UTR instead of random primer as described in previous report [72]. Primer sequences for qRT-PCR analysis are listed in S1 Table.
Protein samples were denatured in an SDS sample buffer, separated by 10% SDS-PAGE gel, and transferred to an Immobilon-P transfer membrane (Millipore). The primary antibodies used were anti-V5 mouse monoclonal (Life Technologies), anti-C19orf66 rabbit polyclonal (Abcam), anti-PABPC1 mouse monoclonal (10E10, Santa Cruz Biotechnology), anti-ISG15 rabbit polyclonal (2743, Cell Signaling), and anti-actin mouse monoclonal (AC40, Sigma) antibodies. Horseradish peroxidase (HRP)-conjugated anti-mouse or anti-rabbit IgG antibody (Cell Signaling) was used as a secondary antibody. For immunoprecipitation analysis, TrueBlot ULTRA anti-mouse IgG HRP (Rockland) was used as a secondary antibody. Proteins were detected using an ImageQuant LAS 4000 mini chemiluminescent image analyzer (GE Healthcare).
To analyze RyDEN expression, HeLa cells preseeded in a 12-well plate at 1 x 105/well density 1 day before treatment were incubated with 10, 100, or 1,000 units/ml of IFN-α/ω at 37°C. Twenty-four hours after treatment, cells were collected and subjected to immunoblotting using an anti-RyDEN antibody. In a parallel experiment, HepG2 cells expressing sh1425 and shCtrl were treated with 300 units/ml of IFN-α/ω, IFN-γ (BioLegend), or IFN-λ1 (PeproTech) for 24 h before assessing RyDEN expression by immunoblotting. For DENV infection, shRNA-expressing HepG2 cells were treated with or without IFN-α/ω (300 units/ml) for 24 h and then inoculated with DENV-2 at an MOI of 1. The culture supernatant was collected 48 h after infection and subjected to plaque assay.
For DENV infection, cells (V5-tagged protein or shRNA-expressing) preseeded in 6-well plate at 5 x 105/well density 1 day prior to infection were infected at an MOI of 0.1, 1, or 10. After 1 h of incubation at 37°C, cells were washed once followed by replacement with growth medium without selection antibiotics. The culture supernatant was collected at indicated time points and subjected to a standard plaque assay. In a similar way, HCV, poliovirus, and EV71 infections were performed by exposing the viruses to V5-RyDEN or V5-DHFR-expressing Huh7.5 cells at an MOI of 2 (HCV) or 1 (poliovirus and EV71), and the culture supernatant was collected 4 days (HCV) or 1 day (poliovirus and EV71) after infection. For WNVKUN, CHIKV, HSV-1, and HSV-2 infections, V5-RyDEN or V5-DHFR-expressing HeLa cells were infected at an MOI of 1 (for WNVKUN and CHIKV) or an MOI of 0.1 (for HSV-1 and HSV-2), and the culture supernatant was collected 48 h (for KUNV) or 72 h (for CHIKV, HSV-1, and HSV-2) after infection. HIV-1 infection of V5-RyDEN or V5-DHFR-expressing Jurkat cells were carried out by exposing the virus (MOI of 0.005) for 2 h, and the level of virus replication was measured with a p24Capsid concentration in a culture supernatant of infected cells [35]. For HAdV-3 infection, V5-RyDEN or V5-RLuc-expressing A549 cells were infected with a virus at an MOI of 1, and the culture supernatant was collected at 24 h.
Virus entry assay was performed as reported by Le Sommer et al. [40]. Huh7.5 cells stably expressing V5-RyDEN or V5-RLuc, which had been seeded in a 24-well plate at a density of 5 x 104/well 1 day before infection, were incubated with DENV-2 at an MOI of 5 at 37°C for 2 h. Uninternalized virus particles were removed by washing the cells twice with cold PBS, followed by a 3-min exposure to 1 M NaCl and 50 mM Na2CO3, pH 9.5. After washing with cold PBS three more times, total RNA was extracted and cell-associated DENV RNA was analyzed by qRT-PCR analysis.
DENV-2 RNA was first extracted from the virus supernatant using QIAamp Viral RNA Mini Kit (Qiagen). For transfecting the isolated viral RNA, Huh7.5 cells that expressed V5-RyDEN or V5-RLuc were preseeded in a 6-well plate at a density of 5 × 105/well 1 day before transfection and transfected with DENV-2 RNA equivalent to 6.7 × 107 PFU using Lipofectamine 2000 (Life Technologies). After 3 days, the culture supernatant was collected to measure the infectious titer of extracellular virus via plaque assay.
A stable A549 cell line expressing a self-replicating DENV replicon was generated by the transfection of in vitro transcribed and 5'-capped genomic RNA of the DENV-2 NGC strain, in which structural genes had been replaced with puromycin-resistant gene and Renilla luciferase gene (DENrepPAC2A-Rluc), and subsequent selection with 5 μg/ml of puromycin as described previously [42]. Established cells were seeded in a 24-well plate at a density of 2 x 104 cells/well and, on the next day, transfected with 4–400 ng of V5-RyDEN or V5-BAP-expressing pcDNA3.1 by Lipofectamine 2000. Forty-eight hours after transfection, cells were harvested and subjected to a luciferase assay using the Renilla Luciferase Glow Assay Kit (Thermo Scientific), as described previously [70]. As an inhibition control experiment, RLuc replicon-expressing A549 cells were also transfected with 10 nM siRNA duplex against DENV NS3 [42] or a scrambled siRNA duplex using siLentFect (Bio-Rad) and analyzed by luciferase assay.
To construct a mutant DENV reporter construct, DENrepPAC2A-Rluc GVD, aspartic acid (D) at position 663 of NS5 was changed to valine (V) [58,73] by QuikChange II XL Site-Directed Mutagenesis Kit (Agilent) using DENrepPAC2A-Rluc as a template plasmid DNA. RNA of DENrepPAC2A-Rluc WT and GVD were in vitro transcribed from XbaI-digested plasmid DNA using MEGAscript T7 Transcription Kit (Life Technologies) in the presence of m7GpppA cap analogue (NEB) and purified by RNeasy Mini Kit (Qiagen). For reporter assay, V5-RyDEN and V5-DHFR-expressing HepG2 cells, which had been preseeded in 24-well plates at 1 x 105 cells/well density, were transfected with 500 ng of transcribed RNA using Lipofectamine 2000, and 4 and 8 h after transfection, the cells were subjected to luciferase activity assay.
V5-RyDEN or V5-BAP-expressing plasmid DNA was constructed using pcDNA3.1/nV5-DEST (Life Technologies) by a Gateway BP reaction. To construct the expression plasmid of STING, ORF of STING, which was fused with the N-terminal HA tag sequence, was generated by RT-PCR using mRNA from HeLa cells and cloned into the EcoRV site of pcDNA3.1 (Life Technologies). Constructed plasmid DNA (500 ng) was transfected to HepG2 cells (preseeded in a 24-well plate at 5 x 104 cells/well density 1 day before transfection) using jetPRIME (Polyplus Transfection) and incubated for 48 h. Total RNA was extracted using the RNeasy Mini Kit (Qiagen) and was subjected to RT-qPCR analysis using SsoAdvanced SYBR Green Supermix and primers listed in S1 Table.
An ORF of RyDEN or BAP was cloned into a lentiviral vector, pYK005C-NTAP-Bla in which a TAP tag consisting of two IgG binding units, a tobacco etch virus (TEV) protease cleavage site, and a streptavidin-binding peptide [44] had been added upstream of the Gateway unit in pYK005C-Bla by LR reaction. A VSV-G-pseudotyped lentiviral vector produced using 293T cells was used to transduce HepG2 cells. The transduced cells were selected in the presence of 10 μg/ml of blasticidin. Expression of N-terminal TAP-tagged RyDEN and BAP proteins were confirmed by immunoblotting using nonspecific rabbit IgG (primary antibody) and HRP-conjugated anti-rabbit IgG as a secondary antibody.
For affinity purification analysis, TAP-fused protein-expressing cells (90% confluence in a 100-mm culture dish) were harvested from a total of 12 dishes, washed twice in PBS that contained 10 mM EDTA, and lysed in 7.8 ml of TAP lysis buffer (50 mM Tris-HCl, pH8.0, 0.5 mM EDTA, 1 mM DTT, 150 mM NaCl, 0.2% NP-40, protease inhibitors) on ice for 40 min. Cell debris was removed by centrifugation for 10 min at 10,000 × g. The supernatants were incubated with 840 μl of IgG Sepharose 6 Fast Flow (50% slurry, GE Healthcare) at 4°C for 2 h. Beads were washed three times with TAP washing buffer (50 mM Tris-HCl, pH 8.0, 0.5 mM EDTA, 1 mM DTT, 300 mM NaCl, 1% NP-40). Proteins were eluted in 1.24 ml of TAP elution buffer (50 mM Tris-HCl, pH 8.0, 0.5 mM EDTA, 1 mM DTT, 150 mM NaCl, 0.2% NP-40) containing 60 U of TEV protease (Life Technologies) at 4°C overnight. The eluted protein sample was concentrated using trichloroacetic acid and separated by a 10% SDS-PAGE gel. Mass spectrometric identification of proteins was performed using MALDI TOF-TOF MS at Protein and Proteomics Centre, Department of Biological Sciences, National University of Singapore.
To construct deletion mutants of RyDEN, cDNA covering amino acid positions 51–291, 101–291, 151–291, and 1–250 were amplified by PCR. Site-directed mutagenesis of RyDEN for substitutions of arginine and lysine to alanine in NLS-L (amino acid positions 121–137) was performed by the overlapping PCR technique using two complementary primers flanking both ends of the RyDEN ORF and two internal mutagenic 25-nucleotide primers. After the first round of PCR, the two mutated DNA fragments (5’ and 3’ parts) were annealed, and a second round of PCR was carried out using the complementary primers. All PCR fragments were gel purified, cloned into pDONR221, and, after confirmation of sequences, transferred to the Gateway unit of pYK005C-Bla. The VSV-G-pseudotyped lentiviral vectors were produced using HEK293T cells and were used to transduce Huh7.5 (for the deletion mutant experiment) or HepG2 (for the NLS-L mutant experiment), followed by selection with blasticidin (10 μg/ml).
Stable cell lines (90% confluence in a 100-mm culture dish) that expressed V5-tagged RyDEN (WT, deletion mutants, and NLS-L mutants), or control RLuc were lysed using 1.1 ml of TAP lysis buffer on ice for 40 min and cleared by centrifugation. Five-hundred microliters of cell lysate were then incubated with 3 μl of an anti-PABPC1 mouse monoclonal antibody (10E10, Santa Cruz Biotechnology) at 4°C for 2 h with rotation, followed by the addition of 30 μl of Protein A/G agarose beads (Santa Cruz Biotechnology) and another 2 h of incubation at 4°C. The bound complexes were washed five times with TAP elution buffer and eluted in SDS sample buffer for immunoblotting analysis. In the co-immunoprecipitation experiments, V5-RLuc was used as a control protein to avoid overlapping with IgG light chain of the anti-PABPC1 antibody (used for pull-down) on immunoblots.
siRNA duplexes that target human PABPC1 (siPABPC1: 5’-AGG CGA UGC UCU ACG AGA AdTdT-3’) and human LARP1 (siLARP1: 5’-GAA UGG AGA UGA GGA UUG CdTdT-3’) and a negative control siRNA duplex (siCtrl) were purchased from SABio (Singapore). HepG2 cells preseeded in a 24-well plate at a density of 1 x 105 cells/well 1 day before transfection were transfected with 50 nM siRNA duplex using jetPRIME and then inoculated with DENV-2 at an MOI of 1 48 h after transfection. Forty-eight hours after infection, the culture supernatant was collected and subjected to plaque assay to determine the viral infectious titer. At the same time, total RNA was extracted from infected cells and used for qRT-PCR to analyze the knockdown efficiency of PABPC1 and LARP1 mRNA using the primers listed in S1 Table.
HepG2 cells that expressed V5- RyDEN (WT and NLS mutants) or V5-DHFR were seeded in a 6-well plate at a density of 5 x 105 cells/well 1 day before infection and exposed to 2.5 x 106 PFU of DENV-2 for 6 h. Cells were then washed with cold PBS three times and lysed with 300 μl of TAP lysis buffer on ice. After centrifugation at 10,000 x g for 10 min, the supernatant was incubated with 3 μl of an anti-V5 mouse monoclonal antibody in the presence of 100 ng/ml of tRNA (Sigma) at 4°C for 2 h with rotation, followed by the addition of 30 μl of protein A/G agarose beads (50% slurry in PBS, Pierce) and another 2 h of incubation at 4°C. The immune complex was washed with 500 μl of TAP washing buffer 5 times and suspended with RNase-free PBS. One-fourth of the suspension was used for immunoblotting to detect V5-tagged proteins, and the rest was used for RNA analysis. DENV RNA was extracted from the suspension using TRIzol (Life Technologies) and subjected to qRT-PCR using DENV 3’UTR-specific primers and a fluorescent probe (S1 Table).
3’UTR sequence of DENV-2 NGC (nucleotide positions 10,271–10,724) was cloned into pEU vector containing SP6 promoter sequence (CellFree Sciences, Japan). A DNA fragment covering the upstream SP6 promoter and the downstream 3’UTR sequences (or DHFR sequence for nonspecific control RNA) was amplified from the pEU-based construct by PCR, which was then used for in vitro transcription in 25 μl of reaction containing 10 mM NTP, 0.25 mM biotinylated UTP (Roche Diagnostics), and 0.8 units/μl SP6 polymerase (CellFree Sciences). Resulting transcripts were column purified, followed by ethanol precipitation to remove free biotinylated UTP.
For production of recombinant proteins, a DNA fragment containing 5’ SP6 promoter, N-terminal tag (consisting of GST and FLAG units, separated by TEV protease cleavage site [GST-TEV-FLAG] for FLAG-tagged proteins [RyDEN WT, RyDEN NLS-L mutant, and DHFR], or GST unit and TEV protease site [GST-TEV] for GST-tagged proteins [PABPC1 and DHFR]), and the protein ORF sequences was amplified from plasmid DNA encoding RyDEN (WT or NLS-L mutant), DHFR, or PABPC1 by previously described split-primer PCR method [74] and used as a template for in vitro transcription. In vitro RNA transcription and subsequent translation of proteins using wheat germ cell-free protein production system were performed in 96-well plate by the bilayer diffusion method using ENDEXT technology (CellFree Sciences) according to the manufacturer’s protocol. The synthesized proteins were captured with glutathione Sepharose 4B (GE healthcare), and the beads were washed with PBS. Proteins were then eluted from beads using elution buffer (50 mM Tris-HCl, pH8.0, 100 mM NaCl) containing 0.4 U/μl TEV protease (for FLAG-tagged proteins) or 10 mM reduced glutathione (for GST-tagged proteins).
In vitro RNA binding assay was performed with 384-well OptiPlate by AlphaScreen technology (PerkinElmer). Twenty nanomolar of FLAG-tagged proteins were mixed with 20 nM of GST-tagged proteins and 3.5 ng/μl biotinylated (or non-biotinylated) DENV 3’UTR RNA (or control RNA) in 15 μl of the binding mixture containing reaction buffer (100 mM Tris-HCl, pH7.5, 100 mM NaCl, 1 mg/ml BSA, 0.01% Tween 20) at 16°C. After 1 h incubation, 10 μl of the detection mixture containing 0.2 μg/ml anti-FLAG mouse monoclonal antibody (Wako), 0.1 μl of streptavidin-coated donor beads and 0.1 μl of protein A-conjugated acceptor beads (PerkinElmer) in reaction buffer was added to the binding mixture, followed by incubation at 16°C for 1 h. Luminescent signal was analyzed by an EnVision microplate luminometer (PerkinElmer) [56].
V5-RyDEN and V5-DHFR-expressing HepG2 cells preseeded in a 12-well plate at a density of 2 x 105 cells/well 1 day before assay were cultured with 10 or 20 μg/ml cycloheximide for 1 h. After the medium was changed, cells were further cultured in the presence of 10 μg/ml of puromycin (Clontech). Cells were harvested 40 min after puromycin pulse, and the cell lysate was subjected to immunoblotting using anti-puromycin mouse monoclonal antibody (3RH11, KeraFAST).
All data are obtained by a representative set of at least three independent experiments, and the average values are shown with error bars indicating the standard deviation (SD). Statistical significance was performed using JMP Pro software version 11 (SAS Institute). P values below 0.05 (P<0.05, *; P<0.01, **; P<0.001, ***) were considered significant.
In this study, the following reference sequences were used to design oligonucleotides: DENV-2 NGC (AF038403.1); C19orf66 (NM_018381); PABPC1 (NM_002568.3); LARP1 (NM_015315.4); BAP (M13345.1); DHFR (J01609.1); ISG54 (NM_001547.4); ISG15 (NM_005101.3); LY6E (NM_002346.2); RIG-I (AF038963.1); IFN-β (M25460.1); GAPDH (NM_002046.5).
|
10.1371/journal.pgen.1007877 | Separate Polycomb Response Elements control chromatin state and activation of the vestigial gene | Patterned expression of many developmental genes is specified by transcription factor gene expression, but is thought to be refined by chromatin-mediated repression. Regulatory DNA sequences called Polycomb Response Elements (PREs) are required to repress some developmental target genes, and are widespread in genomes, suggesting that they broadly affect developmental programs. While PREs in transgenes can nucleate trimethylation on lysine 27 of the histone H3 tail (H3K27me3), none have been demonstrated to be necessary at endogenous chromatin domains. This failure is thought to be due to the fact that most endogenous H3K27me3 domains contain many PREs, and individual PREs may be redundant. In contrast to these ideas, we show here that PREs near the wing selector gene vestigial have distinctive roles at their endogenous locus, even though both PREs are repressors in transgenes. First, a PRE near the promoter is required for vestigial activation and not for repression. Second, only the distal PRE contributes to H3K27me3, but even removal of both PREs does not eliminate H3K27me3 across the vestigial domain. Thus, endogenous chromatin domains appear to be intrinsically marked by H3K27me3, and PREs appear required to enhance this chromatin modification to high levels at inactive genes.
| Eukaryotic genes are packaged in chromatin, and their transcription relies on activators that recruit RNA polymerases and on repressive factors. In multicellular organisms, cell types have distinct patterns of gene expression, and these patterns are controlled by by the expression of cell-type-specific transcription factors and by modulating chromatin structure. The Polycomb system is one major system for the chromatin-mediated silencing of developmental gene expression, where a histone methyltransferase marks extended chromatin domains with trimethylation of lysine-27 of the histone H3 tail (H3K27me3) and forms repressed chromatin. In Drosophila, repressive regulatory elements called Polycomb Response Elements (PREs) are thought to nucleate histone methyltransferase binding which then spreads across these domains. In this study, we demonstrate that two PREs near the developmental vestigial gene have distinct and separable effects on gene activation and chromatin structure. Both PREs are functional repressors in transgenes, but the PRE located near the vestigial promoter is required for gene transcription. This PRE has no effect on histone methylation of the domain. The second PRE located in the middle of the chromatin domain is required for high-level H3K27me3 of the domain, but this methylation is not required to refine vestigial gene expression. Significant chromatin methylation remains when both PREs are deleted. Our findings imply that PREs near promoters may play activating roles in gene expression in the Drosophila genome. We suggest that some domains of H3K27me3 may have little consequence for correctly patterning gene expression.
| The patterns of chromatin histone modifications differ between cell types, reflecting the activity of genes for developmental programs. Tri-methylation of the lysine-27 residue of histone H3 (H3K27me3) typically marks extended chromatin domains, leading to chromatin compaction and epigenetic gene silencing that is maintained as cells differentiate [1,2]. Histone methylation is thought to be initiated at discrete regulatory elements called Polycomb Response Elements (PREs) within domains. These elements bind multiple DNA-binding factors, recruiting the PRC1 and PRC2 complexes, including the Polycomb chromatin factor and the E(z) histone methyltransferase, respectively [3,4]. Transgenes carrying PREs are sufficient to silence reporter genes and to nucleate new H3K27me3 domains [5–7]. However, the function of PREs in their endogenous domains is less clear. Deletion of PREs from the homeobox gene cluster BX-C have limited defects in gene silencing [8–10], but no reduction of histone methylation of this domain. While multiple PREs within the BX-C domain may be redundant, deletion of all mapped PREs near the invected and engrailed genes have no effect on methylation of the locus, and it remains unknown how histone methylation is maintained [11].
Genomic mapping has identified regions where both the PRC1 and PRC2 complexes colocalize, and regions where each complex is found separately. Only about one-half of all Polycomb binding sites are within H3K27me3 domains, and thousands of additional sites are located near the promoters of active genes [12,13], where they may modulate gene expression by holding RNAPII at paused promoters. Here, we characterize the in vivo roles of two PREs near the vestigial gene. While these two PREs are silencers in transgene assays, targeted mutations reveal that the promoter PRE is required for full gene expression. Using a new efficient method for genomic mapping of chromatin factors, we demonstrate that methylation across the domain remains in the absence of both PREs. Our results reveal that PREs stimulate but are not necessary for domain methylation.
To profile chromatin domains in different tissues, we used a chromatin mapping strategy that tethers micrococcal nuclease at factor binding sites. In the CUT&RUN procedure [14], unfixed cells are soaked with a factor-specific antibody, which binds to chromatin. Next, a protein-A-micrococcal nuclease (pA-MNase) fusion protein is soaked in, binding to the chromatin-bound antibody. Activation of the tethered MNase by adding calcium then cleaves exposed DNA around the binding sites of the targeted factor. Sequencing of the cleaved DNA fragments thus maps the location of the targeted chromatin protein. CUT&RUN obviates the need to work with chromatin preparations or to optimize affinity recovery of chromatin particles, and works efficiently with small numbers of cells [15,16]. To implement CUT&RUN for tissue samples, we simply dissected brains and wing imaginal discs from ten larvae, lightly permeabilized the whole tissues with digitonin, and sequentially incubated the tissues with antibody to H3K27me3 and then with pA-MNase. MNase was then activated and finally the cleaved DNA was isolated, subjected to Illumina paired-end sequencing, and mapped to the Drosophila dm6 genome assembly. We similarly mapped the Polycomb protein, which binds at Polycomb Response Elements (PREs).
H3K27me3 domains in larval tissues have been previously mapped by Chromatin Immunoprecipitation [17]. Profiles of H3K27me3 distribution generated by CUT&RUN using substantially less material were similar (Pearson’s r = 0.94). Both methods reveal changes in chromatin methylation that correspond to tissue-specific changes in gene expression. For example, the ANTENNAPEDIA-COMPLEX (ANTP-C) cluster of homeobox genes are encompassed in a H3K27me3 domain in larval brains, consistent with the predominant silencing of this cluster in this tissue [18] (Fig 1A). In contrast, in wing imaginal discs where Antp is transcribed, chromatin over most of this gene is depleted for H3K27me3. Interestingly, some histone methylation remains across the 3’ exons of Antp, which indicate that a shorter isoform of the Antp gene may be transcribed in this tissue. Notably, histone methylation is not completely eliminated from the transcribed Antp gene (Fig 1A).
To quantify changes in chromatin landscapes between tissues, we measured the read count coverage at 125 chromatin domains (listed in S3 Table) with high H3K27me3 in larval brain and wing disc samples. Most chromatin domains are similarly methylated between these tissues, but a small number of domains have lower read counts for H3K27me3 in wing discs compared to larval brains (Fig 1B). One group of domains have low levels of histone methylation in larval brains and lose methylation in wing discs, however, genes in these domains are not expressed in either tissue. A second set of domains encompass genes that are expressed in wing discs but not in brains, including apterous (ap), nubbin (nub), vestigial (vg), and Drop (Dr) (Fig 1B, blue), and histone methylation across these domains is lower in wing discs. Histone methylation is not eliminated, as 25–50% of H3K27me3 levels remains even when domain genes are transcribed, and this is noticeably greater than background levels at random regions outside of domains (red in Fig 1B, Fig 1C). Thus, activation of these genes is accompanied by reduction–but not loss–of the H3K27me3 modification.
We focused on the vg gene locus as a simple model. The vg gene is required for wing determination, is the only gene in a 32 kb H3K27me3 chromatin domain (Fig 1A). The vg gene is inactive in brain tissues, and is expressed only in the pouch of wing imaginal discs. This domain is heavily methylated in larval brains, but reduced to ~50% across the domain in wing discs (Fig 1A and 1C). This is consistent with loss of H3K27me3 when the vg gene is activated, but wing discs are a mixture of cells with and without vg expression. We therefore isolated vg-expressing cells to profile the chromatin status of the active gene. We made a transgene construct containing the vg Quadrant enhancer [19], and the GAL4 transcriptional activator, and used this to drive expression of GFP in the wing pouch. We then used FACS to isolate GFP-positive cells and profiled these cells by CUT&RUN. As expected, profiles of these cells show reduced histone methylation across wing-specific genes like Antp (Fig 1A, Supplementary Information). We found that H3K27me3 across the vg domain is reduced to ~20% of it’s levels in non-expressing cells, but remains ~4-fold more methylated than background levels across the genome (Fig 1A and 1C). Thus, while gene activation is associated with elimination of the H3K27me3 modification, a small amount of methylation remains across many activated domains.
Two potential PREs within the vg domain have been identified by sequence motifs [20] and by chromatin profiling [12,13,21]. The first region, which we term the proximal PRE (pPRE), is located 300 bp downstream of the mapped Transcriptional Start Site (TSS) of the vg gene, which was previously mapped by primer extension [22]. The second distal PRE (dPRE) is located ~25 kb downstream (Fig 2A). We used CUT&RUN to map the Polycomb chromatin protein in larval tissues, and found that this protein is bound at both of these PREs in larval brains (Fig 1A). Surprisingly, Polycomb binding is detectable at both PREs in both wing discs and even in FACS-isolated vg-expressing cells (Fig 1A). We quantified the amount of Polycomb and found that Polycomb is retained to similar levels at the pPRE of the active vg gene, and at reduced levels at the vg dPRE (Fig 1D). These levels are clearly above background levels at a non-target promoter or an insulator element. We also observed retention of Polycomb at the promoters of the Dr and nub genes, which are active in wing pouch cells (Fig 1D). The promoter of the ap gene which changes methylation between brain and wing disc samples has very low levels of Polycomb in repressed brain samples, and is indistinguishable from background in wing pouch cells (Fig 1D). These results indicate that in some cases Polycomb remains present at de-repressed genes.
To test the silencing effects of these PREs, we created transgenes including these two regions, and integrated these at the same landing site in the Drosophila genome. We then tested if these transgenes induce pairing-sensitive silencing (PSS), a diagnostic feature of PREs where a transgene reporter gene is silenced in homozygous animals [23]. Previous studies showed that a dPRE-containing transgene will cause PSS [24], and a similar transgene with the 1 kb dPRE region integrated at the landing site also causes PSS (Fig 3). We found that a transgene with the 300 bp pPRE region also shows strong PSS, demonstrating that the pPRE is also a silencing element. These PREs can interact with each other and cause silencing, as animals heterozygous for a pPRE transgene in one landing site and a dPRE transgene on the homolog also show PSS (Fig 3).
CUT&RUN for Polycomb defined a 200 bp segment where Polycomb binds near the vg promoter (Fig 2A). We used high-resolution mapping by native ChIP in Drosophila S2 cells to precisely define binding sites for three juxtaposed Polycomb-bound sites, one of which is also bound by the Pleiohomeotic (PHO) transcription factor (S1 Fig). Deletion of one of these peaks from the pPRE transgene alleviates PSS (Fig 3), thus this sequence is required for reporter silencing. We analyzed the dPRE similarly. We found that while a transgene including the dPRE induces PSS, deletion of the Polycomb-bound site within the dPRE alleviates this silencing (Fig 3). We conclude that the Polycomb-bound sites in both the pPRE and dPRE elements are required for transgene silencing.
To measure silencing at the endogenous vg locus, we integrated reporter genes by gene-targeting near the pPRE and near the dPRE. The promoter of the engrailed gene is active in the posterior half of the wing imaginal disc [25], including part of the expression domain of vg in the wing pouch. We used an engrailed-GAL4 (en-GAL4) transgene [26] to drive expression of GAL4-dependent UAS-YFP and UAS-RFP reporters in the wing disc. Control reporter gene insertions produce RFP and YFP throughout the posterior half of the wing disc (Fig 4A). In contrast, UAS-YFP reporters inserted in the vg domain are silenced throughout most of the wing disc, with reduced expression only within the posterior part of the wing pouch (Fig 4A). Insertions near the dPRE show similar reduced expression in the wing pouch and silencing in the rest of the wing disc. Thus the vg domain appears to be packaged in repressed chromatin in most of the wing disc, but in derepressed chromatin in the wing pouch (Fig 2A).
We confirmed that silencing in the wing disc is mediated by chromatin by expressing a dominant-negative H3.3K27M mutant histone to reduce chromatin levels of H3K27me3 [27] in the posterior half of the wing disc. Indeed, expression of the mutant histone derepressed the GFP reporter gene throughout the posterior half of the wing disc (Fig 4A). This demonstrates that the vg domain is in two chromatin states in the wing imaginal disc: a silenced configuration, and a derepressed configuration in wing pouch cells where vg is normally expressed. However, expression of the mutant histone does not derepress expression of the vg gene itself (Fig 4B). Thus, the silenced configuration appears to only affect the inserted reporter gene.
To define the function of the PREs near the vg gene, we deleted each element from the endogenous locus (see Methods). Precise breakpoints for each of the recovered deletions were determined by Sanger sequencing, and tested against each other and against previously characterized vg alleles (Fig 2B and 2C; S1 Table; S2 Table). We used the vgnw allele—which deletes part of the coding region of the gene—as a null allele. Ectopic expression of vg converts legs and eyes into wing-like structures [28]; thus deletion of silencing elements should derepress the vg gene and transform non-wing tissues. However, we observed no such transformations in animals homozygous for deletion of the pPRE or of the dPRE. Deletion of both PREs from the vg domain did not transform non-wing tissues. Thus, there appears to be no role for Polycomb silencing in limiting vg expression.
Surprisingly, we found that deletions of the pPRE reduce expression of the vg gene. Wing development is sensitive to the amount of Vg protein, and reduction in vg expression results in progressive notching and deletion of the wing. We found that animals carrying the vgCL1 or vg13A deletions are viable, but have severely reduced wings (Fig 5A). A smaller deletion (vgCZ) which deletes only one of the Polycomb-binding sites within the pPRE has a more limited effect. The vgCZ allele gives no phenotype as a homozygote, but in combination with a null allele adults have notched wings which characteristic of weak vg alleles. We also recovered a similarly weak allele (vgCL2C) that is a single-base pair C-to-T substitution in the pPRE. This substitution lies precisely at the center of the major Polycomb-bound site, in a sequence similar to consensus motifs for Sp1 transcription factors which direct Polycomb binding [29]. Each of these mutations are recessive and viable alleles, and thus are distinct from deletions of the vg promoter, which are recessive lethal alleles [22]. Thus, the pPRE appears to positively contribute to vg expression. In contrast, animals with a deletion of the dPRE (the vgR5 allele) are viable with normal wings, either as a homozygote (Fig 5A). Animals lacking both PREs are wingless like the pPRE single mutant (Fig 5A).
The adult wing blade differentiates from the pouch of the larval wing imaginal disc where vg is expressed, and we therefore imaged Vestigial protein in wing discs of larvae. Animals carrying the pPRE deletion vgCL1 have small wing discs where the wing pouch is reduced, no Vestigial protein is detectable, and the central stripe of Wingless (Wg) signalling ligand that marks the edge of the future wing blade is absent (Fig 5B). The smaller pPRE deletion vgCZ and the point mutant vgCL2C have more limited effects: Vg is produced, but with occasional gaps in wing discs (Fig 5B, yellow arrowheads). These gaps are often associated with gaps in the central Wg stripe, consistent with the notching of adult wings. In contrast, the dPRE deletion vgR5 has no extra staining or defects of Vg, and the central Wg stripe is continuous. Finally, animals carrying both the pPRE and dPRE double deletions have small wing discs similar to the single pPRE deletion, with no Vg protein detected in wing discs (Fig 5B). We conclude that the pPRE is required for expression of the vg gene in wing discs, but the dPRE is not.
We then tested if Polycomb factors are required at the pPRE for vg expression. The vgCL2C basepair substitution is a recessive allele, and animals heterozygous for this mutation have no phenotype. However, in combination with a heterozygous Pc3 allele, vgCL2C/+ animals have deformed wings (Fig 6A). The wings of vgCL2C/vgnw; Pc3/+ animals have more enhanced wing notching, while control Pc3/+ siblings have normal wings (Fig 6A; Table 1). Wing discs from vgCL2C/vgnw; Pc3/+ larvae show gaps in Vg staining and gaps in the central Wg stripe, consistent with the adult notching (Fig 6B). Similarly, mutations in the RING1b homolog Sce enhances the phenotype of vgCL2C/vgnw animals (Fig 6A). These effects suggest that PRC1 components bound at the promoter in active cells (Fig 1D) positively influence vg expression. Finally, combining Polycomb mutations with the vgR5 deletion show no wing defects (Table 1), demonstrating that the genetic interaction between Polycomb and the pPRE allele is specific.
Regulatory structures are present in the 5’UTRs of some transcripts. However, it is unlikely that the mutations we created affect an mRNA function, because they coincide precisely with the chromatin features of the pPRE. Further, the single-base pair mutation vgCL2C is enhanced by Polycomb mutations, supporting the idea that it affects the function of the chromatin element, not an mRNA function. It is unusual for PRC1 to be implicated in transcriptional activity, but there are examples. In one case in the mouse midbrain, Polycomb is required to bring enhancers to the meis2 gene promoter before meis2 is expressed, but then is not required after induction [30]. To test if the vg pPRE is similarly required before activation of the vg gene or if the pPRE is required in cells expressing vg, we generated cells homozygous for a pPRE deletion from heterozygous cells by FLP recombinase-mediated mitotic recombination at different times in development [31]. The pPRE/+ heterozygous animals have no wing defects, but FLP expression produces animals with a range of defects in the wing blade, ranging from notches in the wing margin to complete loss of one wing (Fig 6C), implying that pPRE mutant clones lose vg expression whenever they are induced. Together, these results indicate that the pPRE is continually required to maintain expression of the vg gene.
In transgenes, a PRE is required to nucleate and maintain a Polycomb-regulated domain by recruiting PRC1 and PRC2 complexes [6,7]. We tested if histone methylation of the vg domain depends on the pPRE or on the dPRE. We profiled the chromatin of wildtype and PRE deletion mutants in larval brains, where the vg gene is not active. H3K27me3 levels are high across the vg domain in wildtype larval brains (Fig 7). However, there is no reduction in H3K27me3 across the vg domain In animals lacking the pPRE. In contrast, histone methylation is reduced to ~45% of wildtype levels in animals lacking the dPRE (Fig 7). Finally, histone methylation is reduced to ~20% wildtype levels when both PREs deleted. These results indicate that the dPRE is predominantly responsible for histone methylation of the vg domain, although the pPRE can also contribute to domain methylation. Notably, the residual methylation across the vg domain when both PREs are deleted remains higher than background levels of H3K27me3 in the genome. The amount of residual methylation is similar to that in cells where the vg gene is active, suggesting that this is the minimal level of H3K27 methylation of this domain.
It is possible that minor or cryptic PREs in a domain may direct histone methylation when major PREs are deleted [11]. We therefore profiled Polycomb binding in larval brains from wildtype and PRE deletion mutants, normalizing landscapes to peak heights in the ANTP-C domain (Fig 7A, bottom). Deletion of either the pPRE or the dPRE eliminates only its peak of Polycomb binding. We observe no Polycomb binding when both PREs are deleted, suggesting that there are no alternative or cryptic PREs in the vg domain. Thus, the low level of histone methylation across the vg domain appears to be independent of Polycomb binding sites.
The concept that Polycomb-repressed chromatin domains are nucleated at short factor-binding regulatory elements (PREs) derives from the phenotypes of deletions within the homeobox clusters in Drosophila, where PREs are required for silencing of these genes [32]. However, it has been difficult to determine if PREs control histone modifications in these domains. While transgenes can confer H3K27me3 modification onto their insertion sites [6,33,34], the effects of deleting PREs from endogenous domains has been more ambiguous. In part these ambiguities may be due to the difficulties of measuring changes in histone modifications with limited samples. Using the more efficient CUT&RUN method, we find that we can produce detailed chromatin profiles from small samples with high sensitivity, allowing characterization of specific tissues in mutant animals.
The developmental gene vestigial is contained within chromatin that has the features of a Polycomb chromatin domain, being marked by the H3K27me3 histone modification and silencing inserted reporter transgenes. While the two PREs from this domain both act as silencers in transgenes, our results show that they have distinct roles at the endogenous locus. One PRE is primarily responsible for histone methylation of the domain, but has no effect on silencing or expression of the vg gene. Previous studies have also suggested that PREs differ in their effects. For example, one of two PREs at the dachshund locus also directs methylation of its domain [35]. In contrast, deletions of the known PREs of the engrailed locus had no effect on domain methylation [11]. Cryptic PREs or non-coding RNAs have been proposed to direct histone methylation in these situations. However, there we found that no new PREs appear in the vg domain, and while non-coding RNAs have been identified near one of the PREs in the vg domain [36], these RNAs are deleted in our mutants. Thus, the vg domain appears to have a low level of undirected H3K27-methylation. Such domain methylation might be directed by other histone modifications. Active chromatin regions are often marked by H3K27-acetylation, which is antagonistic to H3K27 methylation [37]. Further, in Drosophila cells, the H3K27 methyltransferase E(z) acts globally and dimethylates ~50% of all nucleosomes [38]. Perhaps sequence features of some regions predispose unacetylated regions to accumulate H3K27-trimethylation, and these regions can then reach high levels of H3K27me when PREs are active.
The second PRE in the vg domain is distinct; while it has little effect on domain methylation, it is required for normal vg expression. PRE localization near promoters is a common feature of the Drosophila genome [12,39], and are well-positioned to regulate gene activity. Polycomb can silence gene expression by inhibiting transition of RNA polymerase II (RNAPII) to its elongating form, and this is a major step for controlling the expression of developmental genes [40,41]. However, ~1000 active promoters in the Drosophila genome are also bound by Polycomb [12]. While this binding has been suggested to reduce transcriptional output, loss of Polycomb results in both loss of silencing at some genes and decreased transcription at others [39]. In genome-wide studies it has been difficult to determine if downregulation is due to pleiotropic effects or a requirement for Polycomb at some genes, The vg gene is the first example of a promoter that requires a PRE for expression during development. Positive effects may be mediated by PREs looping together enhancers and promoters [30] as well as silencing elements. The effects of a specific PRE may then depend on what regulatory elements it brings to a target promoter. Indeed, PREs have been demonstrated to switch between silencing and activating states [42], and in mammals variant Polycomb complexes have been described that activate developmental genes [43,44]. Our observations that the vg pPRE can both silence a reporter gene and promote expression of the endogenous gene suggests that promoters differ in their interactions with PREs, and this may be critical to integrate Polycomb regulation with developmentally-programmed enhancers.
All crosses were performed at 25°C. Transgenes, mutations and chromosomal rearrangements not detailed here are described in Flybase (http://www.flybase.org). The vgnw allele is a deletion of the last two exons of the vg transcript, and so we used this as a standard null allele. New alleles of vg produced in this study are described in S1 Table.
|
10.1371/journal.pntd.0002743 | Identification of Carboxylesterase Genes Implicated in Temephos Resistance in the Dengue Vector Aedes aegypti | Thailand is currently experiencing one of its worst dengue outbreaks in decades. As in most countries where this disease is endemic, dengue control in Thailand is largely reliant on the use of insecticides targeting both immature and adult stages of the Aedes mosquito, with the organophosphate insecticide, temephos, being the insecticide of choice for attacking the mosquito larvae. Resistance to temephos was first detected in Aedes aegypti larvae in Thailand approximately 25 years ago but the mechanism responsible for this resistance has not been determined.
Bioassays on Ae. aegypti larvae from Thailand detected temephos resistance ratios ranging from 3.5 fold in Chiang Mai to nearly 10 fold in Nakhon Sawan (NS) province. Synergist and biochemical assays suggested a role for increased carboxylesterase (CCE) activities in conferring temephos resistance in the NS population and microarray analysis revealed that the CCE gene, CCEae3a, was upregulated more than 60 fold in the NS population compared to the susceptible population. Upregulation of CCEae3a was shown to be partially due to gene duplication. Another CCE gene, CCEae6a, was also highly regulated in both comparisons. Sequencing and in silico structure prediction of CCEae3a showed that several amino acid polymorphisms in the NS population may also play a role in the increased resistance phenotype.
Carboxylesterases have previously been implicated in conferring temephos resistance in Ae aegypti but the specific member(s) of this family responsible for this phenotype have not been identified. The identification of a strong candidate is an important step in the development of new molecular diagnostic tools for management of temephos resistant populations and thus improved control of dengue.
| Temephos is the most important insecticide used in larviciding campaigns to reduce the risk of dengue transmission. This organophosphate insecticide has been in use for over 50 years and resistance to this chemical has been reported in Aedes aegypti populations from Latin America, the Caribbean and from Asia. In other insect species, organophosphate resistance is typically associated with mutations in the target site, acetylcholinesterase, that decrease the insect's sensitivity to the insecticide, or increases in the activity of one or more carboxylesterase enzymes, either by overproduction and/or amino acid substitutions, that reduce the amount of insecticide reaching the target site. Neither of these mechanisms has been previously characterised at the molecular level in dengue vectors. Here we identify an Ae aegypti carboxylesterase gene with expression levels and amino acid sequence polymorphisms correlating with temephos resistance in Thailand. This is a key step in the development of tools to manage resistance in this mosquito species.
| Aedes aegypti is a major vector of dengue fever and yellow fever viruses. Despite an effective vaccine, there are over 200,000 cases of yellow fever each year (WHO source, 2012). With no vaccine currently available for dengue, and no specific drug treatment, approximately 40% of the world's population is at risk of dengue fever and there may be as many as 390 million dengue infections per year [1].
Dengue is endemic in Thailand with the most severe manifestation of dengue, dengue haemorrhagic fever first reported in 1958 [2]. The number of dengue cases has been steadily increasing since 2009 with over 81,000 cases already reported in the first 7 months of 2013 and, predictions of between 100,000 and 120,000 cases for the whole year (Department of Disease Control, Thailand Ministry of Public Health, http://www.ddc.moph.go.th/).
Maintaining Ae. aegypti populations at low levels is crucial for dengue control in Thailand [3]. Environmental management including educational campaigns to remove unnecessary sources of standing water, coupled with covering of permanent water storage vesicles, is recommended to help reduce Aedes populations [4] but this is supplemented by the use of chemical insecticides. In Thailand, adult mosquitoes are predominately targeted with pyrethroid insecticides [5], mainly through the distribution of pyrethroid impregnated materials and the Ultra-Low-Volume (ULV) applications of pyrethroids [6]. Larval control primarily utilises the organophosphate insecticide, temephos, (Department of Disease Control, Thailand Ministry of Public Health) despite the known existence of temephos resistant populations of Ae. aegypti in many regions of Thailand [7], [8].
An understanding of insecticide resistance mechanisms is important for the development of tools and practices that can improve resistance management and thereby the sustainability of control interventions. In many insect species, organophosphate and carbamate resistance is caused by amino acid substitutions in the target site, acetylcholinesterase (ace-1), which reduces the sensitivity of this enzyme to the insecticide. The most common ace-1 substitution in mosquitoes occurs at amino acid residue 119 where the wild type glycine is substituted to serine [9]. However, in Ae aegypti, the codon usage at Glycine 119 makes this substitution very unlikely to occur [10]. Indeed, despite numerous reports of temephos resistance in Ae aegypti populations across the tropics, including at least one report of insensitive AchE [11], no target site mutations linked to organophosphate resistance have been detected to date. Organophosphate resistance can also be caused by elevated levels of esterase enzymes that can both act to sequester the insecticide, reducing the amount of active insecticide that reaches the target site [11], or to increase the rate of turnover of insecticide, by amino acid substitutions in the coding sequences of one or more esterases [12].
Elevated CCE activity has been associated with temephos resistance in several populations of Ae aegypti [13], [14], [15], [16], [17]. A small number of studies [16], [17], [18] have used microarray based approaches to detect genes associated with the resistance phenotype. Although several transcripts of detoxification genes were found to be evelated in temephos resistant populations (including CCEae3a, CYP6Z8 and CYP9M9), a single clear candidate did not emerge from these studies. The current study provides evidence for elevated CCE activity in a temephos resistant population from Thailand and identifies a clear candidate gene that shows both elevated expression and amino acid polymorphisms in temephos resistant populations. Additional genes, potentially involved in temephos and/or permethrin resistance in Ae aegypti larvae are identified and discussed.
Mosquito eggs were collected from four sites of Thailand including Chiang Mai (North, 18°47′25″N 98°59′4″E, 31st October 2011), Nakhon Sawan (central, site 1 : 15°20′45″N 100°29′41″E, 5th March 2012, site 2: 15°52′52″N 100°18′9″E, 27th March 2012) and Phatthalung (south, 7°37′6″N 100°4′24″E, 2th September 2011) (Figure S1). They were chosen based on previous reports of temephos resistance in these districts [8], [19], [20].
Aedes aegypti eggs from Phatthalung and Chiang Mai were collected using modified ovitraps by entomologists from the Department of Disease Control (Ministry of Public Health, Thailand). Eggs from Nakhon Sawan sites were collected by entomologists from office of Disease Prevention & Control 8 (DPC8, Nakhon Sawan). The modified ovitraps consisted of a dark plastic cup with a piece of filter paper over the inner part of the cup and filled with tap water. They were placed in the resting sites of Ae aegypti such as under sinks, beds, cupboards or any cool, humid and dark areas in and around the house. Eggs were then sent to the Liverpool School of Tropical Medicine (LSTM) where they were hatched in distilled water and reared in standard insectary conditions (temperature: 28+/−1°C; relative humidity: 75+/−5%; photoperiod: 12 hours day/night). An insecticide susceptible laboratory colony, New Orleans (NO) strain was used as control in the study. This population was originally collected in the namesake city located in Louisiana, United States.
Standard WHO larval bioassays were conducted to detect the level of susceptibility to temephos [21]. Bioassays were done on late 3rd/early 4th instar larvae using a range of seven temephos (Pestanal, analytic standard, diluted in ethanol) concentrations. Concentrations of insecticides were chosen in order to cover larval mortality range (0–100%). Three replicates of 20 larvae were used for each concentration and 1 ml ethanol was added in control cups. Mortality was recorded after 24 hours of exposure. Larval bioassays using permethrin were also performed to look for any evidence of cross resistance between insecticide classes.
Synergist bioassays were performed on the populations showing the highest temephos resistance levels using a cytochrome P450 inhibitor, piperonylbutoxide (PBO) at 0.3 ppm (piperonylbutoxide 90%, Sigma Aldrich, Inc., Italy), a glutathione S-transferase inhibitor, diethyl maleate (DEM) at 1 ppm (diethyl maleate >97.0% (GC), Sigma Aldrich Chemie GmbH, Austria) and a carboxylesterase inhibitor, S,S,S-tributylphosphorotrithioate (DEF) at 0.5 ppm (S.S.S-tributylphosphorotrithioate 98.1%, Chem service, Inc., USA). Inhibitors were mixed with insecticide dilutions in ethanol and 1 ml of the mixture was added to 99 ml of water according to the protocol of [22]. Different concentrations of synergists were previously tested in order to establish appropriate sub-lethal concentrations [22]. PBO was also used as a synergist in permethrin bioassays. To determine the LC50s and confidence intervals data were analyzed using a Probit model on R software [23].
Activity levels of α esterases and β esterases were measured in the Nakhon Sawan 2 population (NS2), which showed the highest resistance ratio to temephos, and in Phatthalung, the population most susceptible to temephos and permethrin. Procedures were based on mosquito-specific biochemical assay protocols [24], [25], [26]. Briefly, 15 larvae from NS2 and Phatthalung were individually homogenized in 3 mL of 0.01 M potassium phosphate buffer (KPO4), ph 7.2 and 100 µl of each sample homogenate were then transferred by triplicate to a 96-well microtiter plate. Then, 100 µl of α/β naphthyl acetate (3 mM) were added to each well, followed by 15 minute incubation at room temperature. Finally, 100 µl of dianizidine (4 mM) were added, followed by 4 minute incubation, and then absorbance was read at a wavelength of 540 nm. Absorbance values where normalized by measuring protein content using a Bradford assay according to manufacturer's protocol (Sigma, St Louis, MO). Data significance was compared using a Mann-Whitney test (N = 15).
The most resistant population Nakhon Sawan 2 was chosen for the microarray experiment. Phatthalung was used as susceptible population because of its geographical proximity (Figure S1). Three groups of early 4th instar larvae (15 larvae each) were used for total RNA extractions: Phatthalung (P), Nakhon Sawan 2 unexposed (NS 2 Unexp) and Nakhon Sawan 2 larvae (NS 2 Exp) which survived a temephos bioassay inducing 60% mortality (24 hour exposure to 0.032 ppm temephos). Surviving larvae were left to recover in clean water for 24 hours after exposure to reduce the impact of short term gene induction on the transcriptomic profile. The Arcturus Picopure RNA Extraction Kit (Arcturus, California, USA) was used according to the manufacturer's protocol and 100 ng total RNA per biological replicate were amplified and labelled with Cy-5 and Cy-3 dyes with the ‘Two colors low input Quick Amp labeling kit’ (Agilent technologies, Santa Clara, CA, USA) according to manufacturer's instructions. Labelled cRNA were purified with the Qiagen RNeasy kit (Qiagen, Hilden, Germany). Quantification and quality assessment of labeled cRNA were performed with the Nanodrop ND-1000 (Thermo Scientific, DE, USA) and the Agilent 2100 Bioanalyser (Agilent Technologies).
Microarray hybridizations were performed with the 15 k Agilent “Aedes microarray” (ArrayExpress accession number A-MEXP-1966), containing eight replicated arrays of 60-mers oligo-probes representing 14,204 different Ae. aegypti transcripts from AaegL1.2 Vectorbase annotation and several control probes. For each comparison, five hybridizations were performed including two dye-swaps in which the Cy3 and Cy5 labels were swapped between samples. After 17 h hybridization, non-specific probes were washed off with the Agilent microarray hybridization kit according to manufacturer's instructions. Slides were scanned immediately with an Agilent G2205B microarray scanner. Spot finding and signal quantification for both dye channels were performed using the Agilent Feature Extraction software (Agilent Technologies). Data were then loaded into Genespring GX (Agilent Technologies) for normalization and statistical analyses. For each population comparison, only transcripts flagged ‘present or marginal’ in four of five hybridizations were used for further statistical analysis. Mean transcription ratios were then submitted to a one sample Student's t-test (N = 3) against the baseline value of 1 (equal transcription level in both populations) with Benjamini and Hochberg's multiple testing correction. For each selected population, transcripts showing a >2 fold change in either direction and a t-test P-value lower than P<0.01 after multiple testing correction were considered significantly differentially transcribed compared to the susceptible population. Descriptions and GO-terms of transcript-IDs were extracted from VectorBase (www.vectorbase.org) using BIOMART and completed with Blast2GO software (BioBam Bioinformatics S.L. (Valencia, Spain)). GO term Enrichment analysis was performed on the significant up-regulated genes found in both comparisons “NS2 exp vs P” and “NS2 Unexp vs P” using Blast2GO software and Fisher's exact test with FDR<0.05 according to [27]. All microarray data were uploaded to Arrayexpress (E-MTAB-1934, www.ebi.ac.uk/arrayexpress/).
Transcription levels of six genes (four P450s, one CCE and one ABC transporter) found significantly differentially transcribed in at least two comparisons were validated by reverse transcription followed by real-time quantitative PCR (RT-qPCR) as described in [28]. As a secondary control, the susceptible New-Orleans (NO) population was included. Two micrograms of total RNA per biological replicate were treated with DNAse I (Invitrogen, Carlsbad, CA, USA) and used for cDNA synthesis with superscript III and Oligo-dT20 primer (Invitrogen) according to manufacturer's instructions and resulting cDNAs were diluted 50 fold. Real time quantitative PCR reactions of 25 µL were performed on a MX3005P qPCR machine (Agilent technologies, CA, USA) using Brilliant III ultrafast SYBR green mastermix (Agilent technologies, CA, USA), 0.3 mM of each primer and 5 µL of diluted cDNAs. A melt curve analysis was performed to check for the unique presence of the targeted PCR product. Quantification of transcription level was performed according to the ΔΔCt method taking into account PCR efficiency [29] and using two housekeeping genes for normalization: the ribosomal proteins L8 (AAEL000987) and S7 (AAEL009496). Results were expressed as mean transcription ratio (±95% confidence intervals) between Nakhon Sawan 2 and the susceptible populations New Orleans and Phatthalung. All primer sequences are included in supplementary table S5.
Three different groups of 4th instar larvae were used: P, NS2 unexposed and NS2 exposed mosquitoes. NS2 exposed mosquitoes were survivors of a temephos bioassay inducing more than 80% mortality after 24 hours. Genomic DNAs were extracted from 8 individual larvae per group using DNeasy Blood and Tissue Kit according to manufacturer's instructions (Qiagen, Hilden, Germany) and were treated with RNAse A (Qiagen, Hilden, Germany) to remove any RNA contaminants. DNA quantities were assessed on a Nanodrop ND-1000 spectrophotometer. Quantitative PCR reactions were performed as described above on CCEae3a gene (same primers used above) with AAEL000987 (RPL8) and AAEL012167 (Elongation factor) (see table S5 for primer sequences) as housekeeping genes. The relative copy number fold-change was calculated using the 2−ΔΔCt method.
To identify any amino acid polymorphisms that might be associated with temephos resistance, sequencing of CCEae3a cDNA sequence was performed on Nakhon Sawan 2 larvae which survived a concentration of temephos inducing 90% mortality and on unexposed Phatthalung larvae. Total RNAs from 10 individual larvae were extracted using Trizol according to the manufacturer's instructions (Invitrogen, Carlsbad, USA) and total RNA quantities were assessed using a Nanodrop ND-1000 (Thermo Scientific). Genomic DNA contaminants were then digested using DNase I (Invitrogen) and total RNAs were reverse transcribed according to the same protocol used for qPCR validation. Primers were designed (Table S1) to amplify the whole CCEae3a sequence available on Vectorbase (AAEL005112-RA, www.vectorbase.org). PCR amplification was carried using Phusion High-Fidelity DNA Polymerase (Thermo Scientific) using the following conditions: Initial denaturation at 98°C for 30 seconds followed by 35 cycles of 10 sec denaturing at 98°C, 20 sec annealing at 66°C and one minute extension at 72°C. Last extension step 72°C last during 10 min. PCR products were visualized on a 1% agarose gel and purified using a GeneJET Gel Extraction Kit (Fermentas, Vilnius, Lithuania). The PCR products were cloned into DH5 competent cells using pJET 1.2/blunt Cloning Vector kit (Fermentas, Vilnius, Lithuania). Plasmids were extracted using GeneJET Plasmid Miniprep Kit, (Fermentas) and sequenced (Macrogen, Amsterdam, the Netherlands) using pJET primers and two internal primers (Table S5).
The secondary structure and three-dimensional structure of the different polymorphic variants of CCEae3a were predicted by the Protein Homology/analogY Recognition Engine (PHYRE2) (Structural Bioinformatics Group, Imperial College, London). This method uses structural alignments of homologous proteins of similar three-dimensional structure in the structural classification of protein databases to obtain a structural equivalence of residues. The top 20 highest scoring matches of the query to known template structures are used to construct 3D model of the query.
The Phatthalung (P) populations showed the lowest LC50 to temephos and resistance ratios were calculated compared to this population, and according to the standard laboratory susceptible New Orleans (NO). NS 2 showed the highest resistance to temephos (RR at LC50 = 5.9 –9.85 fold) followed by NS 1 (RR at LC50 = 3.3–5.5 fold) and CM (RR at LC50 = 2.1–3.5 fold) (Table 1). Larval bioassays using permethrin showed much higher LC50s in both NS1 and NS2 populations compared to P (RR at LC50 = 29.1 and 31 fold respectively) and intermediate LC50 in the CM population (RR = 8.2 fold) (Table 1). Although permethrin larval bioassays were not performed on a standard lab susceptible strain in this study, two previous studies have reported lab susceptible LC50 for permethrin as approximately 0.0007 ppm [8], [30] which is similar to the 0.0005 value obtained for the P population in the current study.
Synergist bioassays were performed on both NS 1 and NS 2 populations. The use of temephos + PBO or DEM had no significant effect on NS 1 and NS 2 compared to temephos treatment alone. However, the DEF treatment significantly improved the toxicity of temephos by 3.14 fold in NS 1 and 2.48 fold in NS 2 compared to temephos alone. Finally the use of PBO+permethrin in combination showed an improved efficacy by more than two fold in NS 2 larvae compared to permethrin alone.
Comparison of constitutive detoxification enzyme activities between the susceptible population Phatthalung and the most insecticide-resistant population NS 2 revealed increased α- and β-carboxylesterase activities in NS 2 compared to P (2.9 fold and 3.8 fold with P<0.05) (Figure 1).
By using a microarray approach, we detected 2484 transcripts significantly differentially regulated between NS2 Exp and Phatthalung, 2508 between NS2 Unexp and P and 0 between NS2 Exp and NS2 Unexp (RC) (Absolute change >2 fold, corrected P-value<0.01). Validation of microarray data on six selected genes by RT-qPCR revealed an acceptable correlation between transcription patterns obtained by the two techniques (mean R2 = 0.92) except for CYP6Z9 for which transcription pattern among comparisons was not confirmed (Table S1). Between the comparisons “NS2 Exp vs P” and “NS2 Unexp vs P”, 2088 transcripts were commonly found differentially regulated, including 962 up- and 1126 down-regulated transcripts (Figure 2). Among these up-regulated transcripts, GO term Enrichment analysis revealed 8 GO terms over represented compared to the whole microarray (FDR<0.05), all linked with P450 activities (Figure 3a). Within the 962 up regulated transcripts found in both comparisons (Table S2), 42 CYPs were detected, 18 of which belong to the CYP9J family (Table S3).
Larvae from NS2 are resistant to both temephos and permethrin. In an attempt to prioritise genes putatively involved in temephos resistance we applied an additional layer of filtering to derive our candidate gene list. We specifically looked for genes whose fold change compared to the susceptible P population were higher in the NS2 surviving temephos exposure than in the unexposed NS2 vs P comparison. By using an arbitrary ratio threshold of 1.25, the candidate list was reduced to 122 transcripts (Table S4). This threshold was chosen in order to be within the range of differential detection of the microarray technology (in line with recommendations from Agilent Techonologies). These candidates are highlighted in the volcano plot (Figure 4) which also shows all transcripts significantly upregulated in Nakhon Sawan Unexp compared to Phatthalung. Interestingly, among the most overtranscribed genes figured one carboxylesterase CCEae3a (AAEL005112) which was overtranscribed around 60 fold in Nakhon Sawan Unexp compared to Phatthalung and 91 fold in Nakhon Sawan Exp compared to Phatthalung (ratio RS/RC = 1.33). Two other esterases were also found more upregulated in RS comparison compared to CS: CCEae6A (AAEL015264-RA) (29 fold upregulated in RS, 22 fold in CS) and CCEglt1K AAEL006097-RA (4.2 fold in RS, 2.5 fold in CS). Four cytochrome P450s were also present in the candidate genes list: CYP6Z8 (AAEL009131-RA), CYP9M9 (AAEL001807-RA), CYP6AH1 (AAEL007473-RA) and CYP4H28 (AAEL003380-RA). Multiple transcripts coding for cuticular proteins were also found significantly overtranscribed among the 122 transcripts, including 6 paralogous genes belonging to the CPLC group.
Quantitative PCR showed a significantly higher CCEae3a gene copy number in NS 2 unexposed (>165 fold, Pval<0.01) and NS2 Exposed (>350 fold, Pval<0.01) compared to Phatthalung strain (Figure S2).
Sequencing of the cDNA sequence of CCEae3a (AAEL005112-RA) revealed the presence of non synonymous mutations between the sequences from Vectorbase, Phatthalung and the resistant population NS2. The derived amino acid sequence of NS2 had amino acid substitutions AAT positions 373 (GAA to GAC, leading to the change of an aspartic acid to glutamic acid), 374 (AAT to GAT, asparagine to glutamic acid), 538 (CGA to CAA, arginine to glutamine) and 541 (GAA to GAC, glutamic acid to aspartic acid) compared to Vectorbase and Phatthalung sequences (Figure 5).
The models for Nakhon Sawan, Phatthalung, Vectorbase and mutated Vectorbase (Vectorbase sequence with the NS mutations at the positions 373, 374, 538 and 541) sequences were generated using PHYRE2 web server in the intensive mode. For all of them, 99% of the residues were modelled at more than 90% confidence in the final model and the best ranked match was the carboxylesterase αE7 from the Australian sheep blowfly Lucilia cuprina (LcαE7) with 34% identity. The in silico models enabled the polymorphic residues of the analysed variants (E373D, N374D, R538Q and E541D) to be localised and to identify those residues involved in the active site by homology with LcαE7. The most interesting difference between resistant and susceptible forms was found more than 20 Å away from the polymorphic residues and involved residues that belong to the putative substrate-binding site (Y283-G293) (Figure 6).
Previous studies have reported temephos resistant populations of Ae aegypti from Thailand [8], [19]. The objective of the current work was to identify the mechanism(s) responsible for this resistance. Bioassays were conducted on four populations of Thai mosquitoes and a susceptible laboratory population. Larvae from the P population from the southern Phatthalung province were fully susceptible to temephos with a lower LC50 than the New Orleans laboratory strain. Full susceptibility to temephos was also reported in the neighbouring province of Songkhla in 2005 [19]. Two other populations, NS1 and CM, showed low levels of temephos resistance (according to classifications in [31]) and one population, NS2, from central Thailand, showed medium levels of resistance, with RR from 6–10 fold. An earlier study also found the highest levels of temephos resistance in the Nakhon Sawan province [19] and the RRs obtained in the current study are similar to those reported from this province in a 2005 study, despite the use of different lab susceptible populations [8]. Although the current study did not directly assess the impact of the observed resistance on the field efficacy of temephos, earlier studies in Brazil clearly demonstrated an impact of resistance levels of similar magnitudes to the NS2 population on the duration of temephos efficacy in simulated field assays [32]. Hence it is likely that temephos resistance is compromising dengue control in central Thailand but, as noted by others [19], insecticide resistance in Ae aegypti appears to be very focal (note the marked differences in the Temephos LC50 between NS1 and NS2, separated by a distance of 60 Kms).
Permethrin resistance was also detected in Ae aegypti larvae from Chiang Mai and from both populations from Nakhon Sawan province. Again this agrees with earlier bioassays data from Thailand [8]. Pyrethroids are not directly applied as larvicides in Thailand but contamination of breeding sites may occur by the use of pyrethroids as aerial sprays to control dengue epidemics. Alternatively, the co-occurrence of both temephos and permethrin resistance in the same population may be caused by cross-resistance as was proposed following a temephos selection experiment in Cuba [33]. Possible mechanisms for this putative cross resistance are discussed below.
The data from enzyme inhibitors suggests that temephos resistance in the Nakhon Sawan province is linked to carboxylesterase activities. Conversely, the cytochrome P450 inhibitor, PBO, had the biggest impact on permethrin resistance in the NS2 population. However, even after addition of PBO, NS2 remained moderately resistant to permethrin suggesting that pyrethroid target site resistance may be present in the population: two sodium channel mutations associated with permethrin resistance, V1016G and F1534C are known to be widespread in Thailand [34], [35], [36], [37].
Further support for a key role for carboxylesterases in conferring temephos resistance is provided by biochemical assays using alpha- and beta-naphtylacetate as substrates. Significantly higher levels of esterase activity were detected in the NS2 population compared to the susceptible population from Southern Thailand (P). Again, this mimics findings from other temephos resistant populations [14], [15]. Although both changes in gene expression and allelic variation in individual CCE proteins has been associated with organophosphate resistance [38], [39] the latter is typically associated with a decrease in esterase activity, as measured with general esterase substrates [40], [41], [42]. We therefore hypothesised that one or more up-regulated carboxylesterase genes were responsible for the temephos resistance and thus used a microarray platform to identify transcripts that were upregulated in the resistant NS2 population compared to the susceptible Phatthalung population.
Phatthalung was used as a susceptible population, as opposed to a standard laboratory susceptible population, in an attempt to reduce the impact of extended laboratory colonisation and geographical differences on the transciptome data. It was therefore surprising to find over 2000 transcripts significantly differentially transcribed between the two Thai populations. In a three way comparison we compared both NS2 unexposed to insecticides and a subset of NS2 population that had survived temephos exposure and been sacrificed 24 hours after insecticide exposure with the Thai susceptible population. We did not observe any significant differences between the NS2 exposed and unexposed populations but we used these three data sets to filter our candidate list in two steps. Firstly we discarded genes that were only upregulated in the NS2 population in one of the comparisons (Figure 2) focusing initially on the subset of 962 transcripts that were commonly upregulated in the NS2 exposed vs P and the NS2 unexposed vs P. Interestingly, this subset of transcripts contained a large number of cytochrome P450 genes. This was confirmed by the enrichment analysis which showed a clear enrichment of GO terms linked with P450 activities in the overtranscribed genes compared to the whole microarray. Over half of the upregulated P450s belonged to the CYP9J family (Table S3). CYP9Js have been widely implicated in pyrethroid resistance in Ae aegypti populations across the globe [27], [43], [44], and several of these have been biochemically characterized and been shown to metabolize pyrethroids [45]. Further confirmation of the role of this P450 family in pyrethroid resistance comes from transgenic expression of CYP9J28 in Drosophila melanogaster which conferred an elevated level of resistance to pyrethroids [46].
To further refine our list of candidate genes responsible for temephos resistance, we hypothesised that genes putatively conferring this phenotype would exhibit a higher fold change differential in transcript levels in the NS2 exposed versus susceptible comparison than the NS2 unexposed vs susceptible. We therefore reduced our candidate list from 962 to 122 transcripts by dividing the fold changes in “NS2 Exp vs P” comparison by fold changes in “NS2 Unexp vs P” comparison and using an arbitrary cut off of >1.25. Only four cytochrome P450s remained in this refined candidate list (CYP6Z8, CYP9M9, CYP6AH1, CYP4H28), none of which belonged to the CYP9J family, perhaps indicating that the over expression of the CYP9J genes in NS2 contributes to the permethrin resistance phenotype but has a negligible role in conferring temephos resistance. CYP6Z8 has recently been shown to metabolize the 3-phenoxybenzoic alcohol (PBAlc) and 3-phenoxybenzaldehyde (PBAld), common metabolites produced by carboxylesterases [47], and it is possible that elevated levels of this enzyme is an important secondary resistance mechanism.
Three carboxylesterase genes were present within final candidate list. One of these (AAEL006097-RA) encodes a putative glutactin which, although potentially catalytically active as it contains the catalytic triad and oxyanion hole, is not thought to be involved in xenobiotic detoxification. The two remaining carboxylesterases (CCEae3a (AAEL005112) and CCEae6A (AAEL015264)) belong to the alpha esterase clade, a group typically associated with dietary or xenobiotic detoxification functions. CCEae3a was overtranscribed more than 90 fold in NS2 exposed compared to Phatthalung and more than 60 fold in NS2 unexposed compared to P. To verify that this did not simply reflect an exceptionally low level expression in the southern Thai population, we also included the lab susceptible New Orleans in the qPCR. There was no significant difference in the expression of CCEae3A in the two susceptible populations (Table S1). CCEae6A was also highly over expressed in NS2 compared to the P population (29 fold in exposed, 22 fold in unexposed). Of these two alpha esterases, CCEae3a appears a particularly strong candidate for temephos resistance, as this gene is known to be overexpressed in temephos resistant populations from Martinique [16], [48] and Brazil [17]. Interestingly, the copy number of CCEae3a was much higher in the NS2 resistant strain than the susceptible P strain, and also elevated in the subset of the NS2 strain surviving temephos exposure compared to the general NS2 population. This suggests that the overtranscription of CCEae3a may at least be partly due to gene amplification, similar to the mechanism observed in Culex pipiens [11].
In Martinique Island, both CYP6Z8 and CCEae3a were found upregulated together in pyrethroid and organophosphate resistant populations of Aedes aegypti [16], [48] supporting the possible coordinated role of CYP6Z8 and CCEae3a in insecticide detoxification [47].
In addition to the over expression of CCEae3a cDNA sequence, several non-synonymous mutations were found between the sequences from Phatthalung compared to Nakon Sawan 2. In silico structure predictions of CCEae3a, based on the carboxylesterase αE7 from the Australian sheep blowfly Lucilia cuprina (LcαE7) [49] predicted that the polymorphic residues were not adjacent to the insecticide binding site. Nevertheless, the resistant variants lacked the hairpin loop between Y283 and G293 which was found in the susceptible population. It is possible that this loop displaces the F286 residue (homolog to F309 in LcαE7) that seems to be essential in stabilizing OPs in the LcαE7 active site. Further work is needed however to determine whether the allelic variants differ in their enzymatic activity and if either or both forms are capable of sequestering and/or metabolising temephos.
Temephos is one of the key insecticides for dengue control across the tropics but operationally significant levels of resistance are being increasingly reported [18]. Carboxylesterases have long been suspected to play a key role in mediating this resistance but to date no clear candidates had been identified. The identification of strong candidate genes has now laid the foundations for the development of molecular diagnostics to assess the correlation between the overexpression of these genes and temephos resistance across the distribution of Ae aegypti.
|
10.1371/journal.ppat.1006871 | Liver macrophage-associated inflammation correlates with SIV burden and is substantially reduced following cART | Liver disease is a leading contributor to morbidity and mortality during HIV infection, despite the use of combination antiretroviral therapy (cART). The precise mechanisms of liver disease during HIV infection are poorly understood partially due to the difficulty in obtaining human liver samples as well as the presence of confounding factors (e.g. hepatitis co-infection, alcohol use). Utilizing the simian immunodeficiency virus (SIV) macaque model, a controlled study was conducted to evaluate the factors associated with liver inflammation and the impact of cART. We observed an increase in hepatic macrophages during untreated SIV infection that was associated with a number of inflammatory and fibrosis mediators (TNFα, CCL3, TGFβ). Moreover, an upregulation in the macrophage chemoattractant factor CCL2 was detected in the livers of SIV-infected macaques that coincided with an increase in the number of activated CD16+ monocyte/macrophages and T cells expressing the cognate receptor CCR2. Expression of Mac387 on monocyte/macrophages further indicated that these cells recently migrated to the liver. The hepatic macrophage and T cell levels strongly correlated with liver SIV DNA levels, and were not associated with the levels of 16S bacterial DNA. Utilizing in situ hybridization, SIV-infected cells were found primarily within portal triads, and were identified as T cells. Microarray analysis identified a strong antiviral transcriptomic signature in the liver during SIV infection. In contrast, macaques treated with cART exhibited lower levels of liver macrophages and had a substantial, but not complete, reduction in their inflammatory profile. In addition, residual SIV DNA and bacteria 16S DNA were detected in the livers during cART, implicating the liver as a site on-going immune activation during antiretroviral therapy. These findings provide mechanistic insights regarding how SIV infection promotes liver inflammation through macrophage recruitment, with implications for in HIV-infected individuals.
| Liver disease is common in HIV-infected individuals and is one of the leading causes of death in this population. Currently, the factors that contribute to liver disease during HIV infection are not known, as human studies are difficult to conduct. Therefore, pathogenic SIV infection of macaques is a valuable model system for understanding immune changes that occur in tissues during infection, including the liver. Using liver samples from uninfected and SIV-infected macaques, we observed an increase in liver macrophage infiltration that was likely mediated by the macrophage-attracting chemokine, CCL2. Importantly, liver macrophage number strongly correlated with liver SIV levels and the expression of liver inflammatory and fibrosis mediators during SIV infection. Further, we identified an upregulation of immune and inflammatory pathways in the liver of SIV-infected macaques, including a strong antiviral response. Treatment with antiretroviral drugs decreased macrophage infiltration and reduced liver inflammation, consistent with reduced SIV levels. In summary, these findings provide mechanistic insights into how SIV infection promotes liver inflammation through macrophage infiltration, which can predispose this population to liver complications during infection.
| Liver disease has become a leading contributor to morbidity and mortality in HIV-infected people with the occurrence of nonalcoholic fatty liver disease (NAFLD) being one of the most predominant complications [1–5]. Within the setting of HIV infection, not only is NAFLD more common than in the general population, but also is more severe with higher incidence of steatohepatitis (NASH), liver injury, and lobular inflammation when compared to HIV-negative people [4, 6]. Liver inflammation can directly impact systemic circulation through the production of acute-phase proteins and failure to detoxify gut-derived blood to potentially contribute to the chronic inflammatory conditions commonly observed in HIV-infected people [7–9]. Further, the liver plays a role in the metabolism of antiretroviral drugs, and in some studies NAFLD and fibrosis are exacerbated during combination antiretroviral treatment (cART), and is most often associated with low CD4 T cell counts, alcohol abuse, high immune activation, and exposure to certain ART drugs, such as didanosine [4, 10, 11]. Other studies have identified improvement of liver fibrosis during drug treatment, associated with better recovery of CD4 T cells during therapy and younger age [12]. Liver biopsies are currently considered to be the most precise method for diagnosing hepatic disease, however, this invasive procedure can have complications, such as pain, hemorrhage and sepsis, which make human studies difficult to conduct [13, 14].
Experimental simian immunodeficiency virus (SIV) infection of macaques has provided useful models for studying viral pathogenesis and the host immune response, particularly in tissue compartments. Pathogenic SIV infection of macaques has provided key insights in the pathogenesis of HIV/SIV infection, including identifying the gut as a major site of viral replication and CD4 T cell depletion [15, 16], elucidating the mechanisms of mucosal dysfunction [17–20], and defining viral reservoirs [21–25]. In addition, SIV-macaque models have been utilized to establish the liver as a primary site of SIV clearance in vivo [26] and to evaluate liver immune cells during SIV infection, including macrophages, T cells, and NK cells [27–29]. Indeed, one study found that some of the T cells that infiltrate the liver during infection are SIV-specific CD8 cells that localize to the portal triad regions of the liver [30], the area where the portal vein, hepatic artery and bile duct converge. The portal vein, which drains the gastrointestinal tract, gall bladder, pancreas and spleen, is of particular interest as in contrast to other veins, it does not conduct blood back to the heart, but rather supplies the hepatic capillary beds, bringing nutrients and ingested toxins to the liver for processing. This unusual circulatory pattern also provides an opportunity for microbial products from the gastrointestinal tract to enter the liver, especially when intestinal epithelial integrity is compromised. Liver dysfunction leads to incomplete clearance of bacterial products from the blood, and increased presence of translocated microbial products in systemic circulation, which correlates with immune activation/inflammation in SIV-infected macaques [31, 32].
The exact mechanisms of liver disease pathogenesis during HIV/SIV infection are yet to be fully defined. It was previously discovered that CXCL16 production induces NK cell infiltration into the liver during SIV infection [29] supporting a role for chemokines in promoting liver disease during infection. Indeed, chemokine-associated immune cell infiltration has been implicated in a spectrum of liver diseases, including viral hepatitis, fibrosis, and alcoholic liver disease [33]. In particular, the CCL2-CCR2 axis is critical for the progression of acetaminophen-induced hepatotoxicity [34], fibrosis [35], and steatohepatitis [36, 37] through the recruitment of hepatic macrophages. Disruption of this chemokine network reduces both hepatic macrophage number and associated liver pathologies [36, 38, 39] highlighting the central role that infiltrating hepatic macrophages can play in liver disease.
The goal of this study was to delineate the immunologic and inflammatory factors that contribute to liver disease progression during retroviral infection. As liver disease develops over many decades, this cross-sectional study focused on the early mediators that trigger subsequent hepatic dysfunction, and included untreated adult and infant macaques that were infected with SIV, as well as macaques that were SIV-infected and receiving cART. We report that SIV levels in the liver are associated with macrophage infiltration and hepatic T cell numbers. Transcriptomic analysis revealed an inflammatory signature in the livers of SIV-infected macaques that was dominated by a strong antiviral response that was diminished in the livers from the cART treated macaques. These findings provide critical mechanistic insights regarding how SIV infection impacts liver inflammation and viral replication, which will be important for designing therapies to ameliorate liver complications during HIV infection.
During the development of liver disease, the presence of different immune cell subsets provides information regarding the immune cell populations that drive liver inflammation while the intrahepatic localization of these immune cells determines the nature of disease [27, 28]. Utilizing immunofluorescence microscopy, macrophages and T cells were quantified in the liver using cell-specific CD68 and CD3 staining, respectively (Fig 1A). Although elevated numbers of macrophages and T cells, both CD4 and CD8 T cells, were observed in the portal regions during SIV infection (S1 Fig), there was noticeable variation in the levels of immune cells observed within each liver section of infected macaques. Therefore, these quantitative analyses focused on lobular regions of the liver without any portal triads. In SIV-infected infant macaques, a significant increase in T cells was observed compared to uninfected infants. A trend for increased T cells in SIV-infected adult macaques was also observed, although this difference did not reach statistical significance (Fig 1B). Likewise, the number of liver macrophages was also significantly increased in both infants and adults during SIV-infection (Fig 1C). In macaques that were receiving cART, the levels of both liver T cells and macrophages were reduced to levels comparable to uninfected macaques suggesting that this immune cell infiltration was reversible when SIV replication is suppressed.
Hepatic macrophage accumulation, and to a lesser extent infiltration of activated T cells, is largely driven by the CCL2-CCR2 chemokine/receptor axis, whereby CCL2 recruits CCR2-expressing monocytes/macrophages that differentiate into tissue macrophages with the potential to initiate macrophage-mediated inflammation [34–37][40]. Luminex assessment of plasma inflammatory factors revealed a significant increase in circulating levels of CCL2 in the blood of SIV-infected macaques compared to uninfected macaques (67 pg/mL in SIV-infected vs. 16 pg/mL in uninfected, p = 0.0158) (Fig 2A). To evaluate if CCL2 could be influencing T cell or monocyte/macrophage infiltration into the liver, the expression of CCL2 and its receptor, CCR2, were examined by qRT-PCR. We observed an intrahepatic upregulation of both CCL2 and CCR2 transcripts in SIV-infected untreated macaques (Fig 2B and 2C). Importantly, CCR2 expression levels correlate with hepatic CD68+ macrophage levels, providing evidence that CCL2/CCR2 axis contributes to the macrophage infiltration that was observed (Fig 2D). Although CCL2 levels did not directly correlate with macrophage levels, we did observe that in each treatment group those macaques with the highest CCR2 levels also tended to have high CCL2 expression (Fig 2B and 2C, individual macaques denoted with 1, 2, 3). To determine the location of the CCL2-producing cells within the liver, SIV-infected macaque livers were evaluated by immunohistochemistry. Diffuse CCL2 staining was observed throughout the liver with small foci of CCL2-producing cells in portal triad regions (Fig 2E).
To evaluate monocyte/macrophage recruitment in the liver during SIV infection, liver sections were evaluated for recently infiltrated monocytes/macrophages by staining for Mac387. The Mac387 protein recognizes the calcium-binding proteins MRP8 and MRP14 that are restricted to early stage monocyte differentiation and elevated in inflamed tissue, but are absent in mature tissue macrophages [41–43]. Additionally Mac387+ monocyte/macrophages potently migrate toward CCL2 gradients [44]. Elevated levels of Mac387-postive cells that were found localized around portal triads in the livers of SIV-infected macaques (Fig 2F). Furthermore, these Mac387+ monocyte/macrophages were markedly reduced in antiretroviral treated macaques (Fig 2F).
Upon pathogen sensing, many cell types in the liver have been implicated in the production of CCL2, including hepatocytes, hepatic stellate cells, sinusoidal endothelial cells and hepatic macrophages, with hepatic macrophages being the primary source of CCL2 [45]. Therefore, to mechanistically delineate the contribution of macrophages to CCL2 production during viral infection, monocyte-derived macrophages were stimulated with purified viral PAMPS, poly I:C, a TLR3 agonist, and ssRNA40, which is derived from the HIV-1 long terminal repeat and activates TLR7/8. These in vitro studies determined that viral PAMPs have the capacity to upregulate CCL2 expression in monocyte-derived macrophages with TLR7/8 activation using ssRNA40 producing higher CCL2 expression than poly I:C.
To characterize the influx of CCR2-expressing cells into the liver during infection, we assessed the phenotype of T cells and monocytes/macrophages in uninfected (n = 9) and SIV-infected (n = 9) macaques using flow cytometry. Both peripheral blood mononuclear cells (PBMC) and paired liver cell suspensions were first gated on singlets, live cells, and CD45 expression followed by CD3-positive T cells and the proportion of T cells expressing CCR2 (Fig 3A). Cells of monocyte/macrophage lineage were identified by CD14 expression followed by phenotyping using CD16 and CCR2. CCR2-expressing monocytes/macrophages were further characterized based on Mac387 and CD16 expression (Fig 3A). With regards to circulating PBMCs that have the capacity to migrate to CCL2 gradients in tissue, CCR2 was expressed on the majority of monocytes/macrophages while very few T cells expressed CCR2 in the blood (Fig 3B). However, when compared to blood, more T cells in the liver expressed CCR2 in both uninfected and SIV-infected macaques (Fig 3C) likely due to the fact that CCR2 is mostly expressed on activated T cells [46]. Additionally, five of the nine SIV-infected macaques displayed higher levels of CCR2-positive T cells in the liver, however this trend did not reach statistical significance (p = 0.077) (Fig 3C).
With regards to monocytes/macrophages, there was an observed expansion of inflammatory CD16+ monocytes/macrophages in the liver during SIV infection (Fig 3D). A similar trend for increased CD16+ monocytes/macrophages was detected in the blood, however this did not reach statistical significance (p = 0.1615). When considering CCR2+ monocyte/macrophages, there was an increase in the frequency of CCR2-positive monocyte/macrophages that recently infiltrated the liver based on expression of Mac387 (Fig 3E), providing evidence for active recruitment of monocyte/macrophages along the CCR2-CCL2 axis. In addition, there was a highly significant increase in CCR2-monocytes/macrophages that express the inflammatory marker, CD16, again providing evidence for CCR2-associated expansion of inflammatory CD16+ monocyte/macrophages in the livers of SIV-infected macaques. Moreover, these inflammatory monocytes/macrophages correlated with the number of CD68+ macrophages (Fig 3F).
CD14 expression has been historically used to define cells of monocyte/macrophage lineage, however, tissue macrophages display considerable heterogeneity between different species and even in different tissue compartments. For example, CD14-negative tissue macrophages have been described in various tissues, including the liver, the gut, and the spleen [47, 48]. While the above analysis gated on CD14 to assess monocyte/macrophage populations in the liver (Fig 3), this gating strategy may not capture all mature tissue macrophages. CD68, on the other hand, is a well-characterized pan-macrophage marker that has been used to define tissue macrophages in macaques and in humans [49, 50]. Therefore, defining mature tissue macrophages as CD68+CD11b+, a previously defined phenotype of tissue macrophages in macaques [47], allowed us to interrogate the phenotype of mature hepatic macrophages during SIV infection. With regards to function, we observed that hepatic CD68+ cells were the predominant cell type involved in the phagocytosis of labeled E. coli, particularly those cells with higher forward scatter (Fig 4A). Gating first on single, live, CD45+, CD3- cells, mature CD68+CD11b+ macrophages were assessed for the expression of CD14 and CD16 (Fig 4B). Comparable to the assessment of CD68+ macrophages by immunofluorescence microscopy (Fig 1), our flow cytometry analysis also indicates a significant increase in the levels of mature macrophages in the liver during SIV infection (Fig 4C). Interestingly, our flow cytometry analyses revealed variable expression of CD14 on mature CD68+ macrophages with SIV-infected macaques having more CD14+CD68+ macrophages than uninfected macaques (66.9% vs 50.6% of macrophages expressing CD14). A similar trend was observed regarding CD16 expression during infection (63.3% vs 45.4% of macrophages expressing CD16) indicating that these mature macrophages display an altered phenotype during SIV infection with a higher percentage expressing CD14 and CD16 (Fig 4D).
The detection of microbial products within the blood during HIV and SIV infections is well established [31, 32, 51]. To determine if elevated levels of liver macrophages were associated with systemic bacterial translocation, the concentration of plasma LPS-binding protein (LBP) was assessed. A significant increase was observed in circulating levels of LBP in SIV-infected as well as in SIV-infected cART-treated macaques (Fig 5A). The presence of elevated LBP levels suggests that more bacteria are translocating through the liver, as the liver plays an important physiological role in the detoxification of microbial products that enter via the portal vein. Therefore, liver bacteria levels were quantified using the 16S rRNA gene by qPCR. Bacterial DNA was detected in all of the liver samples, indicating that uninfected macaques normally have detectable levels of bacterial DNA (Fig 5B). Interestingly, uninfected infant macaques were found to have elevated levels of bacterial DNA within their livers compared to uninfected adults (0.35 ng vs 0.16 ng 16S DNA/100ng total DNA), a difference that was statistically significant (p = 0.0159) (Fig 5B). Assessing the adult and infant macaques together determined that SIV-infected cART-treated macaques had significantly elevated levels of liver 16S DNA when compared to uninfected (p = 0.0006) and SIV-infected macaques (p = 0.0219) (Fig 5B). The lack of a significant difference in the levels of 16S DNA within SIV-infected macaques when compared to uninfected macaques was due to the higher levels of 16S DNA in the uninfected infants (Fig 5B), as significantly elevated levels of liver 16S DNA was observed when adult macaques were evaluated alone (Fig 5C). The elevation in liver 16S DNA levels in the SIV-infected cART-treated macaques was unexpected, but nevertheless, this elevation of bacterial DNA in the livers of the SIV-infected and SIV-infected cART-treated macaques did not correlate with macrophage levels (Fig 5D), indicating that the macrophage expansion observed during infection is likely not mediated by bacterial translocation to the liver.
To assess the levels of SIV DNA in the liver and to provide insights regarding the role of virus in driving macrophage infiltration, SIV DNA was quantified by quantitative hybrid real-time/digital PCR [52, 53]. We observed a wide range of SIV levels (per 106 cell equivalents) in the livers of SIV-infected macaques with no distinct differences associated with age (Fig 6A). As expected, macaques treated with cART had the lowest levels of SIV with four of the macaques having levels below the limit of detection for the assay (<10 SIV DNA/106 cell equivalents) (Fig 6A). A correlation was observed between the level of SIV DNA in the liver and plasma SIV RNA and the number of hepatic T cells, which were quantified by immunofluorescence microscopy (Fig 1)(Fig 6B). Further analysis identified an even stronger correlation between the levels of SIV DNA hepatic macrophage levels (r = 0.8099, p<0.0001) (Fig 6C). This correlation with liver macrophage number was also observed when plasma SIV RNA levels were assessed (p<0.0001, Fig 6C). These findings demonstrate a direct relationship between SIV load in liver and the levels of hepatic T cells and macrophages implicating SIV burden as a key stimulus in recruitment of immune cells into the liver.
Structurally composed of lobules, the liver is a multifaceted organ with distinct tissue structures that represent unique environments for SIV replication. The high levels of SIV DNA in the liver and the observed correlations with both macrophages and T cells raised questions with regard to where in the liver SIV replication was occurring and which cells, macrophages or T cells, were producing virus. Therefore, utilizing highly sensitive RNAscope in situ hybridization technology [21, 54, 55], we assessed the location and phenotype of SIV RNA-positive cells in the livers of SIV-infected untreated macaques. We identified SIV RNA-positive cells predominately localized to the portal triad regions of the liver with lower levels of SIV-positive cells located in the lobular tissue (Fig 7). SIV RNA-positive cells that were found in lobular regions were typically a single, individual cell, not large aggregates of SIV-positive cells like those observed in the portal regions. To identify the cellular subset associated with SIV replication in the liver, RNAscope was used in combination with antibody staining for CD3 and CD68 to identify T cells and macrophages, respectively. We observed that most SIV RNA was associated with T cells (Fig 8A). Occasionally, SIV RNA-positive signal co-localized with CD68+ macrophages (Fig 8B), however, this was quite rare and often there was diffuse CD3 staining in the same area suggesting this may be a macrophage that has acquired SIV RNA by T cell phagocytosis as has been demonstrated in previous studies [56]. Although SIV DNA was detected in the liver during cART, we were unable to detect SIV RNA-positive cells in the livers of macaques treated with antiretroviral drugs by RNAscope.
To gain further insights into the mechanisms of SIV-induced liver inflammation, the hepatic expression of pro-inflammatory (CCL3, TNFα), pro-fibrosis (TGFβ), and anti-inflammatory (IL-10) mediators were evaluated by qRT-PCR. There was a significant increase in both CCL3 and TNFα in the livers of SIV-infected macaques when compared to uninfected animals (Fig 9A and 9B). A similar upregulation was observed in the pro-fibrosis cytokine, TGFβ, in macaques that were SIV-infected (Fig 9C). Following cART, both CCL3 and TGFβ were significantly reduced in the liver while TNFα remained elevated, particularly in infant macaques (Fig 9A, 9B and 9C). In contrast, assessment of the inflammation-suppressive cytokine, IL-10, showed no differences between treatment groups when both adults and infants were evaluated together (S2 Fig). Interestingly, infant macaques had higher expression of IL-10 transcripts in the liver (S2 Fig) and increased circulating levels of IL-10 in the blood (S2 Fig) when compared to adults, likely reflecting inherent differences in immune system maturity between the age groups.
Correlation analyses demonstrated a significant, positive correlation between hepatic macrophage levels and the expression of liver inflammatory mediators (TNFα, CCL3), consistent with a role for liver macrophages as inflammatory contributors (Fig 9A and 9B, lower panels). Similarly, TGFβ levels also correlated with hepatic macrophage number suggesting that infiltrating macrophages in the SIV-infected macaques may trigger the early events that may lead to liver pro-fibrosis responses (Fig 9C, lower panel).
To delineate the role that monocyte-derived macrophages may have in driving the expression of inflammatory and pro-fibrosis mediators during infection, in vitro generated human monocyte-derived macrophages were stimulated with viral-associated PAMPs. We observed that monocyte-derived macrophages were able to upregulate both inflammatory mediators, CCL3 and TNFα (Fig 9D). This upregulation was observed with ssRNA40, which is derived from HIV and activates a main pathogen-recognition receptor for HIV, TLR7/8. Interestingly, TGFβ was significantly downregulated by viral PAMP stimulation using ssRNA40 (Fig 9D) suggesting that macrophages that infiltrate the liver during infection may not be the producers of TGFβ.
To elucidate the global impact of SIV infection and cART treatment on overall liver homeostasis, the hepatic transcriptomic signature was characterized during chronic SIV infection by microarray expression analysis. Differentially expressed genes were determined by normalizing SIV-infected and SIV-infected cART macaques to age-matched uninfected controls. Overall, liver transcriptional changes in SIV-infected macaques included a number of different functional pathways with the most significant changes occurring in immune and inflammation-associated genes, which are depicted in the functional modules outlined in Fig 10. The SIV-infected adult macaques exhibited an increase in the first four modules, and a down- modulation in the last two pathways as indicted in the heatmap (Fig 10, lane 1). Some of the interesting observations are the upregulation of T cell signaling and NF-kb activation (module 1) and genes associated with communication between innate and adaptive immune systems (module 2), both of which continue to be upregulated, although to a lesser extent, in macaques that were administered cART (Fig 10, lane 2). A strong upregulation in Rig-I-like receptor (RLR) signaling/interferon response (module 4) was also evident in the untreated SIV-infected adult macaques with many of the top upregulated genes being involved in antiviral defense (e.g. ISG15, MX1) (Table 1). Finally, an interesting down-regulation was observed in IL-1 mediated inhibition of retinoid X receptor (RXR) function (module 5) as RXR activity plays a role in metabolic disorders and in the regulation of several macrophage functions, including production of chemokines, pathogen sensing, and macrophage lipid metabolism [57].
In contrast to adults, SIV infection of infant macaques resulted in an upregulation of only two pathways, cytotoxic T lymphocyte-mediated apoptosis (module 3) and RLR/interferon signaling (module 4) (Fig 10, lane 3). Many of the antiviral RLR/interferon genes that were found upregulated in adults were also upregulated in SIV-infected infant macaques (bolded genes, Table 2). Since the RLR signaling/interferon signature was found in both the adults and infants during SIV infection, the specific genes involved in this pathway were examined by pathway analysis. During viral infections, such as SIV, genes in the RLR pathway code for proteins that are involved in sensing cytosolic RNA viruses and/or viral PAMPs, the production of interferons (e.g. IFNα/β) and the downstream activation of interferon-stimulated genes (ISGs) [58]. The pathway analysis revealed the expression of several interferon-induced signaling molecules (e.g. Stat1, IRF9) and downstream interferon stimulated genes (ISGs) (e.g. Mx1, OAS1) was altered in SIV-infected infants (S3 Fig). This was further confirmed through network analysis where many antiviral response genes were identified to be strongly upregulated in the liver, including IRF7, IRF9, ISG15, and Mx2 (S3 Fig). Interestingly, CCL3, one of the pro-inflammatory mediators assessed by qRT-PCR that correlated with liver macrophage number, was also identified in this network analysis and predicted to be part of the host antiviral response (S3 Fig). The pathway analyses for SIV-infected adult macaques, which contained slightly more variation, identified similar gene activity, but fewer genes met the cut-off criteria (p < 0.05, fold change >1.5) (S4 Fig) compared to the infant macaques.
In SIV-infected macaques receiving cART, the liver antiviral signature was reduced, but not completely, especially in infant macaques (Fig 10, module 4, lanes 2 and 4). It could be that residual virus or elevated bacteria levels, which were detected in the liver during cART, may drive some persistent liver inflammation during treatment. As the antiviral profile decreased during cART, we observed that many of the top upregulated genes in both adult and infant macaques are involved in metabolism, cellular homeostasis and the oxidation-reduction process (e.g. SLC22A4, Akr1b7) (Tables 3 and 4). Collectively, this suggests the liver continues to experience perturbation of function during drug therapy, and could explain some of the fatty liver complications observed in HIV-infected individuals.
Liver disease has emerged as one of the most common non-AIDS-related causes of death in those infected with HIV, accounting for 14%-18% of all deaths [59]. Here, a pathogenic SIV-macaque model was utilized to provide insights into liver disease pathogenesis during SIV infection and the impact of antiretroviral treatment. Our findings identify a correlation between the levels of CCL2-CCR2 and immune cell trafficking to the liver during SIV infection, including monocytes/macrophages and T cells. Infiltration of these immune cells during untreated SIV infection is likely driven by SIV burden as correlations between immune cell number, both macrophages and T cells, and SIV levels were observed, in addition to a strong antiviral transcriptiomic signature detected in the microarray analyses. These findings provide mechanistic insights into how infiltration of monocytes/macrophages into the liver can induce tissue inflammation during SIV infection and also suggest that viral suppression during cART can greatly reduce SIV-associated liver inflammation, although these levels are still elevated compared to uninfected macaques.
Expansion of the liver macrophage population has been implicated in multiple forms of liver disease, including acute liver failure, HCV/HBV infection, NAFLD, fibrosis, and alcoholic liver disease [40, 60]. Moreover, studies elucidating the role of monocytes/macrophages in SIV pathogenesis have indicated that monocyte/macrophage accumulation correlates with SIV disease severity and progression, tissue damage in the lung and gut, and SIV-associated encephalitis [61–63]. Importantly, inhibiting macrophage accumulation has shown therapeutic benefits in models of both liver disease and SIV infection [36, 38, 39, 64, 65]. In our study, we observed an increase in hepatic monocyte/macrophages during SIV infection that correlated with both inflammatory (TNFα, CCL3) and fibrosis (TGFβ) mediators suggesting multiple, and maybe even opposing, roles during infection. Interestingly, our experiments utilizing monocyte-derived macrophages, reveals that these monocyte-derived macrophages are potent producers of inflammatory mediators, both TNFα and CCL3, but not TGFβ, at least in the in vitro environment. Therefore, we hypothesize the production of TGFβ in vivo may represent a compensatory mechanism to limit liver inflammation induced by infiltrating monocytes/macrophages and that other cell types in the liver may also contribute to TGFβ production, including sinusoidal endothelial cells, resident Kupffer cells, intrahepatic lymphocytes and dendritic cells [66]. It was interesting that TGFβ transcript levels decreased in the macaques that were treated with antiretroviral drugs when compared to untreated SIV-infected macaques given that some human studies report progression of liver fibrosis during cART. However, many of these studies found liver fibrosis to be associated with other factors during drug therapy, including poor viral control, low CD4 T cell counts, presence of HCV infection, older age, and alcohol abuse [10, 12, 67, 68], which were not recapitulated in our SIV-macaque experiments.
The accumulation of diverse monocyte/macrophage subsets in different tissue compartments during HIV and SIV infection has been described [27, 61–63, 69–72]. For example, CD163+ macrophages, a noninflammatory subset, accumulate in the gut while inflammatory Mac387+ macrophages infiltrate the brain [61, 63] suggesting disparate roles for macrophages during infection. Monocyte/macrophage accumulation in the liver was characterized by an expansion of both inflammatory CD16+CD14+ monocyte/macrophages and mature CD68+ macrophages during SIV infection. CD16+ monocytes/macrophages are presumed to represent a more activated phenotype and are elevated in the periphery in a number of chronic inflammatory conditions, including arthritis, atherosclerosis, Crohn’s disease and even during HIV/SIV infection [73–75]. In the context of liver disease, hepatic CD16+ monocytes/macrophages are potent producers of reactive oxygen species and secrete high levels of inflammatory chemokines and cytokines [76] with the highest levels of these cells found in patients with liver cirrhosis [77]. Although we cannot definitively determine if these inflammatory CD16+ monocyte/macrophages differentiate into tissue macrophages using our current model, Sugimoto et al. utilized BrdU labeling in rhesus macaques to demonstrate the capacity of CD14+CD16- classical monocytes to gradually progress to CD14+CD16+ monocytes and that these monocytes may differentiate into tissue macrophages [78]. Taken together, this suggests that the accumulation of these inflammatory CD16+ monocytes/macrophages is likely a key mediator of liver inflammation and subsequent hepatic dysfunction during SIV infection.
In both humans and rhesus macaques, replenishment and/or expansion of tissue macrophages due to infection or injury is most often associated with monocyte egress from bone marrow and entrance into systemic circulation in the blood, followed by homing to tissue and differentiation into mature tissue macrophages [79]. To further characterize the accumulation of hepatic monocytes/macrophages during SIV infection, we evaluated the CCL2-CCR2 chemokine network, a key inducer of monocyte/macrophage infiltration into the liver. We observed an upregulation of both CCL2 and CCR2 in the liver in macaques during untreated SIV infection with CCR2 expression positively correlating with CD68+ macrophages. Our flow cytometry evaluation of liver cells implicates CCR2 in the active recruitment of immune cells to the liver, where we observed elevated levels of CCR2+ T cells and CCR2+ monocyte/macrophage subsets. Our results suggest that SIV virus, not bacteria, induces immune cell trafficking to the liver as SIV levels in the liver and plasma correlated with both hepatic T cell and macrophage numbers, but bacteria levels did not. In fact, elevation of CCL2 has been described in multiple compartments (blood, gastronintestinal tract, the brain) during HIV infection with CCL2 levels correlating with viremia [80–82]. Moreover, HIV is a potent inducer of CCL2 production from many cell types in vitro, including liver stellate cells [83–85]. Therefore, we speculate that viral stimulation in the liver alters the immune environment through induction of CCL2, and possibly other chemokines, resulting in immune cell infiltration.
Another chemokine, inflammatory CCL3, which also correlated with macrophage number, is of major interest due to its ability to attract CCR5-expressing HIV/SIV target cells to the liver and the fact that this chemokine is involved in the retention of T cells at portal triads [86], which is where we observe SIV-infected cells to be primarily located in the liver. Previous studies have identified macrophages as key producers of CCL3 during viral infection and that CCL3 production is dependent on host interferon signaling [87]. Regarding its role during HIV infection, CCL3 is produced by macrophages following HIV exposure in vitro [88, 89], is upregulated in the livers of HIV-HCV co-infected individuals [90], and is linked to reduced CD4 T cell counts [91]. Collectively, this implies that under viral stimulation, macrophage production of CCL3 may directly influence infiltration of SIV-target cells, mainly T cells, to further amplify SIV levels and inflammation.
The portal region was particularly impacted during infection characterized by massive infiltration of immune cells, production of CCL2, and high levels of SIV-RNA positive cells. During some forms of chronic liver inflammation, portal tracts can organize into lymphoid follicles to form a specialized microenvironment for the recruitment and retention of activated immune cells in the chronically inflamed liver [92]. The factors that determine distribution of immune cells within the liver are poorly understood, but are likely controlled by differential expression of chemokines and adhesion molecules between the portal tract and the parenchyma. For example, homing to the portal triad involves RANTES, CCL2, and CCL3 while IP-10, MIG and ITAC have been implicated in recruitment to the sinusoids [93]. The immune signature captured in this study involving the upregulation of the chemokines CCL2 and CCL3, in combination with the various immune cell subsets aggregating at portal triads is consistent with portal disease during SIV infection.
We observed that T cells were the predominant cell type positive for SIV-RNA in the liver, which is consistent with other tissue compartments during normally progressing HIV and SIV infection [21, 94]. Regarding infection of hepatic macrophages, SIV RNA-positive macrophages were sparsely located throughout the lobular tissue of infected macaques. However, questions still remain as to whether these are truly productively infected macrophages or macrophages engulfing infected CD4 T cells. Several models of SIV infection have demonstrated infection of macrophages in several tissue compartments, including the spleen, lung, and brain, which is generally associated with a loss of CD4 T cells and accelerated disease progression [95–97]. Experimental depletion of CD4 T cells in macaques results in expansion of SIV-infected macrophages and microglia [98] demonstrating macrophage infection increases during disease progression as CD4 T cells decline. With regards to the liver, hepatic macrophages do not support viral replication in vivo [99] and also express low levels of CD4 and co-receptors, CCR5/CXCR4 [27], suggesting they may not be permissive to infection. Therefore, given the low number of SIV RNA-positive macrophages observed in our study, we speculate that macrophage infection occurs at low levels during normal disease progression in the liver.
Regarding the impact of antiretroviral drug intervention on liver inflammation, we observed a loss of macrophage accumulation, a reduced antiviral transcriptomic profile and decreased inflammation, which coincided with reduced levels of SIV DNA. SIV DNA was detected in the liver above the limit of detection for three of the five cART adult macaques analyzed, which has also been reported in human HIV infection during drug therapy [100]. Residual virus in the liver may be attributed to the role that the liver plays in clearance of activated T cells and SIV from circulation [26, 101]. Notably, although viral levels decreased in the liver during therapy, bacteria levels actually increased suggesting that gut barrier dysfunction persisted during cART. Murine studies have demonstrated elevated levels of bacteria compartmentalized in the liver during intestinal inflammation, and in models of experimentally induced liver disease, profound defects in bacterial clearance from the blood and heightened immune activation occur [102]. Therefore, although microbial translocation persists from the gut during antiretroviral therapy, liver function may improve, allowing for better clearance of microbial products. Indeed, we observed that LBP levels, a marker of bacterial translocation, decreased in the plasma during cART, but that liver bacteria levels increased suggesting the liver is filtering more microbial products from portal vein and systemic circulation.
Although the impact of elevated liver microbes in the context of SIV-associated liver dysfunction will warrant additional studies, gut microbial translocation is recognized as a possible cause of fatty liver disease and other metabolic syndrome manifestations, involving increased lipid peroxidation, ROS, and systemic inflammation [103]. Here, in macaques treated with antiretroviral drugs, there was observed alteration in metabolism characterized by downregulation in genes associated with ‘IL-1 mediated inhibition of RXR function’ in the transcriptome. In the context of liver function, perturbation of the retinoid X receptor (RXR) pathway is significant as RXR is a nuclear receptor that regulates transcription of many enzymes involved in the metabolism of lipids, cholesterol, bile acids, and xenobiotics (e.g. cART drugs). Altered activity of RXR with its heterodimers (e.g. PPAR, FXR, LXR) enhances fatty acid synthesis and/or alters fatty acid hepatic export leading to the accumulation of lipids in the liver (steatosis) [104], a common complication of HIV-infected people. Many of the top differentially expressed genes in antiretroviral-treated macaques are involved in metabolism suggesting that metabolism imbalance is prominent in these macaques.
There is limited understanding of how HIV infection impacts liver function in the absence of other factors that confound human studies, such as viral hepatitis co-infection, alcohol use, injection drug abuse, and diet. Utilizing a SIV-macaque model, we were able to undertake a controlled study of retroviral infection, antiretroviral drug treatment, and their impact on the hepatic immune environment. Based on previous studies and the data presented here, we propose the following model of liver dysfunction during HIV/SIV infection that is initiated by infection of SIV target cells, presumably CD4 T cells, that localize around portal triads. Viral stimulation initiates a CCL2 chemokine gradient resulting in infiltration of inflammatory monocytes/macrophages into the liver as part of the antiviral response. These infiltrating macrophages enhance recruitment of SIV-target cells to the liver through production of CCR5 ligands, such as CCL3, setting up a feedback loop to amplify SIV levels and further increase liver inflammation. The use of cART results in a reduction in this feedback loop where a decrease in SIV burden is associated with lower macrophage levels and a shift in the liver transcriptome that more closely resembles uninfected macaques. However, it is important to note that the inflammatory transcriptional signature was not completely ameliorated during therapy. Overall, these findings provide mechanistic insights into how SIV/HIV infection promotes liver inflammation through macrophage infiltration to result in liver complications that are observed in HIV-infected individuals.
All animal studies were conducted in accordance with protocols approved by the Center for Infectious Disease Research (protocol DS-05 UW), Washington National Primate Research Center (protocols 4314–01, 4213–02 and 4213–03) (Seattle, WA), and Yerkes National Primate Research Center (protocol YER-2002662) (Atlanta, GA) under Institutional Animal Care and Use Committees (IACUCs). All macaques in this study were managed according to the the laws, regulations, and guidelines set forth by the United States Department of Agriculture, Institute for Laboratory Animal Research, Public Health Service, National Research Council, Centers for Disease Control, the Weatherall Report titled “The use of nonhuman primates in research”, and the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International. The nutritional plan utilized by the WaNPRC and YNPRC consisted of standard monkey chow supplemented with a variety of fruits, vegetables, and other edible objects as part of the environmental enrichment program established by the Behavioral Management Unit. In addition, other means of enrichment were delivered and overseen by veterinary staff with animals having access to more than one category of enrichment. Paired (uninfected) macaques exhibiting incompatible behaviors were managed by the Behavioral Management staff and managed accordingly. SIV-infected macaques were housed in individual, adjoining cages allowing for social interactions. Primate health was monitored daily by trained staff. All efforts were made to minimize suffering through the use of minimally invasive procedures, anesthetics, and analgesics when determined appropriate by veterinary staff. Animals were painlessly euthanized by sedation with ketamine hydrochloride injection followed by intravenous barbiturate overdose in accordance with the recommendations of the panel of euthanasia of the American Veterinary Medical Association.
The liver tissue samples utilized in this study consisted of adult (uninfected N = 4, SIV+ N = 6, SIV+ cART N = 6) and infant (uninfected N = 7, SIV+ N = 9, SIV+ cART N = 4) Indian rhesus macaques (Macaca mulatta). All SIV-infected adult macaques were infected intrarectally with SIVMAC239x. Adult macaques receiving cART were started on an ART regimen 130 days post-infection consisting of subcutaneous tenofovir (20 mg/kg body weight) and emtricitabine (30 mg/kg) and oral raltegravir (50 mg twice daily). Samples from uninfected adult macaques were obtained through the tissue donor program from animals undergoing routine necropsy at Washington National Primate Research Center. Infant rhesus macaques (less than 9 months of age) were challenged orally with low doses of SIVmac251. Infants that become infected with SIV were placed in the SIV-infected group while infants remaining uninfected post-challenge were used as uninfected age-matched controls. SIV-infected infant macaques treated with cART were also infected with SIVmac251 and were placed on an ART regimen 35 days post-infection consisting of tenofovir, emtricitabine and dolutegravir administered as a triple formulation in a single daily injection. Liver samples and plasma were obtained from each animal at necropsy. To conduct the various experiments presented throughout this paper, liver tissues were either flash-frozen or formalin-fixed. A summary of SIV clinical parameters for the infected animals utilized in this study is presented below (Table 5).
Formalin-fixed, paraffin-embedded 5-um liver sections were stained with hematoxylin and eosin (H&E). A board-certified veterinary pathologist with expertise in nonhuman primate pathology reviewed the H&E stained liver sections in a blinded fashion. Findings from the liver pathology report were considered within normal background of macaques, and include vacuolar hepatopathy and lipid accumulation in stellate cells. No significant differences were observed between the treatment groups.
Liver tissues obtained at necropsy were fixed in 10% formalin and then paraffin embedded. Tissue sections (5-um) were mounted on glass slides and used for immunofluorescence microscopy. Slides were dewaxed in xylene and rehydrated through graded ethanols into distilled water. Antigen retrieval was performed using Antigen Unmasking Solution (Vector Laboratories, H-3300) in a decloaking chamber (Biocare Medical, Concord, CA) at 90°C for 30 minutes and then cooled for 10 minutes before removing. After washing slides twice for 5 minutes in 0.025% TritonX-100 in 1X TBS, tissue sections were blocked for two hours in 0.1% BSA + 1% goat serum in 0.025% TritonX-100 in TBS. Excess liquid was removed from the tissue section by aspiration and then each section was outlined with a hydrophobic barrier. Sections were incubated overnight at 4°C with specific antibodies for CD68 (clone KP1, Santa Cruz, 1:250) to identify macrophages, CD3 (clone SP7, ThermoFisher Scientific, 1:150) to identify T cells, and CD4 (clone BC/1F6, Abcam, 1:250) to identify CD4 T cells. The following day, slides were washed three times in 0.025% TritonX-100 in TBS for ten minutes each followed by incubation with fluorescent secondary antibodies for one hour at room temperature protected from light. CD68 and CD4 staining were detected using AlexaFluor 594 goat anti-mouse (Life Technologies, 1:500) while CD3 was detected using AlexaFluor 488 goat anti-rabbit (Life Technologies, 1:500). After the one hour incubation, slides were washed three times for five minutes each in 0.025% TritonX-100 in TBS. Slides were then mounted using Vectorshield Hard Set Mounting Medium with Dapi (Vector Laboratories, H1500), cured overnight at 4C, and then imaged the following day on a fluorescent microscope. Each liver section was imaged under 100X magnification in eight random fields of view. Macrophages were quantified using the particle analysis feature in ImageJ software while T cells were enumerated with the cell counting platform in ImageJ. Animals A12015 and A13273 were omitted from the microscopy analysis due to lack of fixed liver specimen.
Tissue sections (5-um) were dewaxed in xylene and rehydrated through graded ethanols into distilled water. Antigen retrieval was performed using Antigen Unmasking Solution (Vector Laboratories, H-3300) in a decloaking chamber (Biocare Medical, Concord, CA) at 95°C for 20 minutes. Slides cooled for 20 minutes before removing, and were then placed in water for five minutes followed by wash buffer (0.05% Tween 20 in TBS) for five minutes. Immunohistochemistry was conducted using reagents from the EnVision G2 Doublestain System (Dako, K5361). Slides were incubated in dual endogenous enzyme block for five minutes at room temperatures followed by two washes in wash buffer for five minutes each. Tissues were blocked with 0.25% casein in PBS for 30 minutes at room temperature. Primary antibody, either MAC387 (Abcam ab80084, 1:200) or CCL2 (Invitrogen MA5-17040, 1:500), was added to each slide and incubated for one hour at room temperature. Following two washes for five minutes each in wash buffer, polymer HRP was added for 30 minutes at room temperature, and then two additional wash steps. DAB chromogen (prepared per the manufacturer’s instructions) was added to each slide and the development of color was monitored under the microscope. All slides were stopped at the same time by placing the slides in di-water. Slides were washed one time in wash buffer and then counterstained in CAT Hemotoxylin for 15 seconds. Counterstained sections were rinsed in water until clear and then blued in Scott’s water for 30 seconds. Tissue sections were air-dried overnight and then mounted with Richard Allen Mounting Medium the following day. Images were acquired by Brightfield microscopy.
PBMC were obtained by ficoll-paque density centrifugation and cryopreserved in 90% FBS-10% DMSO at 10 million cells per mL. Liver cell suspensions were acquired from liver tissue at necropsy by flushing tissue with cold RPMI to remove blood, mincing the tissue into small pieces, and then mashing through a 70 μm cell strainer. Liver cells were pelleted, counted, and then cryopreserved in 90% FBS-10% DMSO at 10 million cells per tube. For flow cytometry analysis, cryopreserved cells were thawed, rested for 1 hour in RPMI at 37°C, 5% CO2, and then counted. Two million total cells were used for each stain. CCR2-BV421 clone 48607 (BD Biosciences) was added to each cell suspension and incubated at 37°C for 15 minutes. Next, antibodies for extracellular staining and live/dead staining (Live/Dead Aqua Fixable Dead Stain, Life Technologies) were added and incubated with cells at room temperature for 20 minutes. Extracellular antibodies consisted of CD16-BV605 clone 3G8 (BD Biosciences), CD14-BV785 clone M5E2 (Biolegend), CD11b-PE clone ICRF44 (BD Biosciences), CD45-APC clone MB4-6D6 (Miltenyi), and CD3-APC-H7 clone SP34-2 (BD Biosciences). Cells were washed once in PBS + 2% FBS, pelleted, and then permeabilized in FACS Juice for 10 minutes on ice. Following two washes in PBS + 2% FBS, cells were stained with intracellular antibodies for 20 minutes at room temperature. Intracellular antibodies consisted of CD68-PECy7 clone Y1/82A (BD Biosciences) and S100A9-FITC clone Mac387 (Thermo Fisher). After a final wash in PBS + 2% FBS, cells were resuspended in 1% PFA and acquired on a BD LSRII flow cytometer. Data were analyzed using FlowJo software (version 1.1.0-SNAPSHOT). Gates for cell populations were determined using fluorescence minus one (FMO) stained controls.
Liver cells were freshly isolated from liver tissue as described above. Two million liver cells were pelleted by centrifugation and resuspended in 100 uL of pHrodo E. coli bioparticles (Life Technologies, P35361) prepared at 1 mg/mL in PBS + 2% FBS. Cells were incubated with E. coli bioparticles for 2 hours at 37°C, 5% CO2. A negative control was conducted using 10 uM cytochalasin D. Cells were washed once with PBS + 2% FBS and then stained with CD3-Pacific Blue clone SP34-2 (BD Biosciences), CD4-BV650 clone OKT4 (Biolegend), CD8-APC-H7 clone SK1 (BD Biosciences), CD45-APC clone MB4-6D6 (Miltenyi), CD14-BV785 clone M5E2 (Biolegend), CD68-FITC clone Y1/B2A (eBioscience), and Live/Dead Aqua Fixable Dead Stain (Life Technologies) for 20 minutes at room temperature. After a wash in PBS + 2% FBS, cells were resuspended in PBS + 2% FBS and acquired on a BD LSRII flow cytometer with phagocytosed E. coli bioparticles detected on the PE filter. Data were analyzed using FlowJo software (version 1.1.0-SNAPSHOT).
Flash-frozen liver tissues were pulverized using a Retsch Planetary Ball Mill PM100 under cryogenic conditions using liquid nitrogen. To disrupt the tissue and obtain a fine powder homogenate, each tissue was placed in a 50 mL grinding jar with 20 mm stainless steel balls and subjected to 3 cycles of grinding at 300 rpm for 2 minutes each. The tissue powder was then collected into a sterile tube and stored at -80°C. For nucleic acid isolation, ~10 mg of liver powder was used to obtain RNA using the NucleoSpin RNA isolation kit (Macherey-Nagel) or DNA using the NucleoSpin Tissue Genomic DNA Isolation Kit (Macherey-Naglel). Nucleic acid concentrations for RNA and DNA were obtained using a NanoDrop 2000 Spectrophotometer (Thermo Scientific). Isolated nucleic acids were stored at -80°C until use.
cDNA was prepared from isolated liver RNA (1 ug) using a High-Capacity RNA-to-cDNA kit (Life Technologies) following the manufacturer’s instructions. cDNA was diluted (1:10) in nuclease-free water and used to quantify liver transcript levels of chemokines and cytokines, including TNFα, CCL3 (MIP-1α), TGFβ, IL-10, and CCL2 (MCP-1), and the chemokine receptor, CCR2, using rhesus macaque-specific TaqMan Gene Expression Assays (ThermoFisher Scientific). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the housekeeping gene for data normalization. Each 20 uL reaction contained 4 uL of diluted cDNA, 1 uL of TaqMan Gene Expression Assay Mix, 10 uL of Taqman Gene Expression Master Mix, and 5 uL water. PCR reactions were performed on an ABI 7500 detection system (Applied Biosystems) with one cycle at 95°C for 10 minutes followed by 47 cycles of 95°C for 15 seconds and 55°C for one minute. Any sample displaying high standard deviation between duplicates was omitted from downstream analysis. Relative expression of each transcript of interest was determined using the comparative cycle threshold (Ct) method. Genes of interest were normalized to the endogenous control GAPDH mRNA. Relative mRNA expression levels were then determined using the formula 2-ΔΔCt with an uninfected, age-matched control for determination of relative fold change. Data were then Log2 transformed for statistical and correlation analyses.
RNA was extracted as described above. RNA samples were then verified for purity, and the quality of the intact RNA was assessed using an Agilent 2100 Bioanalyzer. cRNA probes were made from each sample by Agilent one-color Quick-Amp labeling kit. Each cRNA sample was then hybridized to Agilent Rhesus whole-genome oligonucleotide microarrays (4x44k) based on the manufacturer’s instructions. Slides were scanned with an Agilent DNA microarray scanner, and the output images were then analyzed using Agilent Feature Extractor software. For each microarray, raw intensities, probe mappings, and quality-control (QC) metrics were uploaded into a custom laboratory information management system (LabKey Software).
Raw Agilent Microarray files were extracted using Agilent feature extractor version (version 10.7.3.1). Raw Microarray files were downloaded, background corrected using the “norm-exp” method with an offset of 1 and quantile normalized using the limma bioconductor package in the R statistical software environment (version 3.1.3). Replicate probes were mean summarized, and low expressed probes were removed. Exploratory analysis was performed in R. Statistical analysis was performed through the limma package and differentially expressed gene sets were uploaded into Ingenuity Pathway Analysis for Functional Analysis (IPA). Raw and Normalized expression data was submitted to the GEO (accession GSE97676). Gene expression profiles in SIV-infected and SIV-infected cART macaques were compared to age-matched, uninfected macaque expression data using limma. Differentially expressed genes were defined as having greater than 1.5-fold change over uninfected macaques with a Benjamini-Hochberg corrected (adjusted p value) less than 0.05. Co-expression analysis was performed using packages WGCNA and heapmap.2 in R/Bioconductor.
Plasma was obtained by centrifugation (1,000 x g, 10 minutes) from blood collected at necropsy and stored at -80°C until use. Levels of circulating immune-associated factors were determined using a custom 20plex Nonhuman Primate ProcartaPlex Multiplex Immunoassay (eBioscience). Assayed analytes include: IFNα, MIP-1α, IFNγ, IL-1β, TNFα, IL-1RA, IL-10, IL-12p70, IL-17A, IL-18, IL-23, IL-4, IL-6, IL-7, IL-8, IP-10, MCP-1 (CCL2), MIG, MIP-1β, sCD40L. Plasma from cART-treated infants was unavailable for analysis. Each plasma sample was assayed in duplicate according to the manufacturer’s instructions on a Bio-Plex 200 (BioRad, Hercules, California). Plasma concentration for each analyte was calculated from a standard curve using a five-parameter logistic regression after background subtraction. Analytes below the detection of limit were recorded as zero for correlation analyses.
Plasma viral loads were determined for all SIV-infected macaques at the necropsy time-point by real-time reverse-transcription PCR (RT-PCR) based on methods originally described by Suryanarayana et al. [105]. Briefly, plasma purified viral RNA was amplified using oligonucleotides for SIVgag in conjunction with a TaqMan-based probe. Viral load (copies/mL) was determined from a standard curve.
Levels of bacterial 16S DNA in the liver was assessed by qPCR using a Femto Bacterial DNA Quantification Kit (Zymo Research, E2006) per the manufacturer’s instructions. Briefly, total DNA was extracted as described above and diluted in nuclease-free water to a final concentration of 100 ng/uL. Each reaction contained 1 uL of total liver DNA (100 ng) and 18 uL of Femto Bacterial qPCR Premix. A ‘No Template’ Negative Control was also conducted to test for any possible contamination of qPCR reagents. PCR reactions were performed on an ABI 7500 detection system (Applied Biosystems) with one cycle at 95°C for 10 minutes followed by 40 cycles of 95°C for 30 seconds, 55°C for 30 seconds, and 72°C for one minute. One final extension step was performed at 72°C for seven minutes. A standard curve was generated with bacterial DNA standards (supplied with the kit) ranging from 0.00002 to 20 ng of bacterial DNA (R2 = 0.9576). The concentration of bacterial DNA in each liver sample was determined from the standard curve using a nonlinear regression four-parameter variable slope analysis. Duplicates were averaged for each animal and plotted as ng of 16S bacterial DNA per 100 ng of total input DNA.
LPS-binding protein (LBP) was quantified in the plasma of each macaque using an ELISA kit from Biometic (ABIN370809) per the manufacturer’s instructions. Absorbance was read at 450 nm using a Molecular Devices SpectraMax M2 plate reader. Any sample displaying absorbance below background was excluded from downstream analysis. A standard curve was generated (R2 = 0.999) and used to determine the concentration of LBP in each plasma sample assayed in duplicate.
Assessment of SIV DNA in the liver was measured by quantitative hybrid real-time/digital RT-PCR [52, 53]. For each liver sample, ten replicate reactions were run. Animal A13277 (SIV-infected cART adult) was omitted from this analysis due to lack of starting material for DNA isolation. Quantitation of samples showing positive amplification in all ten replicates was determined directly from a standard curve. For samples that did not have positive amplification in all ten replicates, SIV DNA quantification was determined from the frequency of positive amplifications, corresponding to the presence of at least one target copy in a reaction, according to a Poisson distribution of a given median copy number per reaction. Cell number was determined by qPCR for a single copy sequence from the rhesus macaque CCR5 gene and then used to normalize SIV DNA copies per 106 diploid genome equivalents. For three samples (1.2–2.7 X 105 diploid genome equivalents analyzed), SIV DNA was not detected, corresponding to a nominal assay threshold of less than 4–9 copies/106 cell equivalents.
In situ hybridization analysis was conducted using RNAscope technology (Advanced Cell Diagnostics) to identify SIV viral RNA-positive cells in the liver. Target probes, complementary to the SIV plus-RNA strand were designed to hybridize to SIV viral RNA in gag, pol, tat, env, vpx, vpr, nef, and rev genes. Upon binding, a pair of probes forms a double Z configuration and allows for signal amplification, followed by the chromogenic detection of SIV-RNA+ cells using a horseradish peroxidase enzymatic reaction. Liver sections (5-um) on glass slides were baked for one hour at 60°C and then placed in xylene (2 x 5 minutes) followed by 100% ethanol (2 x 3 minutes). Endogenous peroxidases were quenched using RNAscope hydrogen peroxide reagent (catalog no. 322335) for 10 minutes at room temperature followed by one wash in di-water. Antigen retrieval was performed in boiling 1x antigen retrieval buffer (catalog no. 322001) diluted in di-water for 30 minutes, followed by three quick washes in di-water and 1 minute in 100% ethanol. Sections were then air dried at room temperature (~ five minutes) and then outlined with a hydrophobic barrier. Protease III reagent (catalog no. 322331) was utilized at a 1:3 dilution in sterile PBS and incubated on the tissue sections at 40°C in the HybEZ hybridization oven (Advanced Cell Diagnostics) for 20 minutes followed by two rinses in di-water. SIVmac239 target probes (catalog no. 416141) were incubated on the tissue for two hours at 40°C in the HybEZ hybridization oven, followed by two washes in the RNAscope wash buffer (catalog no. 322000) for two minutes each. Signal amplification was conducted at 40°C in the HybEZ oven at the following conditions: Amp1 (catalog no. 322501) for 30 minutes, Amp2 (catalog no. 322502) for 15 minutes, Amp3 (catalog no. 322503) for 30 minutes, and Amp4 (catalog no. 322504) for 15 minutes. After each amplification step, slides were washed two times in RNAscope wash buffer for two minutes before proceeding to the next amplification step. The remaining amplification steps, Amp5 (30 minutes, catalog no. 322509) and Amp6 (15 minutes, catalog no. 322510) were conducted at room temperature with washing using RNAscope Wash Buffer conducted between the steps. Following amplification, signal was developed using DAB-A and DAB-B (mixed in a 1:1 ratio) on the liver sections for two minutes at room temperature followed by two washes in RNAscope Wash Buffer for two minutes each. Slides were then counterstained in hematoxylin for 30 seconds and rinsed with tap water until clear. Tissues were cleared in Scott’s Water for 30 seconds, then dehydrated through graded ethanols and xylene, and mounted using Permount mounting media. Whole liver sections were imaged by a Nanozoomer Scanner (Hamamatsu Photonics) and then analyzed by NDP Viewer software (Hamamatsu Photonics) under 5x and 20x magnification.
For phenotyping SIV RNA-positive cells, TSA Cy3.5 (Perkin Elmer, NEL76300), prepared 1:500 in Amplification Diluent, was substituted for the DAB-A and DAB-B reagent following the amplification steps. After seven minutes of incubation, TSA Cy3.5 treated slides were rinsed in di-water for 10 minutes, then blocked in 4% normal goat serum + 0.25% casein in 1X TBS with 0.05% Tween 20 (TBS-Tween) for 15 minutes. Mouse anti-CD68 clone KP1 (1:200, Santa Cruz) and rabbit anti-CD3 clone SP7 (1:100, ThermoFisher Scientific) were added to TSA Cy3.5 developed slides overnight at 4°C. The following day, slides were rinsed in TBS-Tween wash buffer, and then incubated with goat anti-mouse Alexa647 and goat anti-rabbit Alexa488 (both at 1:200, ThermoFisher Scientific) for one hour at room temperature. Following one wash in TBS-Tween, slides were incubated in Sudan Black for 15 minutes to reduce tissue auto-fluorescence. Slides were then washed one more time in TBS-Tween buffer, counterstained in DAPI (1 ug/mL) for 5 minutes and then mounted using Prolong Gold Mounting Medium (ThermoFisher Scientific). Images were acquired on a Nikon fluorescent microscope under 600x magnification.
Human PBMC were isolated by ficoll-paque density centrifugation from blood obtained from a healthy donor. Monocytes were isolated by negative selection using a Pan Monocyte Isolation Kit (Miltenyi). Isolated monocytes were counted and resuspended at 2 million cells/mL in RPMI containing 10% FBS, Penstrep, and 50ng/mL recombinant human GM-CSF (Peptrotech, 300–03). Mononcytes were plated in 24-well plates at 1 million cells/well (500 uL/well), and differentiated into macrophages over the course of seven days in the presence of GM-CSF (50 ng/mL). On day 3 and 5, the culture media was removed and replenished with fresh media containing GM-CSF. Any unattached cells were removed during feeding. On day 7, the culture media containing GM-CSF was removed and replaced with RMPI + 10% FBS and Penstrep (complete RPMI). Macrophages were rested for 24 hours, and then stimulated with either Poly I:C HMW (Invivogen, tlrl-pic) or ssRNA40 (Invivogen, tlrl-lrna40) at 0.2 and 2 ug/mL in duplicate for 12 hours. Unstimulated control macrophages received only complete RPMI media. After 12 hours, the supernatant was removed and macrophages were rinsed once with warm PBS + 2% FBS followed by lysis in 350uL of RA1 buffer containing 1% beta-mercaptoethanol. Cell lysates were stored at -80°C until RNA extraction. RNA was prepared using the NucleoSpin RNA isolation kit (Macherey-Nagel) and quantified using a NanoDrop 2000 Spectrophotometer (Thermo Scientific). One microgram of RNA was used to prepare cDNA using a High-Capacity RNA-to-cDNA kit (Life Technologies) following the manufacturer’s instructions. cDNA was diluted (1:10) in nuclease-free water and used to quantify transcript levels of CCL2, CCL3, TNFα, TGFβ and GAPDH using human-specific TaqMan Gene Expression Assays (ThermoFisher Scientific). Each 20 uL reaction contained 4 uL of diluted cDNA, 1 uL of TaqMan Gene Expression Assay Mix, 10 uL of Taqman Gene Expression Master Mix, and 5 uL water. PCR reactions were performed on an ABI 7500 detection system (Applied Biosystems) with one cycle at 95°C for 10 minutes followed by 47 cycles of 95°C for 15 seconds and 55°C for one minute. Relative expression of each transcript of interest was determined using the comparative cycle threshold (Ct) method. Genes of interest were normalized to the endogenous control GAPDH mRNA. Relative mRNA expression levels were then determined using the formula 2-ΔΔCt with unstimulated macrophages used for determination of relative fold change. Statistical significance was determined using Mann Whitney T tests comparing technical replicates of stimulated macrophages to unstimulated macrophages.
Statistical analyses were performed using Prism version 5.0f software (GraphPad Software, Inc., San Diego, CA). A Mann-Whitney nonparametric U test was used to compare all groups. For correlation analyses, data were first assessed for normality using D’Agostino & Pearson omnibus normality test. Unless normality was violated, Pearson Correlation analysis was conducted. Alternatively, Spearman Correlation was conducted for data not displaying normality.
|
10.1371/journal.pcbi.1006182 | Assessment of ab initio models of protein complexes by molecular dynamics | Determining how proteins interact to form stable complexes is of crucial importance, for example in the development of novel therapeutics. Computational methods to determine the thermodynamically stable conformation of complexes from the structure of the binding partners, such as RosettaDock, might potentially emerge to become a promising alternative to traditional structure determination methods. However, while models virtually identical to the correct experimental structure can in some cases be generated, the main difficulty remains to discriminate correct or approximately correct models from decoys. This is due to the ruggedness of the free-energy landscape, the approximations intrinsic in the scoring functions, and the intrinsic flexibility of proteins. Here we show that molecular dynamics simulations performed starting from a number top-scoring models can not only discriminate decoys and identify the correct structure, but may also provide information on an initial map of the free energy landscape that elucidates the binding mechanism.
| Determining how proteins fold and form complexes is of crucial importance, for example in the development of novel therapeutics. Experimental determination of structures is costly and lengthy. Computational methods to determine the thermodynamically stable conformation of complexes from the structure of the binding partners are available and constantly improving. Such methods generate a large number of diverse conformations and rank them for their likelihood to be correct. Even a model very similar to the correct structure is rarely the top-scoring one, but, as in the examples presented here, only within the top ~10–100 (the exact number depends on the complexity of the structure, and could be much higher). Here we show through atomistic simulation that good models are kinetically stable and bad models most often are not. More surprisingly, we also see that some bad models spontaneously find the correct (i.e., experimentally determined) conformation. This is remarkable, and could become an additional tool to contribute to structure determination of protein complexes. Such a result can also be expected, because evolution sculpted the free energy landscape in a way that the biologically active state is not only the one of lowest free energy (i.e., the most likely state) but also robustly reachable and kinetically stable (i.e., at the bottom of a funnel on the free energy landscape).
| Most biological processes are mediated by interactions between proteins. The high-resolution structure of protein complexes may help understand those processes at the molecular level and possibly interfere with them by rational design.
Predicting structures of complexes from known protein structures is, at first sight, a simpler task than predicting protein structures from sequences, i.e., ab initio. Yet, the facts that proteins are intrinsically flexible, and that they may change conformational propensity when interacting with other proteins, complicate the task. This entails changes from the very small to the very large (rotamers of interacting amino acids, movements of loops, up to domain orientations, and all combinations of them). Nevertheless, ab initio structure prediction methods constantly improve and can at present, at least for some relatively small systems, generate models that are indistinguishable from the experimental structure of the complex. Unfortunately, this is not the rule, however.
A number of different algorithms have been developed to dock proteins, including ZDOCK[1], PatchDOCK[2], ClusPro[3], ATTRACT[4], Gramm-X[5], DOCK/PIERR[6] and RosettaDock[7]. Some of them, including HADDOCK[8] or CamDock[9], are guided by experimental data. Their performance is assessed periodically in the Critical Assessment of Predicted Interactions (CAPRI), where different research groups compete in blind docking of diverse complexes[10].
Biochemical[11, 12] or evolutionary data[13, 14] may provide the constraints to navigate docking or evaluate decoys. In their absence one usually ends up with dozens of model candidates with similarly good scores. One reason is that scoring functions are only rough approximations of the free energy. Another reason is that many structures exist that are energetically close but structurally very different. Hence, the great challenge is to identify the correct model, i.e., the near-native structure(s).
For this purpose, several rescoring/re-ranking algorithms with different energy functions were developed. One of them is ZRANK which improves the ranking of the near-native poses across a benchmark set of different complexes[15]. FiberDock[16] or GalaxyRefine[17] perform additional backbone and side-chain relaxations. Different metrics to score and re-rank the docked poses are reviewed in Ref. [18]. Although rescoring often helps to narrow down the pool of model candidates, it usually does not unambiguously direct to the correct structure.
Molecular dynamics (MD) simulations have been widely used to study the dynamics of proteins on time scales that are, almost always, orders of magnitude shorter than the folding and binding times. On such short timescales MD may be used to assess the stability of structures of complexes obtained from small molecule docking[19]. In protein-protein docking, MD served typically for the local refinement of near-native decoys[17, 20]. Combined with Markov modeling, MD has been shown to be a powerful tool to recapitulate association kinetics[21]. A crucial property of a correctly docked conformation is that it would be expected to be near the bottom of a funnel in the free energy landscape, separated by sizeable barriers in free energy from incorrect conformations. Wrongly docked conformations, instead, may be either unstable or metastable conformations.
Here, we show that atomistic simulations, starting from a number of diverse, high-scoring models, provide valuable information on the local properties of the free-energy landscape that can be used to discriminate near-native from non-native protein-protein docking poses. For two different complexes, the Designed Ankyrin Repeat Protein G3 (DARPin G3) bound to domain IV of Human Epidermal Growth Factor Receptor 2 (HER2_IV)[22] and extracellular fibrinogen-binding protein (Efb-C) bound to C3–inhibitory domain of Staphylococcus aureus (C3d)[23], we show that the majority of decoys with reasonable scores in the initial docking are kinetically unstable and diffuse away from their initial conformation.
Remarkably, some decoys happen to be within the binding funnel on the free energy landscape and diffuse in our simulations to the correct structure on sub-μs timescales. Thus, these methods appear to capture binding events for models way off the correct structure that may even be trapped intermediates within the binding event. Such binding trajectories may thus also provide valuable information on the free energy landscape and the binding mechanism.
We used RosettaDock in its simplest form, without constraints nor post-processing, to generate a number of poses to be used as starting conformations for molecular dynamics simulations.
All-atom, fully solvated, molecular dynamics simulations were started from each of the best 50 models produced by RosettaDock and have been performed with the assumption that decoys are unstable or metastable states, and the trajectory will thus drift away from the initial structure.
During the room temperature simulations, we observed that most of the high scoring decoys indeed drift away from the initial configuration. After 32 ns at 303 K, 38 out of 50 models are more than 2.5 Å away from their initial structure (Fig 1).
To challenge the structures that during the simulation may get trapped into metastable conformations we increased the temperature by 30 K intervals and continued the simulations for 12 ns for each temperature. After 20 ns at 390 K, only model r37 remains within 2.5 Å RMSD from the initial structure. Model r37 is effectively the model by far the closest to the experimental structure (Fig 2). Ligands in other models cover the vast surface space of the receptor, including regions completely unrelated to the epitope (S1 Fig).
Interestingly, models r41, r48 and especially r23 appear to be kinetically stable, even when simulations are continued at higher temperatures, and only start deviating from the initial conformation when the temperature is raised to 390 K. The kinetic stability of r23 appears to be maintained mainly by Y46 of the DARPin fitting into the proline-rich hydrophobic pocket (P529, P543, P547) of the target, supported by the spatially neighboring R23 that may bridge with E544 (Supplementary Information S2 Fig).
The root-mean-square deviation from the experimental structure for each of the simulations (at 303 K) starting from the different models is shown in Fig 3. The simulation starting from model r37, which was shown already to be very close to the experimental structure, converges to the correct bound state after the initial equilibration.
Interestingly, model r44, initially at about 8 Å RMSD from the experimental structure, starts moving towards the experimental structure after about 15 ns and becomes indistinguishable from it after about 50 ns. This suggests that r44, while considerably off-target, can slide into the correct conformation without encountering sizeable free energy barriers. In other words, the model is likely to fall within a broad funneled region on the free energy landscape that corresponds to the correct bound state.
It is thus of interest to analyze what interactions need to be present for binding to occur fast, i.e., interactions that are likely formed at the transition state for binding. The C-terminal loop of the DARPin in model r44 appears to be positioned similarly to the loop of the near-native r37 (Fig 4A). In both models, the hydrophobic interaction between F112 of the DARPin and the patch formed by F555 and V563 is conserved. This serves as an anchor that allows a smooth transition to the correct pose. Additional hydrophobic contacts are provided by I79 and F81 that may slide around F555 (Fig 4B–4C).
The described hydrophobic clamp is likely the major energetic contribution to the native funnel on the free energy landscape. Model r22, initially closer than r44 to r37 lacks this interaction and cannot diffuse to the correct orientation within the studied time frame. In fact, r22 shares some of the other native contacts, mediated by the bottom of the second repeat and the C-cap of the DARPin. This is the N123-N534 interaction, together with a second important hydrophobic contact between F89 and V533/V552. Nevertheless, unlike for r44, the energetic barrier must be too high to be crossed by sliding without complete unbinding.
We looked at the polar contacts at the interface described in Ref. [22]. Interestingly, the first hydrogen bond is formed after ~24 ns between N123-N534 (Supplementary Information S3 Fig). This then allows further polar contacts to be established. Together with the hydrophobic anchor, the contact N123-N534 forms a hinge that allows pivoting of the wrong model into the correct pose over time.
The fact that two trajectories starting from different conformations converge to the same final one is a strong evidence that the latter is a unique minimum on the free energy landscape. An important caveat, however, is that one cannot conclude from a single simulation that r44 is kinetically closer to the natively bound conformation than any other model. To verify that r44 is effectively kinetically closer to the native structure than the other models, many finite length simulations should be started from each of the model [24]. For this reason, we performed a number of simulations starting from both r22 and r44. Results show that out of 12 simulations, none starting from r22 converge to the correct bound state, while three out of 12 do when started from r44 within 50 ns (Supplementary Information, S4 Fig).
For the Efb-C:C3d complex, the highest scoring model (r1) is a good hit, at an RMSD (Cα) of 1.6 Å from the crystal structure of the complex. However, it is followed by a crowd of false positives, i.e., structures with high Rosetta score and far from the native complex. The best hit (nearest native) is scored 63rd, at just 0.4 Å RMSD from the crystal structure. As the two components for docking (Efb and C3d) were derived from the complex structure and their binding surfaces thus perfectly match each other, it was expected that RosettaDock would provide a more accurate guess than for the DARPin G3:HER2_IV complex. The shape complementarity, ideal in this case, is the major factor considered in all docking functions. Nonetheless, the scores and the RMSD from the native structure correlate poorly (see S5 Fig in Supplementary Information).
The time evolution of the structures of the various models over 40 ns simulations is shown in Fig 5A (for clarity only simulations starting from the ten top scoring models are shown). As in the previous case, the structure deviates very little from the initial one for some trajectories, and these can be identified as likely good models. Indeed, also among these models there are false positives, i.e., metastable decoys. In Fig 5B the maximum RMSD from the initial structure is shown for 22 models (the first 21 and model 63). If simulations are performed at higher temperature (340 K) (Fig 5B), the number of false positives, i.e. models that do not diverge from the initial structure and are not near-native, decreases, while the number of positives (here defined as those at less than 2 Å RMSD from the experimental structure) does not change. In other words, those that do not move away from the initial structure are the nearest native ones. The four models (Fig 5B) that during the 40 ns simulation are always within 3.2 Å from the initial structure turn out to be the ones closest to the correct structure (with an RMSD less than 2 Å from the experimental structure), while all the other models, which end up at more than 3.2 Å from the initial structure during the simulation are the ones more than 6 Å RMSD away from the experimental structure.
As in the other case, we identified a decoy that diffused to the correct orientation. After 23 ns model r18, initially at about 7 Å RMSD from the experimental structure, starts moving towards the native conformation (Fig 6). The most conserved interacting residue between the model and the native state is R131 (Supplementary Information S6 Fig). Although its rotamers differ significantly, both are able to make a salt bridge to D1029 in the receptor. After 25 ns, Q134 and then N138 in the ligand find their right positions.
Discriminating near-native decoys in a crowd of false-positives is a fundamental challenge in protein-protein docking[25]. The score of a model does not generally correlate with the deviation of a model from the correct structure of the complex. This is due to a number of approximations taken by the docking algorithms to allow vast sampling of alternative conformations in reasonable time, ranging from simplified physical forces to purely statistical terms that are biased to the database used for training. Here we show that exploring the free-energy landscape around a hypothetical structure of a complex is a viable method for determining if a structure is close to a minimum of the free energy or within the funneled region of the free energy landscape where binding can occur on short timescales.
The approach we implemented here is simple: it consists in performing all-atom, fully solvated simulations of the systems in question, starting from the top-scoring models of complexes provided, in this case, by RosettaDock. Two scenarios are evident: in some simulations, the complex drifts away from the initial structure, while in others it remains close. The former could be identified as wrong models (“negatives”), and the latter as correct models of the real structure of the complex (“positives”). This was found to be often correct, but false negatives and false positives also occur. False positives result when the structure of the complex remains close to the initial structure because of a kinetic trap of the free energy landscape. False positives can be identified by performing simulations at high temperature, where metastable conformations have more chances to be overcome even in relatively short simulations. Indeed, the results presented here show that correctly docked complexes are stable even at (moderately) high temperature and most incorrectly docked complexes unbind during simulations. In all cases, this approach leads to a considerable reduction of false positives. Simulations longer that those performed here would reduce the number of false positives even further.
DARPin G3:HER2_IV is a remarkable case where essentially all the wrong models could be identified by monitoring the deviation from the initial structure by increasing the temperature up to 390 K. This may have been possible due to the very high stability of both components of the complex, where the proteins do not unfold even at very high temperatures (at least on a ~100 ns timescale). For many proteins such harsh temperature treatment may destroy their native fold.
Here we refer as false negatives to models that are kinetically unstable, but drift rapidly towards the correct structure. This demonstrates the existence of a relatively broad funnel on the free energy landscape around the correctly docked conformation.
An example here is the behavior of model r44 for G3:HER2 and model r18 for Efb-C:C3d that illustrates how a decoy may fall down the native free energy funnel and drift towards the correct structure of the complex. The models are initially at 5–8 Å RMSD from the correct structure but converge to the correct structure within 30–50 ns and remain there for the duration of the simulation.
Knowledge of the correctly docked structure was necessary here to detect such spontaneous binding events. However, even with the relatively small number of simulations, a decision on what the best model is may be made by observing the convergence of pairs of simulations to the same structure (i.e., within an RMSD that is typical of the correctly bound state for the specific complex). In Supplementary Information S1 and S2 Tables we report a list of pairs of models that at the end of the room temperature simulation are similar. For G3:HER2 the three top pairs include the two models that were correct or almost correct and the model that converges to the correct structure. For Efb-C:C3d the closest pair consists of two models that are similar to each other and at about 7 Å RMSD from the correct structure and do not drift away during the simulation; the second closest pair are models that are different (more than 7 Å RMSD from each other), and end up virtually identical after the simulation, which is a strong indication of both being within the binding funnel. Hence, in both cases a candidate for the correct model could have been uniquely identified even if the correct structure had not been available.
The results presented here highlight that empirical scoring functions are relatively good estimators of the thermodynamic stability of a protein-protein complex state, but because of relatively small energetic differences between unbound and bound states of proteins, even small errors in scoring functions may lead to false identification of the native state. A molecular dynamics simulation, where entropic and solvation effects are explicitly present, provides, albeit still approximate, an initial representation of the free energy surface over which the system diffuses. The overall shape and gradients appear to correspond to those of the real free energy surface: states with a low free energy but not confined by barriers in free energy diffuse away; correctly docked conformations appear to be kinetically stable; conformations within the funneled region of the free energy surface rapidly (in tens of ns) reach the correctly docked conformation. One take-home message is that atomistic force fields, particularly when solvent is considered explicitly, are sufficiently reliable. But most importantly, results show that the free energy landscape has been sculpted by evolution to be robust and that small errors in the atomistic force field do not alter the basic feature of there being a funnel around the biologically relevant conformation [26].
On the other hand, protocols such as RosettaDock thoroughly explore a large number of structurally diverse low energy conformers. Exploring the free energy surface by starting many independent short simulations from each of them, and possibly building a Markov state model from those [27], may become a viable way to determine the crucial features (minima and barriers) of the free energy surface and determine with high confidence the correctly bound state. The same holds true for predicting protein structures from sequences, although this is more challenging because of the much larger number of conformers that ab initio structure determination methods need to explore to identify models that are either correct or within the folding funnel. Therefore, all successful methods for structure prediction contain empirical terms taken from known structures to drastically reduce the search space. Similarly, empirical terms inherent in the docking programs mentioned in the Introduction are useful to decrease the initial search space, but are by themselves insufficient for the final ranking.
In conclusion, we have shown that molecular dynamics simulations starting from such models of docked complexes evolve depending on the local properties of the free energy surface. Even relatively short simulations starting from a large number of conformers selected by RosettaDock provide a valuable initial map of the free energy surface revealing the existence of more or less deep regions on the free energy surface.
Docking was performed with RosettaDock, a part of the Rosetta 3 suite, with talaris2014 as the scoring function during the refinement stage (details in Supplementary Methods) [28].
For the DARPin G3:HER2_IV complex, experimental structures of the unbound partners (i.e., not taken from the complex) were obtained from PDB:2jab chain A, residues 21–135 (DARPin G3) and PDB:1n8z chain C, residues 509–579 (HER2_IV). For better comparison to the known crystal structure of the complex, only the sequences corresponding to the resolved residues in PDB:4hrn (chain B and C) were considered. The structures were separately relaxed with all-heavy atom constraints, combined into one PDB file and docked via docking_protocol.linuxgccrelease. The number of trajectories was set to 105 (details in the Supplementary Methods). The top 50 poses (according to total score values) were analyzed. We refer to models according to their ranks (where r1 is best and r50 is worst).
For the Efb-C:C3d complex the structure of the monomers was taken from the bound structure of the complex (PDB:2gox, chain A and B). Monomers were re-docked with Rosetta to generate 5×104 poses. The top 21 models were further analyzed.
All-atom molecular dynamics simulations were performed starting from each of the highest scoring Rosetta models and from the crystal structures of the complexes. After a brief energy minimization, the models were solvated with enough water molecules so that the initial structure, when immersed in a periodic cubic box, is more that 16 Å apart from its closest image, and neutralizing ions (12 K+ for the DARPin G3:HER2_IV complex and 7 Cl- for the Efb-C:C3d complex) were added. The CHARMM36 force field [29] was used for the proteins and the standard TIP3P model [30] for water. A timestep of 2 fs was used to integrate the equations of motion, while all bonds involving hydrogen atoms were constrained. A cutoff of 12 Å was used for the interactions and the particle mesh Ewald method was used for electrostatics. Pressure was kept constant at 1 atm using a Langevin piston, while Langevin dynamics with a low damping coefficient (1 ps-1) were used to keep the temperature constant. Simulations were performed using NAMD [31].
|
10.1371/journal.pcbi.1002065 | Extracting Message Inter-Departure Time Distributions from the Human Electroencephalogram | The complex connectivity of the cerebral cortex is a topic of much study, yet the link between structure and function is still unclear. The processing capacity and throughput of information at individual brain regions remains an open question and one that could potentially bridge these two aspects of neural organization. The rate at which information is emitted from different nodes in the network and how this output process changes under different external conditions are general questions that are not unique to neuroscience, but are of interest in multiple classes of telecommunication networks. In the present study we show how some of these questions may be addressed using tools from telecommunications research. An important system statistic for modeling and performance evaluation of distributed communication systems is the time between successive departures of units of information at each node in the network. We describe a method to extract and fully characterize the distribution of such inter-departure times from the resting-state electroencephalogram (EEG). We show that inter-departure times are well fitted by the two-parameter Gamma distribution. Moreover, they are not spatially or neurophysiologically trivial and instead are regionally specific and sensitive to the presence of sensory input. In both the eyes-closed and eyes-open conditions, inter-departure time distributions were more dispersed over posterior parietal channels, close to regions which are known to have the most dense structural connectivity. The biggest differences between the two conditions were observed at occipital sites, where inter-departure times were significantly more variable in the eyes-open condition. Together, these results suggest that message departure times are indicative of network traffic and capture a novel facet of neural activity.
| The brain may be thought of as a network of regions that communicate with each other to produce emergent phenomena such as perception and cognition. Many potentially interesting aspects of brain networks, such as how information is emitted at different nodes, also tend to be of interest in various types of telecommunication systems, such as telephony. Thus, network properties that are relevant in the context of brain function may be important for telecommunication networks in general. Here we show how neural activity can be partitioned into units of information and analyzed from the perspective of a telecommunication system. We demonstrate that the inter-departure times of such units of information have very similar probability distributions across subjects and that they are sensitive both to regional variation and cognitive state. The approach we describe can be applied in a wide variety of experimental paradigms to generate novel indices of neural activity and open new avenues for network analysis of the brain.
| Recent years have witnessed a remarkable drive to characterize the large-scale structural topology of the brain. The graph model of cortical connectivity – whereby space is discretized and the brain is delineated as a set of regional nodes interconnected by white matter edges – has enabled the application of a whole host of network metrics [1], [2]. The cerebral connectome [3] has been found to possess highly nontrivial properties that do not appear in random networks with comparable connection density and could potentially endow it with a greater capacity to process information. These include small-worldness [4]–[6] and the presence of hubs [7], [8].
However, the functional consequences of this structural foundation are less clear and in general the translation from structure to function has been more difficult to understand. The emergent functional connectome has hitherto been studied by applying similar network analytic measures to graphs extracted from functional data. One approach has been to use these indices as a basis of comparison between networks defined by structural and functional connections. For example, physical links between nodes certainly beget sustained functional interactions and as a result functional brain networks map onto the underlying structural architecture to a great extent [8]–[10]. Another approach has been to study functional networks exclusively and without explicit reference to the underlying structural networks [11], [12].
An important aspect of brain network organization that remains to be investigated is the throughput of information at individual nodes. How does the flux of information vary across regions and under changing external and internal conditions? Do all nodes receive, process and relay messages at the same rate? Questions of this type often arise in relation to many classes of distributed communication networks [13]–[15]. Indeed, the brain must engage in networked computation [16]–[18], a challenge common to multiple types of telecommunication systems [19]. Therefore, it may be possible to learn more about the functional architecture and organizational principles of the brain by treating it as a network of regions that emit units of information.
Here we take the first step in adapting tools from telecommunications research to the problems in neuroscience. Namely, we show how electrophysiological recordings can be plausibly translated into a trace of departing units of information (henceforth referred to as “messages”) and analyzed from the perspective of a telecommunication system. By casting the problem in this light, we may be able to find new ways to describe, quantify and model the flow of information along the distributed brain network. One of the fundamental system statistics for modeling and performance evaluation of communication networks is the distribution of time between successive message departures at each node [13]–[15], [20], [21]. The inter-departure time depends on how messages get processed as well as the nature of their aggregated arrivals to a node and as such it reflects the flux of information through the network. In the present study we devised a method to delineate units of information in gross neurophysiological recordings and to fully characterize the distribution of their inter-departure times.
We first describe an intuitive signal processing approach that can be used to extract such events from the electroencephalogram (EEG). Participants were at rest, with both eyes-open and eyes-closed conditions. The data were resolved in the time-frequency domain using a wavelet transform. We defined message departure times as the local minima in the EEG scalogram, a definition based on the direct physiological interpretation of the EEG. Peaks and bursts in EEG signal power represent the synchronous firing of post-synaptic potentials from a population of neurons. If we take the neuron soma to be grey matter nodes in the network (as the graph model does), then the propagation of post-synaptic potentials to the axon hillock and along the axon may be thought of as the departure of a message. Thus, the troughs preceding each peak mark the point in time at which a unit of information departs from that population of neurons. We show that the distribution of time between successive departures (the inter-departure time) is well described by the family of two-parameter Gamma distributions. These distributions were fitted at each electrode and the two estimated parameters were then treated as dependent variables of neural activity.
If such events do indeed capture some aspect of information flow in brain networks, then we can make several testable predictions. First, the actual paths and sequences of “hops” between nodes will be largely determined by their structural connectivity, so inter-departure time statistics should be region specific and their spatial distribution should be heterogeneous. Second, as external demands change, so too should the manner in which units of information are emitted across the network and the distribution of inter-departure times at individual nodes should also be task-dependent. In particular, we expected the greatest change to be observed at or near occipital channels, given that the biggest difference between the eyes-closed and eyes-open states is the presence of visual input.
The experimental protocol was approved by the Research Ethics Board of the Montreal Neurological Institute and Hospital. Fifty-six (29 male) healthy children 10 years old (mean 10.0, standard deviation 0.393 years) participated in the study (see [22] for details). The participants were asked to keep their eyes open or closed in 8 alternating 30 s epochs (4 each). The electroencephalogram (EEG) was continuously recorded from 128 scalp locations using a HydroCel geodesic sensor net (Electrical Geodesics, Inc., Eugene, OR) referenced to the vertex (Cz). The signal was digitized at a rate of 500 Hz. Impedances did not exceed 60 kΩ. All offline signal processing and artifact correction was performed using the EEGLAB toolbox [23] for MATLAB (Mathworks, Inc.). Data were then average-referenced, digitally filtered [band-pass: 0.5–55 Hz; notch: 60 Hz] and epoched into 30 s segments. Only the middle 20 s of each epoch (5–25 s) were used in the analysis to avoid excessive contamination associated with opening and closing of the eyes. In the absence of a true baseline, the temporal mean was subtracted from each epoch. Ocular (blinks and lateral eye movements) and muscle artifacts were identified and subtracted on a subject-by-subject basis using the Infomax independent components analysis (ICA) algorithm [24] implemented in EEGLAB.
Dynamic spectral changes were estimated using a wavelet transform [25], implemented in the Wavelet Toolbox for MATLAB (Mathworks, Inc.). Trial epochs were convolved with a complex Morlet wavelet in a sliding window and signal power was estimated as the modulus squared of the real-valued wavelet coefficients (Figure 1B). The Morlet wavelet is a Gaussian-modulated complex sinusoid, so it is considered biologically plausible because it is more sensitive to transients in time series (more so than the windowed Fourier transform) and is widely used as an alternative way to model signals such as the EEG [26]. The mother wavelet had center frequency () equal to 1 Hz and envelope bandwidth equal to 2 s. Due to Heisenberg's uncertainty principle, there is a trade-off between the temporal precision and the spectral precision of the transform. Because our primary goal was to localize power fluctuations in the time domain, the bandwidth was deliberately chosen to be as narrow as possible to maximize the temporal precision of the transform, while maintaining at least two full cycles. The mother wavelet was compressed and applied at six scales, corresponding to frequencies of 5–30 Hz, in steps of 5 Hz. The corresponding pseudo-frequencies () were estimated as the inverse of the product of the scale () and digitization interval (): (1)
Departure times were identified by searching for all local minima in the scalogram (Figure 1B). To prevent minute and insignificant troughs from being selected, a local neighborhood threshold was set as a ratio (5%) of the range of the scalogram amplitude. The exact choice of the ratio in the range 2–10% did not impact the functional form or the parameters of the departure time distributions in any significant manner. The time between successive departures (inter-departure time, ) was calculated for each participant, condition, channel and wavelet scale (Figure 1C), producing samples with an average of inter-departure times.
Distributions of inter-departure times were then fitted with the two-parameter Gamma probability distribution function using maximum likelihood estimation (Figure 1D). The two free parameters estimated were the shape and scale . The Gamma probability density has the following form: (2)
The Gamma distribution was not selected a priori, but was determined to be the most appropriate distribution when the data were fitted with 30 common distributions and the goodness of fit was assessed by way of the test using EasyFit software (MathWave Technologies). The test statistic was significantly greater than the critical value for all 30 distributions (including the Weibull, Gaussian, generalized Pareto, etc.), indicating significant departure from all those distributions. However, the Gamma distribution had the lowest value across all fits and was ranked as the best-fitting distribution. Other common goodness of fit tests, such as the Kolmogorov-Smirnov and Anderson-Darling, were deemed inappropriate because they do not adjust the critical value to account for the degrees of freedom lost when parameters are estimated from the data. Upon visual inspection of the histograms it was clear that the two-parameter Gamma distribution offered an excellent fit to the observed data (Figure 2). The superiority of the Gamma distribution is demonstrated in Figure 3, which shows the fits for the Gamma and the next best-fitting distribution, the Weibull.
We treated each of the two parameters from the fitted Gamma distributions ( and ) as measures of neural activity. For each parameter we performed separate mean-centered partial least-squares (PLS) [27]–[29] analyses. PLS is a multivariate statistical technique that can be used to relate a design variable (e.g. experimental conditions) to a dependent measure of brain activity (e.g. or ) that varies across one or more dimensions (e.g. space and frequency). Singular value decomposition (SVD) is used to compute an optimal least-squares fit to the covariance between those two sets of variables (e.g. across all electrodes and conditions). Each solution is termed a “latent variable” (LV) and is expressed in terms of a pair of orthogonal vectors of design saliences and electrode saliences (analogous to component loadings in principal components analysis), as well as a scalar singular value (s). In the present analysis, each LV represented one contrast between conditions (design salience) in relation to a particular pattern of electrodes and frequencies that expressed that contrast (electrode salience). The “cross-block” covariance between the design block and electrophysiological data block that is captured by an LV is reflected by the singular value. Thus, effect size can be estimated as the ratio of the square of the singular value associated with that particular LV to the sum of all squared singular values derived from the decomposition.
Experimental effects captured by each LV were statistically assessed using resampling techniques. The significance of each statistical effect was determined using permutation tests. Each permuted sample was obtained by random sampling without replacement to reassign the order of conditions within participants (500 replications). The p-value was determined by calculating the proportion of permuted singular values that was equal to or exceeded the original singular value. The stability of the multivariate pattern expressed by electrode saliences was indexed by using bootstrap resampling to estimate their standard errors [30]. Bootstrap samples were generated by random sampling with replacement of participants within conditions (500 replications). Saliences were deemed to be reliable if the 99% confidence interval did not include zero. Under the assumption that the bootstrap distribution is unit normal, this condition holds if and only if the absolute value of the ratio of the salience to its bootstrap-estimated standard error is greater than or equal to 2.57 [30].
The empirical inter-departure time () distributions were fitted with the two-parameter Gamma distribution for each condition, subject, electrode and frequency. The Gamma distribution offered a good fit at all frequencies. Despite some individual differences in the parameters of the distribution, the form was remarkably consistent across subjects. This is illustrated in Figure 2, which shows the fits for all 56 subjects at one electrode and one frequency. Nevertheless, there was also substantial variation from subject to subject for both estimated parameters. To illustrate the individual variation of fits across frequencies, we also report the coefficient of variation of each parameter in the Eyes-Open condition, for electrode Cz, for the six frequencies, going from 5 to 30 Hz: 0.22, 0.20, 0.21, 0.20, 0.25 for the shape parameter; 0.33, 0.32, 0.34, 0.34, 0.37 and 0.31 for the scale parameter. The data indicate that both parameters are quite sensitive to individual differences. The spatial distributions of group means for and are displayed in Figs. 4 and 5 and discussed in more detail in the following subsection.
Note that since wavelets effectively act as a band-pass filter, the means of distributions should vary in proportion to frequency, such that departures are expected to occur at a faster rate at higher frequencies, resulting in lower mean inter-departure times. As an example, the group mean inter-departure times for the Eyes-Open condition, channel 60, going from 5 Hz to 30 Hz, were 50.9±1.3, 47.9±1.2, 45.3±1.2, 44.3±1.1, 43.1±1.1 and 40.9±0.8 ms. However, our analyses were concerned with identifying regional and state-dependent statistical effects and did not compare frequencies to each other.
Across all frequencies, the shape parameter of the fitted Gamma distributions was greater over the posterior (occipital and parietal) channels (Figure 4). Moreover, this measure was sensitive to experimental condition and was greater in the eyes-closed than in the eyes-open condition (Figure 4), an observation statistically supported by the PLS analysis (). The statistical effect was most reliable across all frequency bands over occipital channels and to a lesser extent over parietal and frontal channels (Figure 6, top row).
The scale parameter was lower at most posterior and vertical channels and generally much higher over temporal and anterior channels. This pattern was observed at all frequencies (Figure 5). Values were significantly greater in the eyes-open condition () and this effect was most stable over occipital channels (Figure 6, bottom row). There was also some suggestion of frequency dependence in the sense that the bootstrap ratios were slightly higher (i.e. the effect was more robust) at lower frequencies. It is worth noting that the most extreme values of were observed at electrodes close to the eyes (Figure 5), which tend to undergo the heaviest signal processing under most artifact rejection schemes. However, this does not affect the statistical analysis, as the condition differences at these electrodes were not reliable by bootstrap test.
We have described a signal processing method that can be used to identify message departure times from neurophysiological data and quantify the distribution of times between successive departures. The present study demonstrates that the two-parameter Gamma distribution offers a good fit to the inter-departure time distribution. The parameters of inter-departure time distributions were not uniform across the scalp and instead displayed spatial specificity. Namely, distributions recovered from medial posterior electrodes tended to have larger and smaller compared to anterior electrodes. This suggests that inter-departure times may be sensitive to regional differences in connectivity and/or processing capacity. In addition, inter-departure times proved to be sensitive to cognitive engagement, with significantly greater and smaller at occipital channels when participants kept their eyes open.
What does systematic variation in and tell us about the functional capacity of the underlying system? For example, what does it actually mean for a cortical region to produce inter-departure times with greater and smaller in the eyes-open condition? Here it may be instructive to consider other statistics of the distribution that are easier to interpret. For example, the coefficient of variation (, the ratio of the standard deviation to the mean) is a normalized measure of dispersion and for the Gamma distribution is given by (3)
Thus, inter-departure times were more variable at medial posterior channels compared with the rest of the scalp. Moreover, the distributions became more dispersed in the eyes-open condition and the effect was robust at occipital channels. These results suggest that inter-departure times capture a facet of network traffic. For example, traffic traces in telecommunication networks are found to be more variable under conditions of greater spectrum occupancy [31], [32]. The fact that inter-departure times were more variable at parietal channels is consistent with the notion that structures situated in posterior cortex (particularly close to the midline, such as the precuneus and posterior cingulate) enjoy an exalted status in the connectome. These regions tend to occupy positions along the shortest white-matter paths between all other regions of the brain and participate in the greatest number of structural [8], [33]–[35] and functional subnetworks [8, 11 36, 37].
Given that the biggest difference between eyes-open and eyes-closed is the availability of visual input it is not surprising that condition differences were expressed most reliably over the occipital portion of the scalp. This condition-dependent differentiation may reflect the transient reconfiguration of functional networks in response to changes in external input. For instance, as visual processing becomes more prominent in the eyes-open condition, more information should be routed through the occipital cortices. This should influence the rate of information exchange and total flux through the associated subnetworks, making the underlying biological and cognitive operations less regular and less predictable. This is reflected by our results, which indicate that when the eyes are open, both very short and very long inter-departure times become more likely than when the eyes are closed. The expression of condition differences at multiple frequencies precludes the interpretation that they are the result of a simple difference in power spectral density in the frequency band typically observed in visual tasks. For example, condition differences were not specific to activity resolved at 10 and 15 Hz.
From the perspective of telecommunication systems, the fact that inter-departure times were best approximated by the Gamma distribution is significant. The Gamma distribution arises naturally and often in such systems, particularly in relation to waiting times. For instance, the round-trip delay time for a packet on the Internet (the time it takes to travel from the source node to the destination and back to the source) is best modeled using the Gamma distribution [38]. In particular, when the shape parameter is a positive integer, the Gamma distribution can be thought of as the sum of independent exponentially distributed random variables, each with a rate parameter . This situation arises when a message must be processed or receive some type of service over a series of stations or stages at a server (termed an Erlang server, Figure 7), each of which has an exponential service time distribution. For instance, the server may represent a population of neurons (as in the graph model). The stages are simply a sequence of processes that take place before a unit of information is emitted. In the context of a neuronal ensemble, these processes may represent the interactions among cells within the ensemble. The time spent at the stage, , is drawn from the probability density function (4)
Since the service times are exponential, the expectation and variance for are given by: (5)(6)
The total time spent at the server (traversing the stages) is the sum of independent identically distributed random variables drawn from the distribution . Therefore, the expectation and variance of the total processing time can be calculated by summing across the stages: (7)(8)
Importantly, the coefficient of variation of the total service time is given by resulting in a hypoexponential service time distribution, named to denote the fact that the coefficient of variation for this distribution is smaller than that of the exponential distribution (i.e. 1) [13]. Hypoexponential service times indicate that the underlying processing stages are arranged in series (Figure 7). If there is any branching and some stations are arranged in parallel, service time distributions will be hyperexponential, with a coefficient of variation greater than 1 (for a detailed derivation see [13]). In the present data, inter-departure times were found to be hypoexponential, which under this theoretical framework is indicative of the former arrangement. This view is biologically plausible, because it suggests that once a unit of information arrives to a node, the sequence of operations performed on that unit is set and does not change from unit to unit. Note however, that although these stages may represent a transformative process, they do not necessarily alter the information content of each unit. Importantly, this derivation should not be misinterpreted as a statement about whether large-scale cognitive processes are coordinated in series or in parallel. Our data merely suggest that there is no variation in the sequence of steps performed on each unit.
The Laplace transform of the exponentially-distributed service time random variable with rate is (9)and the transform of the sum of such random variables is the product of their transforms (10)
The transform can then be inverted to give the distribution of total service time: (11)which is a special case of the Gamma distribution (Eq. (2)) where is a positive integer and the scale parameter is the inverse of the exponential rate parameter ().
Overall, this conceptualization of neural dynamics provides a novel narrative of information flow in the brain. This view suggests that units of information may be processed in a series of independent stages. Moreover, the number of processing stages () and the service rate at each stage () vary across regions of the brain and depend on internal and external conditions. The presence of visual input appears to engender a mode of operation with fewer processing stages but slower service rates. Thus, although it is not the only possible explanation, a telecommunication-based perspective offers a simple and biologically meaningful interpretation for the observed hypoexponential Gamma-distributed inter-trough times and the associated parameters and .
The idea to delineate signal units in the EEG and to characterize the sequence of inter-departure times is directly inspired by research in telecommunication networks. However, it is important to consider the physiological validity of the telecommunication model. To what degree are units of information recovered from the EEG scalogram comparable to data transmitted in a typical telecommunication network? In our approach, emitted peaks and troughs are de facto the basic units of information transfer, whereas in neural systems the more likely candidates would be action potential spikes or spike trains [39]. The key is that we would like to know how information emitted across the scalp changes under different experimental conditions. For this context and by virtue of their spatial scale and coverage, gross neurophysiological recordings such as the EEG which represent aggregated postsynaptic potentials from entire populations of neurons are the more appropriate measure of neural activity from which to isolate inter-departure times compared to single cell recordings. It is also interesting to note that, although action potential spikes are often modeled as a Poisson point process, inter-spike intervals (ISIs) measured from single cells often do not appear exponential but take on a functional form rather more similar to the Gamma distribution described here (e.g. Figure 1C in [40]).
The goal of the present study was to establish a foundation upon which the effects of experimental perturbations on communication in the brain could be studied, rather than to advocate any specific structural or functional similarities between telecommunication and brain networks. We sought to delineate physiologically meaningful units of information from gross electrophysiological recordings and to apply analytical tools from telecommunications research to describe how they are emitted across the network. However, some authors have articulated possible parallels between the brain and specific types of telecommunication networks. For example, Graham and Rockmore [41] posited that the brain may actually route and relay information in a manner analogous to packet-switching on the Internet, whereby a message is chopped up into a number of “packets” which are then transmitted along different paths to the destination, where they are re-assembled. The paths taken by individual packets are not pre-determined at the source and instead get adjusted dynamically at each node along the path according to network conditions. Under the current scheme for extracting inter-departure times it is not possible to infer the routes of individual messages. How information flow is directed in the brain and whether the mechanism bears any similarity to a packet-switching network remains to be determined.
However, the benefit of accurately characterizing inter-departure time distributions will be to inform future computational models and to test hypotheses about how information is directed in the brain. By combining physiologically realistic connectivity and realistic inter-departure time statistics, it will be possible to construct simulations with multiple types of routing mechanisms and dynamics unfolding over a cortical foundation. Such models will allow detailed examination of the communication capabilities of the cerebral cortex. For example, they could be used to answer a variety of interesting questions, such as which combinations of nodes and paths are particularly prone to congestion and which nodes become bottlenecks.
How will the present method generalize to other experimental settings, such as an event-related design with multiple shorter trials? One of the keys to fitting distributions to empirical data is sufficient sample size. In other words, to estimate the distribution of packet inter-departure times with a reasonable degree of confidence, one must generate many such packet departures. In a more traditional setting where time series are epoched into shorter segments the same procedure could be applied by calculating in all individual trials and collating them into a single sample to be fitted. In addition, it remains unclear what impact, if any, time-locked evoked responses would have on and this certainly warrants further investigation.
The EEG is vulnerable to volume conduction and therefore the spatial precision with which we were able to describe changes in inter-departure time distributions is naturally limited. Moreover, the present method treats all units of information in the same vein, even though peaks in the EEG scalogram vary in their amplitude. In other words, our method implicitly allows the possibility that units of information transmitted in the brain may vary in size. However, even if differences in message size were to be taken into account, this would not change the inter-departure time statistics extracted from the time series.
In the present study we applied tools from teletraffic engineering to the study of neural activity patterns. We have developed a way to identify electrophysiological events that may be interpreted as departing units of information and we have shown that the times between departures are distributed according to the Gamma probability distribution. In addition, we have demonstrated that this facet of neural activity is meaningful from the perspective of cognitive function. Namely, distributions of inter-event times are highly dependent on cognitive state and spatial location. We conjecture that inter-departure times reflect the flow of network traffic and index the communication capability of the brain's functional architecture.
|
10.1371/journal.pgen.1005410 | The Shelterin TIN2 Subunit Mediates Recruitment of Telomerase to Telomeres | Dyskeratosis Congenita (DC) is a heritable multi-system disorder caused by abnormally short telomeres. Clinically diagnosed by the mucocutaneous symptoms, DC patients are at high risk for bone marrow failure, pulmonary fibrosis, and multiple types of cancers. We have recapitulated the most common DC-causing mutation in the shelterin component TIN2 by introducing a TIN2-R282H mutation into cultured telomerase-positive human cells via a knock-in approach. The resulting heterozygous TIN2-R282H mutation does not perturb occupancy of other shelterin components on telomeres, result in activation of telomeric DNA damage signaling or exhibit other characteristics indicative of a telomere deprotection defect. Using a novel assay that monitors the frequency and extension rate of telomerase activity at individual telomeres, we show instead that telomerase elongates telomeres at a reduced frequency in TIN2-R282H heterozygous cells; this recruitment defect is further corroborated by examining the effect of this mutation on telomerase-telomere co-localization. These observations suggest a direct role for TIN2 in mediating telomere length through telomerase, separable from its role in telomere protection.
| The shelterin complex protects telomeres from being processed by the DNA damage repair machinery, and also regulates telomerase access and activity at telomeres. The only shelterin subunit known to promote telomerase function is TPP1, which mediates telomerase recruitment to telomeres and stimulates telomerase processivity. Mutations in shelterin components cause Dyskeratosis Congenita (DC) and related disease syndromes due to the inability to maintain telomere homeostasis. In this study, we have identified TIN2-R282H, the most common DC-causing mutation in shelterin subunit TIN2, as a separation-of-function mutant which impairs telomerase recruitment to telomeres, but not chromosome end protection. The telomerase recruitment defect conferred by TIN2-R282H is likely through a mechanism independent of TIN2’s role in anchoring TPP1 at telomeres, since TPP1 localization to telomeres is unaffected by the mutation.
| The multi-subunit shelterin complexes bind along mammalian telomeres, shielding the natural chromosome ends from engaging the DNA damage signaling and repair machinery [1]. Among the shelterin components, TRF1 and TRF2 bind directly to duplex telomeric repeats [2], while POT1 binds to the single-stranded regions of telomeres [3]. TPP1 forms a heterodimer with POT1 and enhances the affinity of POT1 for telomeric ssDNA [4]. Depletion of TPP1 or POT1 results in the deregulation of the single-stranded telomeric terminal overhang and the induction of a DNA damage response at telomeres [5–7]. TIN2 directly interacts with TRF1, TRF2 and TPP1, assuring structural integrity of the complex [8–10]. Depletion of TIN2 causes profound telomere deprotection phenotypes including destabilization of the shelterin complex, activation of telomeric DNA damage signaling, and increased apoptosis [9,11–14].
Increasing evidence suggests that the shelterin complex also regulates access of telomerase to telomeres and hence telomerase action on them. The best evidence for a shelterin-specific role in telomerase regulation comes from analysis of TPP1, which interacts with the telomerase catalytic subunit through the N-terminal OB-fold domain of TPP1 [15–18]. This interaction is crucial for recruiting telomerase to telomeres, as assessed by co-localization of telomerase RNA to telomeres through in situ hybridization analysis [19]. The TPP1/POT1 heterodimer also promotes telomerase processivity, as demonstrated by an in vitro direct telomerase activity assay [4,20]. Notably, mutations in the TPP1 OB-fold domain compromise telomerase-dependent telomere extension but not telomere end protection [18,21], indicating that TPP1 performs a role in telomerase regulation which is distinct from its contribution to chromosome end protection.
Whether other shelterin components also directly contribute to telomerase regulation has been less well characterized. Depletion of TIN2, which associates with TPP1, leads to reduced levels of TPP1-mediated telomerase association to telomeres [19], although this result might simply reflect an indirect function for TIN2 as a regulator of telomerase recruitment through anchoring TPP1 at telomeres. Intriguingly, an N-terminally truncated form of TIN2 lacking the TPP1 interaction domain can still induce significant telomerase-dependent telomere extension [8,22], suggestive of a TPP1-independent role for TIN2 in telomerase regulation.
An important resource for genetic defects in both telomerase and shelterin has come from Dyskeratosis Congenita (DC) patients. DC is an inherited disorder caused by abnormally short telomeres [23]. Clinically diagnosed by the mucocutaneous abnormalities, DC patients are prone to developing bone marrow failure, multiple types of cancers and a spectrum of diseases collectively characterized as “telomere syndromes” [24]. DC-causative mutations have been found in various telomerase ribonucleoprotein components affecting enzymatic activity (i.e. the telomerase catalytic subunit TERT and the RNA subunit TER) as well as telomerase biogenesis and trafficking (i.e. Dyskerin, NHP2, NOP10 and TCAB1) [25–30]. Recently, additional DC-causative mutations have been identified in shelterin components (i.e. TIN2 and TPP1) and other proteins involved in telomere replication (i.e. RTEL1 and CTC1) [31–38]. TIN2 mutations in DC patients correlate with aberrantly shortened telomeres and early onset of DC. Almost all patients reported thus far are heterozygotes, harboring only one mutated allele of TIN2. The vast majority of the disease-related TIN2 mutations are missense point mutations that cluster in a highly conserved yet uncharacterized region in the TIN2 C-terminus (a.a. 280–291), outside of the known TRF1, TRF2 or TPP1 interaction regions [10,39], with R282H being the most frequently observed mutation.
Using an ectopic expression system to examine the consequences of TIN2 DC mutations on telomere length in human cells, one recent study reported that overexpression of TIN2 DC mutants caused accelerated telomere shortening in HT1080-derived HTC75 cells, through a telomerase-dependent pathway [40]. However, another study challenged these conclusions and reported that overexpression of wild-type TIN2 and TIN2 DC mutants produced indistinguishable telomere length changes in HT1080 cells [41]. We note that overexpression studies have serious limitations as models to characterize the mechanistic basis for TIN2 dysfunction in DC patients: First, as observed by one of the above studies [40], ectopic expression of TIN2 increased endogenous TRF1 and TPP1 levels (both of which have roles in telomere length regulation). In contrast, neither TPP1 nor TRF1 accumulates to higher levels in TIN2 DC cells (under conditions where the mutant TIN2 protein is expressed at endogenous levels) than in wild-type cells. Second, since TIN2 binds to multiple shelterin proteins but not directly to telomeric DNA, overexpression of TIN2 can potentially sequester other shelterin proteins from telomeres. These side effects complicate the interpretation of TIN2 overexpression studies and may have caused the discrepancy between the overexpression studies. Collectively, these results have left unaddressed whether TIN2, like TPP1, has a direct role in telomerase regulation that can be distinguished from its telomere end protection activity.
Here we recapitulate TIN2-R282H heterozygosity in cultured telomerase-positive human cells using a zinc finger nuclease mediated knock-in approach to generate a TIN2 DC allele expressed at its normal level from its endogenous locus. Unlike the DC patient-derived cells which have been carrying the TIN2 mutations for years and may have acquired secondary compensatory mutations or epigenetic changes that potentially complicate the interpretation of phenotypes, our approach allows analysis of the immediate telomere phenotype in isogenic human cell clones that differ only in their TIN2 status. Analyses of these cells demonstrate that the TIN2-R282H heterozygosity has no impact on the telomere protection function of TIN2. Instead, we show that this separation-of-function defect in TIN2 leads to impaired telomerase recruitment, resulting in a reduced frequency of telomerase-mediated telomere extension events. These observations identify a second subunit of shelterin that mediates telomerase function, thereby further extending the premise that shelterin performs a dual role at telomeres.
Zinc finger nuclease (ZFN) mediated gene targeting was used to knock a TIN2 DC mutation into the human colon carcinoma cell line HCT116. HCT116 cells were chosen because they have active telomerase and wild-type shelterin components, they maintain a stable diploid karyotype suitable for gene targeting, and they are intact for most DNA damage-dependent checkpoints [42]. To knock TIN2 DC mutations into HCT116 cells, we designed a pair of ZFNs that specifically recognize unique sequences within TIN2 exon 6 [43] (S1A and S1B Fig). Two donor template constructs were used in the knock-in study: one carries a G to A mutation within exon 6 to introduce a single amino acid change from Arg to His at position 282 (R282H, one of the most frequently observed TIN2 mutations in DC patients); the other carries the wild-type sequence (WT) to generate the isogenic wild-type control cells. Translational silent mutations were introduced into the donor template at the ZFN recognition sites to prevent binding and cleavage of the construct by the ZFN pair.
Two knock-in clones heterozygous for the TIN2-R282H mutation (clone R282H.1 and R282H.2) and two knock-in clones wild-type for the TIN2 gene (clone WT.1 and WT.2) were established (S1C, S1D, and S1E Fig; see also Materials and Methods for details). Two TIN2 splice variants had been previously identified in human cells. The R282H mutation lies in the middle of the sixth exon shared between the splice variants (S1B and S2A Figs). As shown in S2B Fig, the expected TIN2 splice variants were produced in both the HCT116 parental cells and the knock-in clones, and sequencing of the reverse transcription products confirmed that both the mutated allele and the wild-type allele were transcribed into mRNA in the TIN2-R282H heterozygotes (S2C Fig). The functional difference between the two TIN2 splice variants is not yet characterized. In multiple human cell lines, only TIN2S could be detected [14,44,45], possibly due to the low abundance of the TIN2L protein. Immunoblotting analysis with an anti-TIN2 antibody raised against an N-terminal epitope of TIN2 (a.a. 44–58) detected one TIN2 protein band (TIN2S) of ~42KD in all clones (S2D Fig), consistent with the other reports. TIN2 protein levels in the R282H heterozygous clones were indistinguishable from those in the WT clones and parental HCT116 cells (S2D Fig), demonstrating that the R282H mutation did not significantly change TIN2 protein stability. Levels of the other five shelterin proteins were indistinguishable between TIN2-WT and TIN2-R282H clones as well (S2E Fig). Furthermore, immunoprecipitation analysis showed that there was no change in the interaction between TIN2 and its shelterin binding partner TRF1, TRF2 or TPP1 in the TIN2-R282H heterozygotes (S2F Fig).
To examine the effect of the TIN2-R282H mutation on telomeres in cells with active telomerase, we monitored telomere length of the HCT116 knock-in cells over successive cell divisions. Because telomere length is a heterogeneous trait, sub-clones of human cell lines with the same genotype can display variations in mean telomere length, as shown in Fig 1A and previously observed [46]. Despite differences in initial telomere length, telomeres in both TIN2-R282H heterozygote clones shortened progressively until they reached a mean telomere length slightly above 2kb; as expected, telomeres in TIN2-WT cells maintained stable lengths (Fig 1A). This establishes that a primary consequence of the TIN2-R282H heterozygous mutation is a progressive reduction in telomere length that occurs even in the presence of telomerase.
Notably, TIN2-R282H heterozygotes had a similar proliferation rate as TIN2 wild-type cells, even at late PDs (Fig 1B). Furthermore, immunostaining analysis showed that there were no detectable changes of telomeric localization for either TIN2 or other shelterin proteins in TIN2-R282H heterozygotes (Fig 1C and 1D). In particular, we found no evidence for TPP1 delocalization in response to the TIN2-R282H heterozygous mutation (Fig 1E), arguing that the telomere length phenotype conferred by the TIN2-R282H defect was not simply due to an indirect effect on TPP1 delocalization.
These initial observations also suggested that telomere protection was not impaired by the TIN2-R282H defect. To assess this more rigorously, we evaluated telomere dysfunction-induced foci (TIF) formation in early and late PD knock-in cells by performing immunostaining with an antibody against the DNA damage marker 53BP1 and telomeric fluorescent in situ hybridization (FISH) with a telomeric peptide nucleic acid (PNA) probe. Although HCT116 cells were fully functional for telomere dysfunction-induced DNA damage signaling (as indicated by the localization of 53BP1 to telomeres in HCT116 depleted of TRF2), there was no significant increase of TIFs, even in late PD TIN2-R282H heterozygote cells (PD51 and PD76) (Fig 1F). Together these results show that the heterozygous TIN2-R282H mutation does not cause shelterin redistributions or gross deprotection of telomeres.
TIN2 knock-in cells were also collected at early and late PDs for FISH analysis of metaphase spreads and examined for telomere abnormalities. No significant differences were found between TIN2-WT and TIN2-R282H cells at early population doublings. By PD51, the TIN2-R282H heterozygotes had a statistically significant increase in chromosome ends lacking detectable telomeric signals, which presumably reflects the very short telomeres in these cells. Notably, however, we did not observe an increase in fragile telomeres or chromosome end-to-end fusions (Fig 1G), further supporting the conclusion that TIN2-R282H heterozygosity did not lead to chromosome end deprotection.
The progressive telomere shortening in TIN2-R282H heterozygotes can be caused by either a defect in the telomerase pathway or by a telomerase-independent process such as increased telomere degradation. To distinguish between these two possibilities, we asked whether combining the TIN2-R282H mutation with a telomerase defect would confer an additive effect on telomere shortening, which would argue that TIN2-R282H mediated its effect on telomere length through a telomerase-independent mechanism. We ectopically overexpressed a dominant-negative form of telomerase catalytic subunit (DN-hTERT) [47] in parallel in TIN2-R282H heterozygotes and TIN2-WT cells (at PD8). DN-hTERT overexpression suppressed telomerase activity to undetectable levels as expected (Fig 2A), and caused progressive telomere shortening in all HCT116 knock-in clones (Fig 2B and 2C). Measurement of telomere length changes between PD8-PD28 cells showed that the expression of DN-hTERT led to similar rates of telomere shortening in TIN2-WT and TIN2-R282H cells (Fig 2C).
The above observations led us to consider that the mechanism underlying the progressive telomere shortening in TIN2 heterozygote cells was due to a defect in the telomerase pathway. To address this, we first monitored activity levels of the telomerase enzyme. Quantitative PCR analysis of endogenous telomerase RNA in HCT116 knock-in clones showed that the TIN2-R282H mutation did not cause a significant change in telomerase RNA levels (S3A Fig). Telomerase TRAP analysis also showed that telomerase enzymatic activity in TIN2-R282H heterozygotes was indistinguishable from that in TIN2-WT cells (S3B Fig), indicating that the progressive telomere shortening in TIN2-R282H heterozygotes was not caused by a reduction of the core telomerase enzymatic activity.
Since telomerase levels did not appear to be altered, we employed two assays to monitor the extent of telomerase activity (this sub-section) or telomerase recruitment (the next sub-section) at individual telomeres. To directly measure telomerase activity in vivo, we designed a novel assay to measure telomerase extension events at individual telomeres, by adapting a FISH-based assay which utilizes a telomerase enzyme that adds variant telomeric repeats to telomeres [48]. When the mutant template telomerase RNA, 47A-hTER, is expressed in telomerase-positive cell lines, it assembles with endogenous telomerase catalytic subunit hTERT into active telomerase to direct the incorporation of TTTGGG variant repeats at telomeres [49–51]. FISH with a (CCCAAA)3 PNA probe specifically detects the TTTGGG variant repeats, whereas the (CCCTAA)3 PNA probe specifically detects the canonical TTAGGG telomeric repeats (Fig 3A). In prior experiments with the mutant 47A-hTER telomerase RNA, it was detrimental to cell growth, causing chromosome end-to-end fusions [48,50]. However, in these previous experiments, the steady state expression level of 47A-hTER was at a ≥10:1 ratio relative to the endogenous hTER. In contrast, when the 47A-hTER was expressed at low levels (at ~1:1 ratio relative to the endogenous hTER; Fig 3B) for a limited number of cell divisions in HCT116 cells, we did not observe either chromosome fusion or growth inhibition effects (Fig 3A and S4A and S4B Fig), even though mutant TTTGGG repeats were added by telomerase to chromosome termini, as assessed by FISH (Fig 3A). The presence of very low levels of a variant telomerase enzyme allows us to quantify two aspects of telomerase activity at individual telomeres: (i) the relative frequency of telomere extension events, as determined by counting the fraction of telomeres that had incorporated TTTGGG repeats; and (ii) the relative length of extension at individual telomeres, as determined by measuring the fluorescence intensity of TTTGGG repeats at these newly extended telomeres. This provides a powerful assay to directly monitor the activity of telomerase at individual telomeres, at a resolution that has not been previously attainable by other approaches.
Expression of 47A-hTER at a 1:1 ratio relative to the endogenous hTER in each of the knock-in clones (Fig 3B) led to the assembly of enzymatically active 47A-hTER-containing telomerase at comparable levels in the knock-in clones, as shown by the 47A-hTER-specific TRAP analysis in Fig 3C (see Materials and Methods and S5 Fig for TRAP assay conditions that distinguish between 47A-hTER telomerase and wild-type telomerase). TRAP analysis also revealed that the expression of 47A-hTER caused only ~10% decrease of wild-type telomerase activity in each of the knock-in clones (Fig 3D), suggesting that 47A-hTER titrated away ~10% of hTERT from the endogenous telomerase complex. The reconstitution of only a small proportion of endogenous telomerase into 47A-hTER-containing telomerase in this time frame (8 days) likely reflects the extreme stability and very long half-life reported for telomerase RNA in telomerase-positive human cancer cells [52].
Due to the very low levels of the reconstituted 47A-hTER telomerase, only a small subset of telomeres in the knock-in clones had incorporated TTTGGG variant repeats (Fig 3E). Using this assay, we observed that the fraction of chromosome ends incorporating the variant repeats per metaphase in TIN2-R282H heterozygote cells was significantly less than that in TIN2-WT cells (Fig 3E), indicating that the R282H mutation caused a reduction in the frequency of telomere extension events. Strikingly, although the number of chromosome ends elongated by telomerase (based on the incorporation of variant TTTGGG repeats) was reduced in TIN2-R282H heterozygous cells, the amount of TTTGGG repeats added by the 47A-hTER telomerase at individual telomeres was not affected. This was revealed by measuring the fluorescence intensity of the TTTGGG variant repeat tracts at individual termini. As shown in Fig 3F, the distribution of the TTTGGG signal intensity in TIN2-R282H heterozygote cells was indistinguishable from that in TIN2-WT cells, indicating that the lengths of extension by the 47A-hTER telomerase at individual extended telomeres were comparable irrespective of the TIN2 status. These observations argued that the frequency, but not the extension rate, of telomerase extension events was reduced in response to the TIN2-R282H mutation.
The reduced frequency of telomere extension by the reconstituted 47A-hTER telomerase in TIN2-R282H heterozygotes suggested that telomerase recruitment to telomeres was compromised by the TIN2-R282H mutation. As a final step in our analysis of telomerase function, we directly examined the recruitment of endogenous wild-type telomerase to telomeres by their co-localization in the HCT116 knock-in cells. To do so, cellular localization of telomerase was monitored by RNA-FISH using established oligonucleotide probes complementary to the telomerase RNA component [18,53,54]. A mix of three oligonucleotide probes (~55 nt long) were used, each covalently labeled with five red fluorescence dyes, hence marking one telomerase RNA molecule by as many as fifteen fluorescence dyes, significantly amplifying the signal [55]. We carried out immunofluorescence staining against telomeric shelterin proteins TRF1 and TRF2, followed by FISH for telomerase RNA and analyzed the co-localization between telomerase RNA and telomeres (Fig 4A). Strikingly, we observed that the co-localization between telomerase RNA and telomeres was significantly lower in TIN2-R282H heterozygotes than in TIN2-WT cells (Fig 4B). These observations, combined with those shown in Fig 3, provide direct evidence that the heterozygous TIN2-R282H mutation impairs recruitment of telomerase to telomeres.
In this study, we have identified a novel function for the TIN2 subunit of shelterin, through the analysis of a separation-of-function TIN2 allele recovered from human Dyskeratosis Congenita patients. Using karyotypically stable, telomerase-positive human HCT116 cells, we have generated knock-in clones heterozygous for the DC-associated TIN2-R282H mutation. Notably, the resulting TIN2-R282H heterozygote cells do not display any characteristics of a telomere end protection defect. Instead, two independent systems for interrogating in vivo telomerase function at individual telomeres—a mutant repeat incorporation assay and the co-localization of telomerase RNA to telomeres—show that the TIN2-R282H mutation impairs telomerase recruitment, resulting in a reduction of the frequency of telomere extension by telomerase. Our model is consistent with a previous report that ectopically expressed TIN2-R282H pulls down less telomerase activity than wild-type TIN2 [40].
Originally defined of its role in chromosome end protection, the shelterin complex was thought to negatively regulate telomere length by sequestering telomeres away from telomerase. Functional characterization of the shelterin subunit TPP1, however, revealed two surprising roles of TPP1 in promoting telomerase recruitment and telomerase processivity, shifting the view of shelterin as solely an end protection complex which blocks telomerase from acting on telomeres. The work described here further demonstrates that a shelterin-dependent role in promoting telomerase function is not unique to the TPP1 subunit, as revealed by the pronounced telomere shortening in the TIN2-R282H mutant cells. Since TPP1 localization to telomeres is unaffected in TIN2-R282H heterozygotes, this argues that the telomerase recruitment defect of the TIN2-R282H mutant is not conferred through a defect in anchoring TPP1 at telomeres. However, it remains possible that the TIN2-R282H mutation renders TPP1 incompetent for interaction with telomerase. Furthermore, our data show that although telomerase is recruited to telomeres at a reduced frequency in TIN2-R282H heterozygotes, the average length of the extension product at those telomeres that are extended by telomerase is unaffected, suggesting that once telomerase is recruited to telomeres, it is as active as in TIN2-WT cells. Whether telomerase recruitment proceeds through a single coordinated pathway that involves the cooperative behavior of both TIN2 and TPP1, or through independent contributions by these two shelterin subunits, will be a subject for future investigation. For example, whereas TPP1 may have a direct role in recruitment, TIN2 may modify the conformation of telomeres, making them more accessible to telomerase. Interestingly, TIN2 binds to the heterochromatin protein 1γ (HP1γ) through the same region where DC-associated TIN2 mutations cluster. Disrupting TIN2-HP1γ interaction impacts both telomere cohesion and telomere length regulation [22]. HP1γ is required for establishing appropriate sister telomere cohesion and may be involved in shaping the local telomeric chromatin into a more favorable structure for telomerase association.
Of great interest is why the telomere maintenance defects in DC patients carrying heterozygous TIN2 mutations are usually worse than in those carrying heterozygous mutations in the telomerase enzymatic components. Studies of telomerase in induced pluripotent stem cells (iPSCs) have shown that the expression levels of telomerase catalytic subunit and telomerase RNA were both up-regulated significantly during the induction of pluripotency [53,56]. One potential explanation for the more severe form of the disease observed in DC patients carrying TIN2 mutations may be that during early embryonic development, the amount of TIN2, but not the core telomerase components, is the limiting factor for regulating telomerase activity. If so, this also suggests that the recruitment function of TIN2 may be a more tractable target for inhibition of telomerase activity during oncogenesis.
Finally, we point out that our results complement the recent analysis of another TIN2 DC mutation (TIN2-K280E) in knock-in mouse system. This mutation was found to confer both telomerase-dependent and -independent telomere shortening (the exact mechanisms remain to be characterized), as well as cause subtle telomere replication problems [41]. Whether the differences between the two studies are due to a difference in the molecular defect(s) of the two TIN2 DC alleles (TIN2-K280E versus TIN2-R282H) and/or to the differences in telomere maintenance between the two systems (normal mouse cells versus human cancer cells) remains to be determined.
The targeting construct (S1B Fig) was assembled by combining the following segments through overlapping PCR: a 2.7kb genomic fragment containing the human TINF2 gene, 1.8kb DNA fragment containing the puromycin N-acetyltransferase expression cassette flanked by loxP sites, and 1.8 kb of 3’ flanking DNA of the human TINF2 gene. MluI sites were engineered at the 5’ and 3’ ends of the construct to clone it into the pBluescript SK vector. Site-directed mutagenesis was used to engineer the silent mutations at the ZFN binding region (5’-CCATGCCAGACCCTGGGGGGAAGGGCTCTGAAG-3’ to 5’-CCTTGCCAGACACTGGGAGGCAGAGCTCTGAAG-3’). For the R282H targeting construct, site-directed mutagenesis was used to introduce the R282H (5’-GAGCGCCCC-3’ to 5’-GAGCACCCC-3’) mutation in exon 6. The ZFN recognition site is ~160bp from the R282H mutation. Full length sequences of the targeting constructs (~6.3kb) were confirmed by DNA sequencing before use.
TIN2 exon 6-targeting heterodimeric ZFNs (named T2-X6-L5+R4) were expressed in pCMV-FokI(DA+RV) plasmid system. The specificity and efficiency of the ZFNs were described in [43]. 2.5x105 HCT116 cells were plated in 6-well plate 24 hours before transfection. Cells were transfected with 4μg of linearized donor plasmid and 0.5μg of each ZFN encoding plasmid using 10μl of JetPrime transfection reagent (Polyplus). Puromycin selection was applied two days after transfection. Individual colonies were then picked and expanded.
Targeted HCT116 cell clones were screened by Southern blotting of NdeI+KpnI digested genomic DNA. 19 out of 768 clones were identified as correctly targeted clones. PCR analysis was performed on correctly targeted clones using the following primers: F1, 5’-TCTAGCTGGCCGACACTTCAATCT-3’; R1, 5’-CCTGCTAACCCTTTTAGGCACAGC-3’; R2, 5’-CTACCGGTGGATGTGGAATGTGTG-3’. R1 is specific to the unedited allele and R2 is specific to the targeting construct. F1+R1 and F1+R2 PCR products were sequenced to identify clones that contain only the designed change in sequences. PCR was also performed to amplify TIN2 genomic sequences encompassing all coding regions and sequenced to verify no additional mutations were present. To confirm both edited allele and unedited allele of TIN2 gene were transcribed, RT-PCR analysis was conducted on total RNA using the following primers: RT-F1 (spanning exons 5 and 6), 5’- TGGCTGCTTCCAGAGTGCTCTGTT-3’; RT-R1, 5’- TGGCTTCCTGGCCCTAGGAGGTAA-3’. The target sequence for the TIN2 shRNA is 5’- GAATCCTCCTCAGCAACAA-3’. After the screen procedure, we identified two TIN2-R282H heterozygous clones and four TIN2-WT knock-in clones. All four WT clones maintain stable telomere length over prolonged passaging. Two WT clones that have initial mean telomere lengths comparable to the respective R282H heterozygous clones were then selected for additional telomere maintenance analysis.
Nuclear extracts were made using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (PIERCE). Protein concentrations were determined by performing the Bradford assay (Bio-Rad). Samples were suspended with 2x Laemmli sample buffer, resolved with 10% SDS-PAGE, and detected by Western blotting using the following primary antibodies: mouse anti-TIN2 (Imgenex), mouse anti-TPP1 (Abnova), mouse anti-TRF1 (Genetex), mouse anti-TRF2 (Millipore), rabbit anti-Rap1 (Bethyl), and rabbit anti-POT1 (Abcam ab21382), followed by horseradish peroxidase-conjugated donkey anti-rabbit or anti-mouse IgG (Jackson ImmunoResearch), and visualized by the ECL prime reagent (GE Healthcare). The nuclear protein p84 was detected with a mouse monoclonal anti-p84 antibody (Genetex) as loading controls. Intensities of TIN2 bands were quantified by densitometry using the ImageJ software. Intensities of p84 were used to normalize between different samples.
Nuclear extracts were diluted 1:2 in TNE buffer containing 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 2 mM EDTA, 1% NP-40, and protease inhibitors. The diluted extracts were precleared by incubating with protein G-Sepharose beads (Sigma-Aldrich) at 4°C for 30 min. Immunoprecipitation was carried out by incubating the precleared supernatant with a rabbit polyclonal antibody [8] and protein G-Sepharose beads at 4°C overnight. The beads were washed six times with TNE buffer before fractionation on a 10% SDS-PAGE and Western blotting analysis. 10% of the amount used for the immunoprecipitation was fractionated directly on SDS-PAGE as input.
4–5 μg of genomic DNA was digested with RsaI and HinfI, fractionated by 0.5% agarose gel, then transferred to a Hybond XL membrane and hybridized to an end-labeled telomeric probe (CCCTAA)4. Signals were detected by phosphorimaging (Molecular Dynamics). Mean telomere lengths were analyzed by the ImageQuant software. Briefly, the telomere signal intensity over each lane was measured and plotted. The mean telomeric lengths were determined assuming a Gaussian distribution and calculated according to the positions of molecular weight markers run on the same gel.
The pHR’CMV lentiviral expression vector system used in this study was provided by Dr. Didier Trono. Telomerase RNA expressing lentiviral vectors contain the wild-type hTER or 47A-hTER cDNA driven by the IU1 promoter and a GFP gene driven by the CMV promoter [50]. The 47A-hTER template sequence is 3’-CAAACCCAAAC-5’. DN-hTERT- or luciferase- expressing lentiviral vectors contain the DN-hTERT or luciferase cDNA driven by the CMV promoter, followed by an internal ribosome entry site and a hygromycin resistance gene. DN-hTERT contains the D712A and V713I mutations which confer it catalytically inactive [47].
The day before infection, 2x105 cells were seeded on a 10cm plate and allowed to attach overnight. For viral infection, cells were incubated with virus-containing culture media supplemented with 8μg/ml polybrene. For in vivo telomerase function analysis using 47A-hTER, approximately 2 transducing units (TU) of lentivirus per cell were used for each infection. Cells were infected with >90% efficiency as indicated by a GFP expressed from the same lentiviral vector. After 24 hours, the virus-containing media was replaced with fresh media. Infected cells were pooled and passaged for subsequent analysis. For overexpression of DN-hTERT and control luciferase protein, approximately 30 TU of lentivirus per cell were used for each infection.
For combined immunofluorescence staining-telomere FISH, cells grown on coverslips were fixed with 4% formaldehyde and permeabilized with 0.5% NP40. Immunostaining was carried out by incubating with one of the following primary antibodies: anti-TRF1 (Genetex), anti-TRF2 (Millipore), anti-TPP1 (Abnova), anti-TIN2 (Imgenex), or anti-53BP1 (BD Transduction Laboratories), followed by incubating with secondary antibody conjugated with Alexa Fluor 488 (Molecular Probes). The cells were fixed again with 4% paraformaldehyde and dehydrated by successive incubation in 70%, 95% and 100% ethanol before subjected to telomeric FISH analysis using a TMR-OO-5’-(CCCTAA)3−3’ PNA probe as described previously [57]. DNA was stained by 0.1μg /mL DAPI. Coverslips were then mounted onto glass slides in Prolong Gold Antifade Reagent (Invitrogen).
Combined immunofluorescence staining-telomerase RNA FISH was carried out as described [58]. Briefly, cells grown on coverslips were fixed with 4% formaldehyde and permeabilized with 0.1% NP40. Immunostaining was performed by incubating with a mix of anti-TRF1 (Genetex) and anti-TRF2 (Millipore) antibodies to amplify telomere signal, followed by incubating with secondary antibody conjugated with Alexa Fluor 488 (Molecular Probes). The cells were fixed again with 4% paraformaldehyde and dehydrated by successive incubation in 70%, 95% and 100% ethanol. The cells were subsequently rehydrated in 50% formaldehyde in 2XSSC, incubated in prehybridization solution containing 10% dextran sulfate, 50% formamide, 2XSSC, 1mg/ml E. coli tRNA, 1mg/ml RNase-free BSA, 0.5mg/ml salmon sperm DNA, and 2mM vanadyl ribonucleoside complexes. Telomerase RNA FISH was performed by adding a mixture of three Cy3-conjugated telomerase RNA probes (30ng of each per coverslip) [54] to the prehybridization solution and incubating at 37°C in a humidified chamber overnight. The cells were then washed sequentially by 50% formamide in 2XSSC at 37°C, 0.1% NP40 in 2XSSC, 1XSSC and PBS. DNA was stained by 0.2 μg/ml DAPI and the coverslips were mounted onto glass slides in Prolong Gold Antifade Reagent (Invitrogen).
Cell images were acquired with a Nikon Ti-U microscope using a 100x objective and collected as a stack of 0.2 μm increments in the z-axis. After deconvolution using the AutoQuant X3 software, images were viewed with the Maximal Projection option on the z-axis. All image files were randomly assigned coded names to allow blinded scoring for spots co-localization and fluorescence intensity quantification.
Metaphase spreads and telomere fluorescence in situ hybridization was performed as described [57], using an Alexa488-OO-5’-(CCCTAA)3−3’ and a TMR-OO-5’-(CCCAAA)3−3’ PNA probe (Panagene). Images were acquired with a Nikon Ti-U microscope using a 60x objective. All image files were randomly assigned coded names to allow blinded scoring of variant repeats incorporation and fluorescence intensity. Telomere fluorescence intensity was quantified using the ImageJ software. Telomeric variant repeats signals on metaphase chromosomes were segmented manually and the integrated intensity from every segment was quantified. For each metaphase, the average background intensity was determined and subtracted from individual telomere signals.
Total RNA was extracted with the TRIzol reagent (Invitrogen). cDNA was prepared using the High Capacity RNA-to-cDNA kit (Invitrogen). Real-time PCR was performed using the Power SYBR green PCR master mix (Invitrogen), with respective set of primers at 50nM concentration, on a StepOnePlus Real-Time PCR machine. Telomerase RNA levels were normalized against GAPDH mRNA levels. Primer sets used: hTER forward 5’- GGTGGTGGCCATTTTTTGTC-3’, hTER reverse 5’-CTAGAATGAACGGTGGAAGGC-3’; GAPDH forward 5’-CATGTTCGTCATGGGTGTGAACCA-3’, GAPDH reverse 5’-ATGGCATGGACTGTGGTCATGAGT-3’.
Unless otherwise specified, telomerase activity was analyzed using the TRAPeze kit (Millipore) per manufacturer's directions. The telomeric extension products were separated by 10% TBE-PAGE and visualized by phosphorimaging (Molecular Dynamics).
For 47A-hTER or WT-hTER specific TRAP assay, TRAP reaction was carried out as described in [59] except that the return primer 5’-GCGCGGTACCCATACCCATACCCAAACCCA-3’ was used to detect 47A-hTER activity, and the return primer 5’- GCGCGGTACCCTTACCCT TACCCTAACCCT-3’ was used to detect WT-hTER activity. TRAP products intensity in each lane were quantified by the ImageQuant Software and normalized to the respective internal control intensity.
|
10.1371/journal.pbio.2003067 | Spliced integrated retrotransposed element (SpIRE) formation in the human genome | Human Long interspersed element-1 (L1) retrotransposons contain an internal RNA polymerase II promoter within their 5′ untranslated region (UTR) and encode two proteins, (ORF1p and ORF2p) required for their mobilization (i.e., retrotransposition). The evolutionary success of L1 relies on the continuous retrotransposition of full-length L1 mRNAs. Previous studies identified functional splice donor (SD), splice acceptor (SA), and polyadenylation sequences in L1 mRNA and provided evidence that a small number of spliced L1 mRNAs retrotransposed in the human genome. Here, we demonstrate that the retrotransposition of intra-5′UTR or 5′UTR/ORF1 spliced L1 mRNAs leads to the generation of spliced integrated retrotransposed elements (SpIREs). We identified a new intra-5′UTR SpIRE that is ten times more abundant than previously identified SpIREs. Functional analyses demonstrated that both intra-5′UTR and 5′UTR/ORF1 SpIREs lack Cis-acting transcription factor binding sites and exhibit reduced promoter activity. The 5′UTR/ORF1 SpIREs also produce nonfunctional ORF1p variants. Finally, we demonstrate that sequence changes within the L1 5′UTR over evolutionary time, which permitted L1 to evade the repressive effects of a host protein, can lead to the generation of new L1 splicing events, which, upon retrotransposition, generates a new SpIRE subfamily. We conclude that splicing inhibits L1 retrotransposition, SpIREs generally represent evolutionary “dead-ends” in the L1 retrotransposition process, mutations within the L1 5′UTR alter L1 splicing dynamics, and that retrotransposition of the resultant spliced transcripts can generate interindividual genomic variation.
| Long interspersed element-1 (L1) sequences comprise about 17% of the human genome reference sequence. The average human genome contains about 100 active L1s that mobilize throughout the genome by a “copy and paste” process termed retrotransposition. Active L1s encode two proteins (ORF1p and ORF2p). ORF1p and ORF2p preferentially bind to their encoding RNA, forming a ribonucleoprotein particle (RNP). During retrotransposition, the L1 RNP translocates to the nucleus, where the ORF2p endonuclease makes a single-strand nick in target site DNA that exposes a 3′ hydroxyl group in genomic DNA. The 3′ hydroxyl group then is used as a primer by the ORF2p reverse transcriptase to copy the L1 RNA into cDNA, leading to the integration of an L1 copy at a new genomic location. The evolutionary success of L1 requires the faithful retrotransposition of full-length L1 mRNAs; thus, it was surprising to find that a small number of L1 retrotransposition events are derived from spliced L1 mRNAs. By using genetic, biochemical, and computational approaches, we demonstrate that spliced L1 mRNAs can undergo an initial round of retrotransposition, leading to the generation of spliced integrated retrotransposed elements (SpIREs). SpIREs represent about 2% of previously annotated full-length primate-specific L1s in the human genome reference sequence. However, because splicing leads to intra-L1 deletions that remove critical sequences required for L1 expression, SpIREs generally cannot undergo subsequent rounds of retrotransposition and can be considered “dead on arrival” insertions. Our data further highlight how genetic conflict between L1 and its host has influenced L1 expression, L1 retrotransposition, and L1 splicing dynamics over evolutionary time.
| Long interspersed element-1 (L1) is a non-long terminal repeat (non-LTR) retrotransposon that comprises approximately 17% of human genomic DNA [1]. Over 99.9% of human L1s cannot retrotranspose due to 5′ truncations, internal DNA rearrangements, or point mutations that inactivate the L1-encoded proteins [1–4]. However, the average diploid genome harbors approximately 80–100 full-length retrotransposition-competent L1s (RC-L1s) [5], including a small number of expressed [6–8], highly active (i.e., “hot”) L1s [5,9–11] that can retrotranspose efficiently in cultured cells or cancers. RC-L1 retrotransposition affects both intra- and interindividual human genetic variation (reviewed in [12,13]) and, on occasion, can lead to disease-producing mutations [14–16].
Human RC-L1s are approximately six kilobases (kb) in length [17,18]. They contain a 5′ untranslated region (UTR) that harbors both sense [19] and antisense [20] RNA polymerase II promoters (Fig 1A) as well as an antisense open reading frame (ORF0) [21], which encodes a protein that may mildly enhance L1 retrotransposition efficiency. The 5′UTR is followed by two open reading frames (ORF1 and ORF2) that are separated by a 63–base pair (bp) inter-ORF spacer that contains two in-frame stop codons [18,22] (Fig 1A). L1s end with a 3′UTR, which contains a conserved polypurine motif, a “weak” RNA polymerase II polyadenylation signal, and a variable length polyadenosine (poly(A)) tract (Fig 1A) [17,23–25].
ORF1 encodes an approximately 40-kilodalton (kDa) protein (ORF1p) [26] that contains an amino-terminal coiled-coil (CC) domain required for ORF1p trimerization [27–29], a centrally located noncanonical RNA recognition motif (RRM) domain [29,30], and a carboxyl-terminal domain (CTD) harboring conserved basic amino acid residues [29–32] (Fig 1A). The RRM and CTD are critical for ORF1p nucleic acid binding [32–36]; the nucleic acid chaperone activity is postulated to play a role in L1 integration [30,36,37]. ORF2 encodes an approximately 150-kDa protein (ORF2p) [38–40] that contains conserved apurinic/apyrimidinic-like endonuclease (EN) [41,42] and reverse transcriptase (RT) domains [31,43,44] as well as a conserved cysteine-rich (C) domain [31,45] (Fig 1A). Biochemical activities contained within both ORF1p and ORF2p are required for canonical EN-dependent L1 retrotransposition in cultured human cells [31,41].
A round of human RC-L1 retrotransposition begins with the internal sense-strand promoter initiating transcription at or near the first nucleotide of the 5′UTR [13,19,46,47]. The resultant bicistronic L1 mRNA is exported to the cytoplasm, where it undergoes translation [22,46,48,49]. Following translation, ORF1p and ORF2p preferentially bind to their encoding mRNA in cis to form a ribonucleoprotein particle (RNP) [33,35,50–53]. The 3′ poly(A) tail of L1 mRNA is a critical Cis-acting determinant for recruitment of nascent ORF2p to L1 RNA [51,54]. Components of the L1 RNP gain access to the nucleus by a mechanism that does not require nuclear envelope breakdown [55, 127]. L1 integration likely occurs by target-site primed reverse transcription (TPRT) [41,53,56,57]. During TPRT, the L1 EN makes a single-strand endonucleolytic nick at a thymidine-rich sequence (e.g., 5′-TTTT/A-3′, 5′-TTTC/A-3′, etc.) present on the “bottom” strand of a target site in genomic DNA to liberate a 3′ hydroxyl (3′-OH) group [41,57,58]. Microhomology-based annealing between the L1 poly(A) tail and thymidine residues at the L1 EN cleavage site in genomic DNA enhances the ability of the L1 ORF2p RT to use the resultant 3′-OH group as a primer to initiate reverse transcription of L1 mRNA [53,59]. How TPRT is completed requires elucidation. However, as demonstrated for the related R2 non-LTR retrotransposon from Bombyx mori [60], it is possible that enzymatic activities associated with L1 ORF2p participate in both second-strand (i.e., “top” strand) genomic DNA cleavage and second-strand L1 cDNA synthesis.
Although retrotransposition assays and biochemical studies revealed the L1-encoded proteins preferentially retrotranspose their encoding mRNA in cis [50,53,61,62], L1 ORF1p and/or ORF2p can act in trans (Trans-complementation) to retrotranspose RNAs encoded by nonautonomous short interspersed elements (SINEs; e.g., Alu RNA [63,64] and SINE-R/VNTR/Alu [SVA] RNA [65–67]). Additionally, the L1-encoded protein(s) can act in trans to retrotranspose noncoding RNAs [12,68–73] and cellular mRNAs, with the latter process leading to the formation of processed pseudogenes [50,62,73–77].
The evolutionary success of L1 requires the faithful retrotransposition of full-length L1 mRNAs. Previous studies have revealed the presence of functional splice donor (SD), splice acceptor (SA), and premature polyadenylation signals in primary full-length RC-L1 transcripts [24,78–81]. Paradoxically, the use of these sites during posttranscriptional RNA processing leads to the production of truncated and/or internally deleted L1 mRNAs [24,78–81], which could adversely affect L1 retrotransposition. Thus, it is somewhat surprising that Cis-acting sequences that could negatively affect L1 retrotransposition have not been removed by negative selection during L1 evolution.
Here, we address how the retrotransposition of spliced L1 mRNAs leads to the generation of spliced integrated retrotransposed elements (SpIREs). We describe two classes of SpIREs: those that splice within the 5′UTR (intra-5′UTR SpIREs) and those that splice from within the 5′UTR into the ORF1 sequence (5′UTR/ORF1 SpIREs). Additionally, we suggest a mechanism for why some apparently deleterious Cis-acting splice sites within L1 mRNA are conserved throughout L1 evolution. Finally, we provide experimental evidence revealing that L1 splicing dynamics are altered by structural changes within the 5′UTR that allow L1s to evade host repression and that retrotransposition of the resultant spliced variants can lead to the generation of new classes of SpIREs. Thus, these data provide a snapshot of how an “arms race” between L1 and host repressive factors may affect the evolutionary trajectory of L1 5′UTRs. In sum, we conclude that SpIREs are deficient for retrotransposition and likely represent evolutionary “dead-ends” in the L1 retrotransposition process.
Using fosmid-based discovery methods, we previously identified a polymorphic L1 (fosmid accession #AC225317) in the human population that contains a 524-nucleotide deletion within its 5′UTR [10]. Upon closer inspection, we determined that this deletion likely resulted from the retrotransposition of a spliced L1 RNA that used a previously identified SD (G98U99) [78] and an unreported SA (A620G621) within the L1 5′UTR (numbering based on L1.3, accession #L19088; [9,82]) (Fig 1A and 1B). The structure of this element resembled previous L1s characterized by Belancio and colleagues, supporting the hypothesis that spliced L1 transcripts can complete retrotransposition in the human genome [78,79]. We named these L1s SpIREs to distinguish them from bona fide full-length genomic L1s. The three SpIREs investigated here all use the same SD (G98U99) but use different SA sequences that reside within either the L1 5′UTR (SA: A620G621 or SA: A788G789) or L1 ORF1 (SA: A974G975) (Fig 1B, 1C and 1D).
We used the BLAST-like alignment tool (BLAT) (https://genome.ucsc.edu) [83] (in which transposable element—derived DNAs are not masked) to search the human genome reference (HGR, GRCh38/hg38) for SpIRE G98U99/A620G621 sequences (referred to as SpIRE97/622). The HGR contains an annotated record of L1s that have accumulated over evolutionary time (i.e., millions of years); thus, searching the genome should reveal how SpIREs contribute to the genomic L1 repertoire. We used a 100-nucleotide in silico probe that spans the intra-5′UTR splice junction present in SpIRE97/622 (nucleotides 47–97 and 622–672 of L1.3) to query the HGR. We identified 116 SpIRE97/622 sequences, which span the youngest L1PA1 subfamily (also known as L1Hs, members of which are currently amplifying in the human population) through the older L1PA6 subfamily (which amplified approximately 27 million years ago [MYA]), but none in older (L1PA7–L1PA17, L1PB, and L1MA) L1 subfamilies (S1A Fig) [84,85]. Thus, 116 out of 6,609 (about 1.8%) of previously annotated full-length L1s in the L1PA1–L1PA6 subfamilies are actually SpIRE97/622 sequences (S1 Fig; S1 Data; S1 Table).
Almost half of the SpIRE97/622 sequences we identified belong to the L1PA3 subfamily (53 sequences, comprising about 3.4% of previously annotated full-length L1s in that subfamily) (S1A Fig; S1 Data; S1 Table). The L1PA1 subfamily harbors six SpIRE97/622 (comprising about 2.0% of previously annotated full-length L1s in that subfamily) and the L1PA6 subfamily contains only one SpIRE97/622 (comprising 0.1% of previously annotated full-length L1s in that subfamily) (S1 Fig; S1 Data; S1 Table). Seven SpIRE97/622 sequences could not be unambiguously assigned to a specific L1 subfamily and are classified as either L1PA2–L1PA3 or L1PA4–L1PA6 sequences (S1 Fig; S1 Data; S1 Table) [86].
Given the above data, we used BLAT to search the HGR for additional L1s containing G98U99/A788G789 and G98U99/A974G975 splicing events identified by Belancio and colleagues (referred to as SpIRE97/790 and SpIRE97/976, respectively) [78,79]. These searches confirmed the presence of four previously identified SpIRE97/790 sequences in the L1PA1–L1PA2 subfamilies (S1A Fig; S1 Data; S1 Table) [78]. We also discovered an additional SpIRE97/976 sequence in addition to the ten previously identified SpIRE97/976 sequences (S1A Fig; S1 Data; S1 Table) [78,79]. In total, these three classes of SpIREs comprise a small but notable (131/6,609 or about 2%) percentage of previously annotated full-length L1s from the L1PA1–L1PA6 subfamilies. The SpIRE97/622 sequences discovered here represent the majority (116/131 or about 89%) of identified SpIREs.
We next characterized the 131 SpIRE97/622, SpIRE97/790, and SpIRE97/976 sequences. We first examined the post-integration (i.e., filled) site of each SpIRE in the HGR sequence. We then used the genomic sequences flanking each SpIRE to reconstruct a putative pre-integration (i.e., empty) site. Many of the SpIRE sequences, especially those from older L1 subfamilies, have degenerate poly(A) tails at their 3′ ends, which, in some cases, made it difficult to reconstruct the putative pre-integration site to bp resolution (S1 Data; S1 Table). These analyses revealed that SpIREs generally are flanked by target site duplications that ranged in size from about 6–25 bp, end in a 3′ poly(A) tract, and integrated into an L1 EN consensus cleavage site (5′-TTTT/A-3′ and variants of that sequence) (S1 Data; S1 Table). Consistent with previous studies, approximately 39% (51/131) of the SpIREs are present within the introns of RefSeq (https://www.ncbi.nlm.nih.gov/refseq/) [87] annotated genes [69,88,89], and the majority (32/51 or about 63%) of these SpIREs are present in the opposite transcriptional orientation of the annotated gene (S1 Table) [90,91]. Other structural features of the SpIREs are shown in S1 Data and S1 Table. In sum, our analyses strongly suggest that SpIREs represent a subset of genomic L1 insertions and retrotranspose by the canonical process of L1 EN-dependent TPRT.
The formation of SpIRE97/622 results in the deletion of five known transcription factor binding sites within the L1 5′UTR [47,92–97] (Fig 1A); thus, we hypothesized the SpIRE97/622 5′UTR would have reduced promoter activity. To test this hypothesis, we subcloned the wild-type (WT) L1.3 [9,82] or SpIRE97/622 5′UTR sequences upstream of a promoter-less firefly (Photinus pyralis) luciferase gene (vector pGL4.11), creating pPLWTLUC and pPL97/622LUC, respectively (Fig 2A). We then characterized the promoter activity of these 5′UTRs using functional assays.
We first conducted northern blot analyses using polyadenylated mRNAs isolated from untransfected HeLa-JVM cells and HeLa-JVM cells transfected with the luciferase expression vectors (Fig 2). An RNA probe complementary to ribonucleotides 7–99 of the L1 5′UTR (Fig 2A; purple line) detected a strong signal at the expected size of about 2.7 kb in mRNAs derived from HeLa-JVM cells transfected with pPLWTLUC, but not in mRNAs derived from HeLa-JVM cells transfected with pPL97/622LUC or pGL4.11 or from untransfected HeLa-JVM cells (Fig 2B, first panel). Similar results were obtained using riboprobes complementary to either ribonucleotides 103–336 of the L1 5′UTR (Fig 2A, red line; Fig 2B, second panel) or the 3′ end of luciferase (Fig 2A, blue line; Fig 2B, third panel). These data are consistent with previously published findings [47], which demonstrated that L1 transcription begins at or near the first nucleotide of the L1 5′UTR. Control experiments verified the integrity and quality of the mRNAs (Fig 2B, actin probe).
We were able to detect a faint band representing the predicted approximately 2.2 kb mRNA from HeLa-JVM cells transfected with pPL97/622LUC upon the prolonged exposure of the northern blots using probes complementary to ribonucleotides 7–99 of the L1 5′UTR, but not using a probe complementary to ribonucleotides 103–336 of the L1 5′UTR (S2A Fig; purple arrow). The absence of the predicted approximately 2.2-kb band in HeLa-JVM cells transfected with pPL97/622LUC using a probe complementary to the 3′ end of the luciferase gene is likely due to the limits of detection in our assay (S2A Fig). The origin of the approximately 2-kb transcript remains unclear (Fig 2B, S2A Fig, orange arrow); however, it could be representative of transcript initiation downstream of the canonical transcriptional start site within the 5′UTR [47,98]. These data suggest that the SpIRE97/622 5′UTR retains weak promoter activity. Because the splicing events that gave rise to SpIRE97/790 and SpIRE97/976 led to larger deletions of the 5′UTR when compared to SpIRE97/622, we reasoned that they would lead to a similar, if not a greater, reduction in transcriptional activity; thus, they were not tested in this assay.
To corroborate the northern blot analyses, we conducted dual luciferase expression assays on whole cell lysates (WCLs) derived from HeLa-JVM cells co-transfected with firefly luciferase-based vectors (pPLWTLUC, pPL97/622LUC, or pGL4.11) and a constitutively expressed Renilla (Renilla reniformis) luciferase internal control plasmid (pRL-TK; Methods). Consistent with the northern blot data, HeLa-JVM cells transfected with pPLWTLUC exhibited an approximately 267-fold increase of normalized firefly luciferase activity when compared to cells transfected with the promoter-less pGL4.11 vector (Fig 2C; S2 Table). By comparison, HeLa-JVM cells transfected with pPL97/622LUC exhibited only about a 7-fold increase of normalized firefly luciferase activity when compared to cells transfected with the promoter-less pGL4.11 vector (Fig 2C; S2 Table). Together, the above data suggest that the splicing event leading to the generation of SpIRE97/622 severely compromises its promoter activity.
Given that splicing reduces L1 promoter activity, we examined why the G98U99 SD may be conserved in the L1 5′UTR. Previous studies revealed that a RUNX3 binding site within the 5′UTR is important for maximal L1 promoter activity [96]. Interestingly, the SD site used to generate the three classes of SpIREs reported here is contained within the core sequence of a RUNX3 binding site that is conserved from the L1PA1–L1PA10 subfamilies (Fig 1A; SD: G98U99; S1B Fig) [84]. Thus, we hypothesized that this SD is retained to maintain an active RUNX3 transcription factor binding site. To test this hypothesis, we mutated the SD sequence within the WT 5′UTR (U99C, creating pPLSDmLUC) [99] and tested if this mutation affects 5′UTR promoter activity. Northern blot analyses using the previously described riboprobes detected a signal at about 2.7 kb in mRNAs derived from HeLa-JVM cells transfected with pPLSDmLUC. However, there is markedly less of this mRNA when compared to cells transfected with pPLWTLUC (Fig 2B; about 18% of pPLWTLUC). In contrast, mutating the SA site within the WT 5′UTR (A620C, creating pPLSAmLUC) did not drastically affect L1 promoter activity (Fig 2B). Thus, our data are consistent with previous findings [96] and suggest that the retention of the complete RUNX3 site containing the G98U99 SD is critical for L1 promoter activity.
We next sought to identify spliced L1 mRNAs that might have given rise to SpIREs. To this end, we conducted end-point reverse transcription PCR (RT-PCR) experiments using poly(A) mRNAs isolated from HeLa-JVM cells transfected with a series of L1/firefly luciferase expression vectors (S2B Fig; Methods). The REV-LUC oligonucleotide (S2B Fig, purple line) was used to initiate L1/firefly luciferase first-strand cDNA synthesis; the cDNA products then were PCR amplified using FWD-5′UTR (S2B Fig, red line) and REV-LUC (S2B Fig, purple line) oligonucleotide primers. The resultant cDNAs were separated on an agarose gel, cloned, and characterized using Sanger DNA sequencing. Control experiments conducted in the absence of RT revealed that the characterized PCR products were derived from the amplification of cDNAs (S2C Fig).
We detected the predicted full-length L1/firefly luciferase cDNA products from HeLa-JVM cells transfected with pPLWTLUC, pPLSDmLUC, and pPLSAmLUC (Fig 2D, yellow “*” in lanes 1, 3, and 4) as well as the shorter predicted L1/firefly luciferase cDNA product from HeLa-JVM cells transfected with pPL97/622LUC (Fig 2D, yellow “#” in lane 2). In agreement with our northern blot experiments (Fig 2B), we did not detect cDNAs consistent with SpIRE97/622 splicing in pPLWTLUC transfected HeLa-JVM cells (Fig 2D). However, we did detect an L1/firefly luciferase cDNA that corresponds to the SpIRE97/790 splicing event from cells transfected with pPLWTLUC and pPLSAmLUC (Fig 2D, yellow “+”, lanes 1 and 4; Fig 1C) [78]. Importantly, this product was not detected in HeLa-JVM cells transfected with either pGL4.11 or pPLSDmLUC or untransfected HeLa-JVM cells.
We next tested whether intra-5′UTR splicing affects L1 mRNA translation. L1 sequences were cloned into an episomal pCEP4 expression vector that contains a hygromycin B resistance gene and a cytomegalovirus (CMV) early promoter, which augments L1 expression. HeLa-JVM cells were transfected with a WT L1 (pJM101/L1.3), an L1 that contains a 5′UTR deletion (pJM102/L1.3), or an L1 containing the SpIRE97/622 deletion (pPL97/622/L1.3) (Fig 3A) [9,50].
Western blot analyses were conducted using WCLs that were derived from hygromycin-resistant HeLa-JVM cells transfected with the above constructs 9 days post-transfection. An ORF1p polyclonal antibody (α-N-ORF1p; directed against amino acids +31 to +49 in L1.3 [100] [UniProtKB accession #Q9UN81]) detected an approximately 40-kDa product in cells transfected with pJM101/L1.3, pJM102/L1.3, and pPL97/622/L1.3 but not in cells transfected with the pCEP/GFP control (Fig 3B). HeLa-JVM cells transfected with pPL97/622/L1.3 exhibited a slight reduction in the steady-state level of ORF1p when compared to HeLa-JVM cells transfected with pJM101/L1.3 or pJM102/L1.3 (Fig 3B). Because a CMV promoter augmented L1 transcription, it is unlikely that this reduction is due to reduced L1 expression. It is possible that the slight reduction in ORF1p is due to an alteration of the L1 5′UTR RNA secondary structure and/or minor changes in the stability of pPL97/622/L1.3 mRNA when compared to pJM101/L1.3 and pJM102/L1.3 mRNAs.
The splicing event yielding SpIRE97/976 results in an amino-terminal ORF1 deletion of 66 nucleotides, including the canonical ORF1p methionine start codon (Fig 3C, black AUG, 40 kDa). We hypothesized that ORF1p synthesis might initiate from two methionine codons (AUG) that are located in weak Kozak consensus sequences either 102 or 270 ribonucleotides downstream from the canonical AUG start codon (Fig 3C) [101]. If the downstream methionine codons are used for translation initiation, we expect to detect amino terminal truncated ORF1 proteins of about 33 kDa and 27 kDa, respectively.
Western blot analyses were conducted as above using WCLs derived from hygromycin-resistant HeLa-JVM cells transfected with pJM101/L1.3, an L1 containing the SpIRE97/976 deletion (pPL97/976/L1.3), or pCEP/GFP control vectors (Fig 3A) [22]. As predicted, the α-N-ORF1p and α-C-ORF1p antibodies detected an approximately 40-kDa protein in WCLs derived from HeLa-JVM cells transfected with pJM101/L1.3 but did not detect a protein in WCLs derived from HeLa-JVM cells transfected with the pCEP/GFP control (Fig 3D, left and right panels). The α-N-ORF1p antibody detected an approximately 33-kDa protein in WCLs derived from HeLa-JVM cells transfected with pPL97/976/L1.3 (Fig 3D, left panel), whereas the α-C-ORF1p antibody detected approximately 33-kDa and approximately 27-kDa proteins in the same extracts and an unknown cross-reacting protein at about 25 kDa (Fig 3D, right panel). Similar results were obtained when RNP extracts were used in western blot experiments, although western blots performed with the α-C-ORF1p antibody did not detect the cross-reacting approximately 25-kDa protein (S3A Fig). To confirm that the approximately 33-kDa and 27-kDa products were ORF1p derived, we introduced a T7-gene10 epitope tag to the 3′ end of ORF1, creating pPL97/976/L1.3-T7. Western blots using a α-T7 antibody recapitulated our previous results and, similar to RNP preparations, did not identify the cross-reacting approximately 25-kDa protein (S3B Fig). Thus, the 5′UTR/ORF1 splicing event leads to the generation of an mRNA that, if translated, results in the synthesis of amino-terminal truncated derivatives of ORF1p.
Our data indicate that SpIRE97/622 contains a defective promoter and, if transcribed, SpIRE97/622 mRNA is translated at slightly lower levels than WT L1 mRNA. Thus, we hypothesized that an intra-5′UTR spliced L1 mRNA would be capable of undergoing an initial round of L1 retrotransposition. However, the resultant full-length retrotransposition events would contain a defective promoter, which may compromise subsequent retrotransposition.
To test the above hypothesis, we examined whether RNAs derived from a cohort of L1 expression constructs could retrotranspose using a cultured cell retrotransposition assay [31]. The 3′UTR of each construct contains a retrotransposition indicator cassette (mneoI). The mneoI cassette consists of an antisense copy of a neomycin phosphotransferase gene whose coding sequence is interrupted by an intron that resides in the same transcriptional orientation as the L1 [31,102]. This arrangement ensures that the expression of a functional neomycin phosphotransferase gene will only be activated upon L1 retrotransposition, thereby conferring cellular resistance to the drug G418 [31,102]. Retrotransposition efficiency then can be quantified by counting the resultant numbers of G418-resistant foci [31,61].
Consistent with previous reports (e.g., [31,41]), mRNAs derived from RC-L1s that contain both CMV and 5′UTR (Fig 4A, pJM101/L1.3, black bar; S3 Table), CMV only (Fig 4A, pJM102/L1.3, black bar; S3 Table), or 5′UTR only (Fig 4A, pJM101/L1.3ΔCMV, gray bar; S3 Table) promoters could efficiently retrotranspose. By comparison, the pPL97/622/L1.3 expression construct produced mRNAs that could undergo efficient retrotransposition when a CMV promoter augmented L1 expression (Fig 4A, black bar, about 70% the activity of pJM101/L1.3; S3 Table), but not when L1 expression was driven from the 5′UTR harboring the intra-5′UTR splicing event (Fig 4A, pPL97/622/L1.3ΔCMV gray bar, about 7% the activity of pJM101/L1.3; S3 Table). Consistent with this observation, control experiments revealed that an L1 lacking promoter sequences (Fig 4A, pJM102/L1.3ΔCMV; S3 Table) [50] was unable to retrotranspose. Additional controls demonstrated that an L1 containing a missense mutation (pJM105/L1.3; D702A) that disrupts ORF2p RT activity [50] severely reduced L1 retrotransposition efficiency (Fig 4A; S3 Table). Thus, the data suggest that the SpIRE97/622 intra-5′UTR splicing event severely compromises L1 5′UTR promoter activity as well as subsequent rounds of L1 retrotransposition.
The retrotransposition of an mRNA derived from a 5′UTR/ORF1 splicing event would generate a SpIRE (e.g., SpIRE97/976) that contains a defective promoter and, if transcribed and translated, would produce amino-terminal truncated versions of ORF1p. If the truncated version(s) of ORF1p were nonfunctional, we reasoned that the 5′UTR/ORF1 splicing event would lead to an L1 mRNA that is compromised for an initial round of retrotransposition in cis. Indeed, RNAs derived from pPL97/976/L1.3 could not retrotranspose despite expression being driven by CMV (Fig 4B; S4 Table).
We next hypothesized that a source of WT ORF1p would be required to act in trans to promote the retrotransposition of L1 mRNAs containing a 5′UTR/ORF1 splicing event. To test this hypothesis, we co-transfected pPL97/976/L1.3 (whose expression is augmented by a CMV promoter) with a series of “driver” L1 expression plasmids that lack the mneoI retrotransposition indicator cassette [22,50]. The co-transfection of pPL97/976/L1.3 with “driver” plasmids that express WT ORF1p, pJBM561 (a monocistronic ORF1p expression vector), pJM101/L1.3NN, or pJM105/L1.3NN, resulted in low levels of pPL97/976/L1.3 RNA retrotransposition in trans (Fig 4C; columns 1, 2, and 3, respectively; S5 Table). By comparison, the co-transfection of pPL97/976/L1.3 with “driver” plasmids that do not express ORF1p (pORF2/L1.3NN [a monocistronic ORF2p expression vector] or pCEP4) did not support retrotransposition in trans (Fig 4C; columns 4 and 5, respectively; S5 Table) [22]. Thus, the expression of ORF1p, but not ORF2p, can promote low levels of retrotransposition of mRNAs derived from pPL97/976/L1.3 in trans.
RT-PCR experiments using L1/firefly luciferase expression vectors uncovered evidence of SpIRE97/790 splicing events (Fig 2D). Intriguingly, SpIRE97/790 sequences are only present in the L1PA1 and L1PA2 subfamilies (S1A Fig; S1 Data; S1 Table) [84]. Indeed, the analysis of 1,000 genomes data [103] revealed that the L1PA1 SpIRE97/790-3 sequence (S1 Data; S1 Table) is polymorphic with respect to presence in the human population (about 41% homozygous “filled”; 35% heterozygous; 24% homozygous “empty”), whereas L1PA2 SpIRE97/790 sequences appear to be fixed with respect to presence in humans. Additionally, we identified four non-reference L1PA1 SpIRE97/790 sequences in data from the 1000 Genomes Project (S1 Data; S1 Table). Thus, SpIRE97/790 sequences may represent an evolutionarily younger SpIRE subfamily than the SpIRE97/622 and SpIRE97/976 sequences, which are predominantly found in older L1 subfamilies (S1 Data; S1 Table).
Recently, an elegant study from the Haussler laboratory demonstrated that the Krüppel-associated Box-containing Zinc-Finger Protein 93 (ZNF93) could bind within L1PA3 and L1PA4 5′UTRs to repress their expression [104]. Intriguingly, a 129-bp deletion that eliminates the ZNF93 binding site within the L1PA2 and L1PA1 5′UTRs allowed them to evade ZNF93-mediated repression [104]. This 129-bp sequence resides between a putative branch site and the SA sequence used to generate the spliced L1 RNA that gave rise to SpIRE97/790 sequences (Fig 5A). Thus, we hypothesized this 129-bp deletion may have altered L1 5′UTR splicing dynamics by relocating the SpIRE97/790 SA (A916G917 in L1PA3) to a favorable splicing context in L1PA2 and L1PA1 subfamily members (Fig 5A).
To test the above hypothesis, we generated L1/firefly luciferase expression vectors containing the 5′UTR of a “hot” L1 (L1RP [accession #AF148856]) [106] or a version of the L1RP 5′UTR that includes the 129-bp L1PA4 sequence containing the ZNF93 binding site [104] upstream of a promoter-less firefly luciferase gene (pGL4.11), creating pJBMWTLUC and pJBMWT129PA4LUC, respectively (Fig 5B, top panel). We also created a control vector that has a “scrambled” version of the 129-bp L1PA4 sequence (pJBMWT129SCRLUC) [104] (Fig 5B, top panel). Dual luciferase assays using WCLs derived from HeLa-JVM cells co-transfected with pJBMWTLUC, pJBMWT129PA4LUC, pJBMWT129SCRLUC, or pGL4.11 and a constitutively expressed Renilla luciferase internal control plasmid (pRL-TK; Methods) revealed that pJBMWTLUC and pJBMWT129PA4LUC exhibited an increase (about 345- or about 320-fold, respectively) of normalized firefly luciferase activity, when compared to the promoter-less pGL4.11 vector (Fig 5B, bottom panel; S6 Table). By comparison, pJBMWT129SCRLUC exhibited a significant, though less pronounced, increase (about 88-fold) of normalized firefly luciferase activity (Fig 5B, bottom panel; S6 Table). Thus, in general agreement with previous studies [104], the 129-bp L1PA4 insert does not negatively affect L1RP5′UTR transcriptional activity. As an additional control, we confirmed that the 129-bp L1PA4 sequence did not significantly affect L1 activity using an EGFP-based retrotransposition assay (Fig 5C; S7 Table) [104].
To test whether the presence or absence of the 129-bp L1PA4 sequence affects intra-L1 5′UTR splicing, we used a slightly modified version of the end-point RT-PCR strategy depicted in Fig 2D. In agreement with experiments performed with pPLWTLUC (Fig 2D), we detected the predicted full-length L1RP 5′UTR cDNAs as well as SpIRE97/790 spliced cDNAs in cells transfected with pJBMWTLUC (Fig 5D, yellow “*” and yellow “+,” respectively, lane 3). By comparison, HeLa-JVM cells transfected with pJBMWT129PA4LUC yielded the predicted full-length 5′UTR L1 cDNA (Fig 5C, yellow “**” lane 4), but did not yield cDNAs corresponding to the SpIRE97/790 splicing event. Instead, we detected a new spliced cDNA that used the same G98U99 SD and a new SA that resides within the 129-bp L1PA4 sequence (A851G852), which is not present in the WT L1RP sequence (Fig 5A and 5D, lane 4, yellow “@”). Finally, we detected the predicted full-length L1RP 5′UTR cDNAs from cells transfected with pJBMWT129SCRLUC, as well as a biologically irrelevant product that utilized the same G98U99 SD and an SA that resides within the 129-bp L1PA4 scrambled sequence (Fig 5D, lane 5, yellow “***” and yellow “$,” respectively). Thus, our data demonstrate that the loss of the 129-bp sequence from L1PA3 resulted in a new splicing pattern that led to the emergence of SpIRE97/790 sequences (Fig 5A and 5D).
Finally, we examined whether the new cDNA detected from cells transfected with pJBMWT129PA4LUC corresponds to a SpIRE. Indeed, a BLAT search of the human genome using an in silico probe that spans the intra-5′UTR splice junction present in this putative SpIRE (nucleotides 47–97 and 853–903 of pJBMWT129PA4LUC) yielded nine additional SpIRE97/853 sequences (S1 Data; S1 Table). These additional SpIREs retain L1 structural hallmarks (S1 Data; S1 Table), indicating that canonical EN-dependent TPRT led to their generation.
The evolutionary success of L1 requires the continued retrotransposition of full-length L1 RNAs. Thus, it was surprising when Belancio and colleagues identified a small number of L1 retrotransposition events in the HGR that apparently were derived from spliced L1 RNAs [78,79]. Here, we confirmed and extended those findings and report a novel group of retrotransposed L1s that are derived from an L1 RNA containing an intra-5′UTR splicing event (SpIRE97/622; Fig 1). SpIRE97/622 is 10 times more prevalent than previously identified SpIREs and comprises about 1.8% of the annotated full-length L1 retrotransposition events accumulated during the past approximately 27 million years (MY) (S1 Fig).
Numerous studies have demonstrated that L1 ORF1p and L1 ORF2p exhibit Cis-preference and preferentially bind to their encoding mRNA to promote its retrotransposition (Fig 6A) [33,35,38,50–53]. Using a cell culture based retrotransposition assay in HeLa cells, we demonstrated that L1 mRNAs that contain intra-5′UTR splicing events can produce ORF1p and ORF2p and undergo an initial round of retrotransposition in cis (Fig 6B). However, the resultant SpIREs lack Cis-acting sequences required for efficient L1 transcription (Fig 2) [47,94,96] and, as a result, are compromised for subsequent rounds of retrotransposition (Figs 4A and 6B).
In contrast to intra-5′UTR splicing events, L1 mRNAs containing 5′UTR/ORF1 splicing events produce nonfunctional, amino-terminal truncated versions of ORF1p (Fig 3C; S3A and S3B Fig). As a result, these mRNAs are retrotransposition defective in cis and must rely on exogenous sources of ORF1p to promote their retrotransposition by Trans-complementation (Figs 4B and 4C and 6C). Notably, these experiments also provide genetic evidence that ORF2p can be translated from the 5′UTR/ORF1 spliced L1 mRNAs. In the rare cases in which Trans-complementation occurs, the resultant 5′UTR/ORF1 SpIRE will lack Cis-acting sequences required for efficient L1 transcription and, if transcribed, would produce nonfunctional versions of ORF1p. The loss of Cis-acting sequences and the requirement for Trans-complementation make it highly unlikely that the resultant 5′UTR/ORF1 SpIREs could undergo subsequent rounds of retrotransposition (Fig 6C).
The above data strongly indicate that SpIREs represent evolutionary “dead ends” in the L1 amplification process. It is possible that a small number of SpIREs could give rise to new L1 retrotransposition events. For example, the insertion of a SpIRE97/622 downstream of a cellular promoter could, in principle, enhance its expression and subsequent retrotransposition. However, any resultant retrotransposition event would contain a defective promoter and ultimately be compromised for subsequent rounds of retrotransposition. Thus, we conclude that splicing negatively affects L1 retrotransposition.
The SpIREs examined in this study each use a common SD site (G98U99) but different SA sites (A620G621, A788G789, A851G852, or A974G975) [78,79]. These findings raise the following question: if splicing adversely affects L1 retrotransposition, why are these splice sites retained in L1 RNA? The G98U99 SD site is about 46 MY old, is conserved in the L1PA1–L1PA10 subfamilies (S1B Fig) [84], and resides within a core binding site for the RUNX3 transcription factor [96]. Indeed, previous studies indicated that mutating U99 in the L1 5′UTR impairs RUNX3 binding and decreases 5′UTR transcriptional activity [96]. Consistent with these findings, we found that mutating the SD sequence leads to an approximately 5-fold reduction in L1 steady-state RNA levels (Fig 2B). Together, these data strongly suggest that the benefit conferred by the RUNX3 transcription factor binding site at the DNA level outweighs the cost of harboring the SD site (G98U99) in L1 RNA.
Despite the evolutionary conservation of the G98U99 SD, northern blotting experiments revealed that the vast majority of L1 5′UTRs are not subject to splicing (Fig 2B). SpIREs are therefore most likely formed when L1 RNAs containing rare splicing events undergo retrotransposition. The reason(s) G98U99 is not efficiently utilized as a functional SD site requires elucidation. However, it is possible that the G98U99 sequence is sequestered into a secondary structure within L1 RNA that restricts its access to U1 small nuclear RNA (snRNA) (reviewed in [107,108]). Alternatively, a cellular protein(s) might bind at or near the SD site, thereby blocking its ability to interact with U1 snRNA. Either scenario provides a plausible mechanism for how L1 maintains a functional SD sequence in its mRNA and could, in part, explain why SpIREs only represent about 2% of annotated full-length L1 retrotransposition events that occurred during the past approximately 27 MY.
SA sites within the 5′UTR might also reside in functional transcription factor binding sites or functionally conserved regions of ORF1p. For example, the A788G789 SA is about 70 MY old and is conserved through the L1PA15B subfamily (S1B Fig), suggesting that it may reside in a conserved Cis-acting motif. The ORF1 A974G975 SA resides at codon positions two and three of lysine 22, and any nucleotide change at codon position two would result in an amino acid substitution in ORF1p that may adversely affect its function. Thus, it is possible that some functional splice sites are embedded in sequences that are critical for 5′UTR and/or ORF1p function.
Our data reveal how host-factor—driven L1 5′UTR evolution can alter L1 splicing dynamics. We demonstrated that structural changes in the 5′UTR can lead to collateral intra-5′UTR splicing changes, which have resulted in the generation of new SpIRE97/790 sequences (Fig 5A and 5D). In addition to yielding insights into the evolution of human L1 5′UTR sequences, these experiments demonstrate the utility of our luciferase-based reporter constructs to prospectively detect ancestral L1 splicing events that led to the generation of an older SpIRE (SpIRE97/853) subfamily (Fig 5A and 5D; S1 Data; S1 Table). Although the SpIRE97/622 sequence is the most abundant SpIRE in the HGR, only SpIRE97/790 sequences were detected in our RT-PCR experiments. These data, as well as the finding that five of eight SpIRE97/790 sequences are polymorphic with respect to presence/absence in the human population, suggest that SpIRE97/790 sequences are currently amplifying in modern human genomes.
It is noteworthy that the splicing events detected from engineered L1 mRNAs in transfected HeLa cells recapitulate many splicing events that led to SpIRE formation in the human genome (Figs 2D and 5D, and [78,79]). It has recently been shown that a small number of distinct genomic L1 loci are expressed in a cell type—specific manner [6–8]. Moreover, L1 splicing and/or premature polyadenylation patterns vary among human tissues and cell types [79,80,109,110], host proteins involved in splicing and polyadenylation associate with L1 RNPs [100,111–113], and overexpression of the Epstein-Barr Virus SM protein alters L1 splicing and premature polyadenylation patterns [79]. Thus, it is tempting to speculate that L1 posttranscriptional processing may suppress expression and/or retrotransposition of full-length L1s in a developmental or cell type—specific manner.
In sum, our data strongly indicate that L1 mRNA splicing is detrimental to L1 retrotransposition and further strengthen the hypothesis that ORF1p and ORF2p predominantly retrotranspose their encoding full-length L1 RNAs to new genomic locations in cis. In addition, we demonstrated that despite harboring evolutionarily conserved functional SD and SA sites within their 5′UTR, the vast majority of L1 transcripts apparently evade splicing. Finally, we provide experimental evidence revealing that changes within the L1 5′UTR, which are driven by the escape from host-factor repression, lead to collateral changes in L1 splicing profiles. Together, these data provide insights into the evolutionary dynamics of the L1 5′UTR and raise the intriguing possibility that host factors that promote L1 splicing or alter L1 splicing profiles may represent a mechanism by which the cell can disrupt full-length L1 RNA to prevent unabated L1 retrotransposition.
All plasmids were propagated in DH5α Escherichia coli (genotype: F- φ80lacZΔM15 Δ(lacZYA-argF) U169 recA1 endA1 hsdR17 (rk-, mk+) phoA supE44 λ- thi-1 gyrA96 relA1) (Invitrogen, Carlsbad, CA). Competent cells were generated as described previously [114]. Plasmids were prepared using the Plasmid Midi Kit (Qiagen, Germany) according to the protocol provided by the manufacturer.
HeLa-JVM cells (obtained from Dr. Maxine Singer and originally cited in reference [31]) were cultured in high glucose Dulbecco’s Modified Eagle Medium (DMEM) lacking pyruvate (Invitrogen). DMEM was supplemented with 10% fetal bovine calf serum (FBS) and 1X penicillin/streptomycin/glutamine to create DMEM-complete medium, as described previously [31]. HeLa-JVM cells were grown in a humidified tissue culture incubator (Thermo Scientific, Waltham, MA) at 37°C in the presence of 7% CO2.
BLAT [83] was used to screen build 38 (GRCh38/hg38) of the UCSC genome browser (https://genome.ucsc.edu) using 100 bp in silico probes that spanned (50 bases upstream and downstream) the splice junctions of SpIRE97/622, SpIRE97/790, and SpIRE97/976. The in silico probes were designed using the L1.3 sequence (accession #L19088 [9,82]) as a template. A 100-bp in silico probe that spanned (50 bases upstream and downstream) the splice junction of SpIRE97/853 was designed using the pJBMWT129PA4LUC sequence. Putative SpIREs shared >95% sequence identity with the in silico probes.
Putative SpIREs were downloaded from the UCSC genome browser and manually curated with the aid of repeat masker (http://repeatmasker.org). Each sequence was inspected to ensure it contained a splicing event and represented a bona fide SpIRE. For four events that were prematurely 3′ truncated, we analyzed 4 kb of genomic DNA flanking the 3′ end of the SpIRE to determine if it shared >95% sequence identity with L1.3 using the Serial Cloner alignment tool (http://serialbasics.free.fr/Serial_Cloner.html). We were unable to identify any L1 sequence in the flanking DNA; thus, we cannot determine the reason for the apparent 3′ truncation in these four SpIREs. Structural hallmarks of L1 integration events that occur by canonical TPRT (e.g., the presence of target site duplications, the presence of untemplated nucleotides at the 5′ genomic DNA/L1 junction [47,69,89,115], a 3′ poly(A) tract, and putative L1-mediated sequence transductions) [23,116,117] were determined manually by analyzing sequences flanking the 5′ and 3′ ends of each SpIRE [69,116,118]. The L1 “empty site” for all SpIREs is inferred; the 3′ TSD was considered the ancestral “empty site” and any nucleotide differences between the 5′ and 3′ TSD are annotated in the 5′ TSD only. Sequences are named based on the splicing event contained within the SpIRE (SpIRE97/622, SpIRE97/790, SpIRE97/976, or SpIRE97/853) and a corresponding number for easy referral between S1 Data and S1 Table (for example; SpIRE97/622-1 is the first of the analyzed 116 SpIRE97/622 sequences).
Khan et al. 2006 provided full-length L1 subfamily consensus sequences of L1PA1 (L1Hs) through L1PA16 and assembled an alignment of the respective 5′UTRs [84]. We manually inspected these alignments to determine the oldest L1 subfamily that contained the 5′UTR SD/SA sequences utilized in generating the reported SpIREs. We next determined the conservation of the ORF1 SA sequence (A974G975) by aligning full-length L1 consensus sequences provided in Khan et al. 2006 using the ClustalW alignment function [84,119] from the MegAlign (http://www.dnastar.com/t-megalign.aspx) software. As with the 5′UTR, we manually inspected the resulting alignment to determine the oldest L1 subfamily that contained the ORF1 SA sequence (A974G975).
To identify putative splicing branch point sequences, we utilized the L1.3 5′UTR (accession #L19088) sequence and the pJBMWT129PA4LUC 5′UTR sequence and submitted them for analysis using the Human Splicing Finder v3.0 online prediction program (http://www.umd.be/HSF3/HSF.html) [105]. The resultant analyses identify potential SD, SA, and branch point sequences and assign consensus value scores for each motif [105]. Motif scores greater than 80 represent “strong” splice sites; sequences with scores less than 80 represent “weaker” splice sites. The 5′UTR sequence of each L1 was uploaded and analyzed by the general “Analyze a Sequence” function. We then selected predicted branch points that might pair with the known SA: A788G789 (L1.3) and A851G852 (pJBMWT129PA4LUC) based on their proximity to the SA sequence [120]. We identified a putative branch point (A763C764C765T766C767A768C769) with a score of 95.75 that could pair with the SA: A788G789 in the L1.3 5′UTR. We also identified a putative branch point (T795C796C797A798G799A800G801) with a score of 75.73 that could pair with the SA: A851G852 in the pJBMWT129PA4LUC 5′UTR (Fig 5A).
We performed in silico genotyping of four SpIRE97/790 loci using reads from the 1000 Genomes Project [103,121]. Read pairs anchored within 600 bp of each locus were extracted from each of 2,453 samples from the 1000 Genomes Project. Extracted read pairs were aligned to reconstructed reference (insertion) and alternative (empty site) sequences and the most likely genotype for each sample was determined based on the number and mapping quality of read pairs aligned to each allele [121]. Read pairs that aligned entirely within the L1 sequence as well as read pairs that show equivalent alignments to both the reference and alternative sequences were ignored in the analysis.
We utilized an anchored read pair mapping approach to identify additional non-reference SpIRE97/790 insertions in the 1000 Genomes Project samples. We searched alignment files from 2,453 samples for read pairs in which one read is aligned across the splice junction in one of the four SpIRE97/790 sequences represented in the genome reference sequence and the other read is uniquely aligned elsewhere in the genome. We then intersected the resulting anchored locations with a recently published map of non-reference L1 insertions discovered in the same samples [122], identifying four insertions supported by multiple SpIRE-associated read pairs. To further characterize these loci, we extracted insertion-supporting read pairs for each locus and performed a de novo read assembly using the CAP3 assembler [123]. This analysis results in a collection of short contigs for each locus, which extend into the flanking edges of each inserted L1 element. The resulting contigs were filtered for repeat content, aligned to the genome reference, and annotated for characteristics indicative of bona fide SpIRE97/790 insertions (S1 Data and S1 Table).
The following L1 constructs contain a derivative of an RC-L1 (L1.3, accession #L19088 [9,82]) cloned into the pCEP4 plasmid backbone (Life Technologies), unless indicated otherwise. Cloning strategies used to create these constructs are available upon request.
pJM101/L1.3 contains a full-length version of L1.3 in the pCEP4 backbone. The 3′UTR of L1.3 contains the mneoI retrotransposition indicator cassette [9,31,82].
pJM101/L1.3ΔCMV is identical to pJM101/L1.3, but the CMV promoter was deleted from the pCEP4 plasmid [9,31,82].
pJM101/L1.3NN is a derivative of pJM101/L1.3 that lacks the mneoI retrotransposition indicator cassette [50].
pDK101/L1.3 is a derivative of pJM101/L1.3 that expresses a version of ORF1p that contains a T7 gene10 epitope tag on its carboxyl-terminus [35].
pJM105/L1.3 is identical to pJM101/L1.3, but contains a D702A missense mutation in the ORF2p RT active site [50].
pJM105NN is a derivative of pJM105/L1.3 that lacks the mneoI retrotransposition indicator cassette [50].
pJM102/L1.3 is a derivative of pJM101/L1.3 that lacks the L1 5′UTR [58].
pJM102/L1.3ΔCMV is identical to pJM102/L1.3, but the CMV promoter was deleted from the pCEP4 plasmid [50].
pPL97/622/L1.3 is a derivative of pJM101/L1.3 that contains a 524 intra-5′UTR deletion (L1.3 nucleotides 98–621) present in SpIRE97/622 [10].
pPL97/622/L1.3ΔCMV is identical to pPL97-622/L1.3, but the CMV promoter was deleted from the pCEP4 plasmid.
pPL97/976/L1.3 is a derivative of pJM101/L1.3 that contains an 878-bp 5′UTR/ORF1 deletion (L1.3 nucleotides 98–975) present in SpIRE97/976.
pPL97/976/L1.3-T7 is a derivative of pPL97-976/L1.3 that expresses a version of ORF1p that contains a T7 gene10 epitope tag on its carboxyl-terminus.
pORF2/L1.3NN is a monocistronic L1 ORF2 expression plasmid that lacks the mneoI retrotransposition indicator cassette [22].
pJBM561 is a monocistronic L1 ORF1 expression plasmid that lacks the mneoI retrotransposition indicator cassette.
pCEP/GFP is a pCEP4-based plasmid that expresses a humanized Renilla green fluorescent protein (hrGFP) from phrGFP-C (Stratagene). A CMV promoter drives the expression of the hrGFP gene [22].
The following L1 constructs contain a derivative of an RC-L1 (L1RP, accession #AF148856.1 [124]) cloned into the pCEP4 plasmid backbone (Life Technologies) lacking the CMV promoter.
pL1RP-EGFP contains a full-length version of L1RP element. The 3′UTR contains the EGFP retrotransposition indicator cassette [124].
pL1RP(JM111)-EGFP: a derivative of pL1RP-EGFP that contains two missense mutations in ORF1 that abolish retrotransposition [31,124].
L1Hs+129L1PA4: a derivative of pL1RP-EGFP that carries a 129-bp sequence element from the L1PA4 5′UTR that is not present in L1Hs [104].
L1Hs+129scrambleL1PA4: a derivative of pL1RP-EGFP that carries a scrambled version of the 129-bp sequence element from the L1PA4 5′UTR that is not present in L1Hs [104].
The following plasmids are based on the pGL4.11 promoter-less firefly luciferase expression vector (Promega, Madison, WI). Oligonucleotides and cloning strategies used to create these constructs are available upon request.
pPLWTLUC is a derivative of pGL4.11 that contains the WT L1.3 5′UTR upstream of the firefly luciferase reporter gene.
pPL97/622LUC is a derivative of pGL4.11 that contains the pPL97-622/L1.3 5′UTR deletion derivative upstream of the firefly luciferase reporter gene.
pPLSDmLUC is a derivative of pPLWTLUC that contains a U99C SD mutation in the L1.3 5′UTR upstream of the firefly luciferase reporter gene.
pPLSAmLUC is a derivative of pPLWTLUC that contains an A620C SA mutation in the L1.3 5′UTR upstream of the firefly luciferase reporter gene.
pRL-TK is an expression plasmid where the HSV-TK promoter drives Renilla luciferase transcription (Promega).
pJBMWTLUC is a derivative of pGL4.11 that contains the L1RP 5′UTR from the plasmid pL1RP-EGFP [124] and was cloned upstream of the firefly luciferase reporter gene.
pJBMWT129PA4LUC is a derivative of pGL4.11 that contains the 5′UTR from L1Hs+129L1PA4 [104] and was cloned upstream of the firefly luciferase reporter gene.
pJBMWT129SCRLUC: the 5′UTR from L1Hs+129scrambleL1PA4 [104] that was cloned upstream of the firefly luciferase reporter gene.
RNA isolation was performed as previously described with minor modifications [100]. Briefly, 8×106 HeLa-JVM cells were seeded into a T-175 Falcon tissue culture flask (BD Biosciences, San Jose, CA). On the following day, transfections were conducted using the FuGene HD transfection reagent (Promega, Madison, WI). The transfection reactions contained 1 mL of Opti-MEM (Life Technologies), 120 μl of the FuGene HD transfection reagent, and 20 μg of plasmid DNA per flask. The tissue culture medium was changed 24 hours post-transfection. The cells were collected 48 hours post-transfection. Briefly, cells were washed in ice-cold 1X phosphate buffered saline (PBS) (Life Technologies). The cells then were scraped from the tissue culture flasks, transferred to a 15-mL conical tube (BD Biosciences), and centrifuged at 3,000 × g for 5 minutes at 4°C. Cell pellets were frozen at −20°C overnight. The frozen pellets were thawed and total RNA was prepared using the TRIzol reagent following the protocol provided by the manufacturer (Life Technologies). Poly(A) RNAs then were isolated from the total RNAs using a Oligotex mRNA Midi Kit (Qiagen), suspended in UltraPure DNase/RNase-Free distilled water (Thermo Fisher Scientific, Waltham, MA), and quantified using a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific). For the RT-PCR experiments in Fig 5D, total RNA was collected using an RNeasy kit (Qiagen), and polyadenylated RNA was isolated from total RNA using Dynabeads Oligo (dT)25 (Ambion).
Northern blot experiments were performed as previously described [100]. Briefly, Northern blot experiments were conducted using the NorthernMax-Gly Kit (Thermo Fisher Scientific) following the protocol provided by the manufacturer. Briefly, aliquots of poly(A) RNAs (2 μg) were incubated for 30 minutes at 50°C in Glyoxal Load Dye (containing DMSO and ethidium bromide) and then were separated on a 1.2% agarose gel. The RNAs were transferred by capillary action to a Hybond-N nylon membrane (GE Healthcare, Marlborough, MA) for 4 hours and cross-linked to the membrane using the Optimum Crosslink setting of a Stratalinker (Stratagene, La Jolla, CA). Membranes were then baked at 80°C for 15 minutes. Membranes were prehybridized for approximately 4 hours at 68°C in NorthernMax Prehybridization/Hybridization Buffer (Thermo Fisher Scientific) and then were incubated overnight at 68°C with a strand-specific RNA probe (final concentration of probe, 3×106 cpm/ml). The following day, the membranes were washed once with low stringency wash solution (2x saline sodium citrate (SSC), 0.1% sodium dodecyl sulfate [SDS]) and then twice with high stringency wash solution (0.1x SSC, 0.1% SDS). The washed membranes were placed in a film cassette (Thermo Fisher Scientific, Autoradiography Cassette FBCA 57) and exposed to Amersham Hyperfilm ECL (GE Healthcare) overnight at −80°C. Films were developed using a JP-33 X-Ray Processor (JPI America Inc., New York, NY).
Northern blot probes were prepared as previously described [100]. Strand-specific αP32-UTP radiolabeled riboprobes were generated using the MAXIscript T3 system (Thermo Fisher Scientific). Briefly, oligonucleotide primers were used to PCR amplify portions of the L1.3 5′UTR [100] (L1.3 nucleotides 7–99 or L1.3 nucleotides 103–336) or the 3′ end of the luciferase gene (see below). The resultant PCR products were separated on a 1% agarose gel and were purified using QIAquick gel extraction (Qiagen). The labeling reaction was carried out at 37°C using the following reaction conditions: 500 ng of gel purified DNA template, 2 μL of transcription buffer supplied by the manufacturer, 1 μL each of unlabeled 10 mM ATP, CTP, and GTP, 5 μL of αP32-UTP (10 mCi/mL), and 2 μL of T3 RNA polymerase. The reaction components then were mixed and brought to a total volume of 20 μL using nuclease-free water in a 1.5-mL Eppendorf tube, which was incubated at 37°C for 10 minutes in a heating block. Unincorporated nucleotides were subsequently depleted using the Ambion NucAway Spin Columns (Thermo Fisher Scientific) following the protocol provided by the manufacturer. To generate a control β-actin riboprobe, the pTRI-β-actin-125-Human Antisense Control Template (Applied Biosystems) was used in T3 labeling reactions. Biological triplicates of each northern blot exhibited similar results.
Oligonucleotide sequences were used to generate northern blot probes. A T3 RNA polymerase promoter sequence was included on the antisense (AS) primer used to generate the antisense riboprobe (underlined below):
L1.3 5′UTR 7–99 Sense: 5′-GGAGCCAAGATGGCCGAATAGGAACAGCT-3′
L1.3 5′UTR 7–99 AS: 5′-AATTAACCCTCAAAGGGACCTCAGATGGAAATGCAG-3′
L1.3 5′UTR 103–336 Sense: 5′-GGGTTCATCTCACTAGGGAGTG-3′
L1.3 5′UTR 103–336 AS: 5′-AATTAACCCTCACTAAAGGGTATAGTCTCGTGGTGCGCCG-3′
Luciferase 3′ FFLuc Sense: 5′-GGCAAGATCGCCGTGAATTCTCAC-3′
Luciferase 3′ FFLuc AS: 5′-AATTAACCCTCACTAAAGGGCCTGGCGCTGGCGCAAGCAGC-3′
Northern blot bands were quantified using the ImageJ software (https://imagej.nih.gov/ij/download.html) [125]. The intensity of the bands in the pPLWTLUC and pPLSDmLUC lanes were determined and normalized to the actin loading control. Three independent northern blots were subject to quantification. We then computed that average intensity of the bands and calculated a standard deviation. As reported in the text (Fig 2B), the steady-state level of pPLSDmLUC mRNA is about 18% the level of pPLWTLUC mRNA with a standard deviation of ±3.1%.
Luciferase assays were performed using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI) following the manufacturers protocol. Briefly, 2×104 HeLa cells were plated into each well of a 6-well plate (BD Biosciences). Approximately 24 hours later, each well was transfected using a transfection mixture of 100 μl Opti-MEM (Life Technologies), 3 μl of FuGENE6 transfection reagent (Promega), and 1 μg plasmid DNA (0.5 μg of a firefly luciferase test plasmid and 0.5 μg of an internal control Renilla luciferase expression). Each transfection was performed as a technical duplicate (i.e., in two wells of a 6-well tissue culture plate). Approximately 24 hours post-transfection, the transfected cells were washed once with ice-cold 1X PBS and the cells in each well were subjected to lysis for 15 minutes at room temperature using 500 μl of the 1X Passive Lysis Buffer supplied by the manufacturer. Following homogenization of the lysate by manual pipetting, 60 μl of the lysate from each well of the 6-well tissue culture plate was distributed equally in 3 wells of a 96-well white opaque, optically transparent top plate (BD Biosciences), allowing six luminescence readings for each transfection condition (six technical replicates—3 readings per well of a 6-well plate). The 96-well plate then was subject to luciferase detection assays using a GloMax-Multi Detection System (Promega) following the manufacturer’s protocol. Luminescence readings from the six technical replicates were averaged to give a single normalized luminescence reading (NLR). This assay then was performed in biological triplicate, yielding three independent NLRs. The resultant data were subsequently analyzed using a Student one-tailed t test. Error bars indicate the standard deviation. Luminescence readings from lysis buffer alone and from lysates derived from untransfected cells were included used as negative controls.
Poly(A) selected mRNA from transfected HeLa-JVM cells in a T-175 tissue culture flask was collected as previously described for northern blots. The resultant mRNAs were subjected to targeted RT-PCR using SuperScript III One-Step RT-PCR System, with Platinum Taq DNA Polymerase (Thermo Fisher Scientific), following the manufacturer’s protocol. The REVLUC primer was used to synthesize first-strand cDNA. The FWD5′UTR and REVLUC primers then were used to amplify the resultant cDNAs (see sequences below). For RT-PCR experiments in Fig 5D, cDNA was synthesized from polyadenylated RNA with a SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen) using the REVLUC primer. The resultant cDNA was then subjected to PCR using the FWD5′UTR and REVLUC primers and Platinum Taq DNA polymerase (Invitrogen) (30 cycles; annealing temp: 54°C; 1-minute extension). The RT-PCR products were separated on a 2.0% agarose gel, excised from the gel using QIAquick gel extraction (Qiagen), and cloned using the TOPO TA Cloning Kit (Thermo Fisher Scientific). Sanger DNA sequencing performed at the University of Michigan DNA Sequencing Core verified the cDNA sequences in the resultant plasmids. Biological triplicates of this experiment yielded similar results.
The following oligonucleotide sequences were used in the RT-PCR experiments:
FWD5′UTR: 5′-GGAACAGCTCCGGTCTACAGCTCCC-3′
REVLUC′ 5′-CCCTTCTTAATGTTTTTGGCATCTTCC-3′
The plating and transfection of HeLa-JVM cells in T-175 tissue culture flasks was performed as detailed above in the mRNA isolation section except that HeLa-JVM cells were subjected to selection in DMEM-complete medium supplemented with 200 μg/ml of hygromycin B (Thermo Fisher Scientific) 48 hours post-transfection and the selection medium was changed every other day for 7 days. The hygromycin resistant HeLa-JVM cells were harvested 9 days post-transfection as described in the mRNA isolation section. The cell pellets were frozen at −80°C overnight. The following day, pellets were lysed for 15 minutes on ice by incubation in 0.5 mL of lysis buffer: 10% glycerol, 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.1% NP-40 (IGPAL) (Sigma-Aldrich, St. Louis, MO), and 1X Complete Mini EDTA-free Protease Inhibitor Cocktail (Roche Applied Science, Germany). The resultant protein lysates then were centrifuged at 15,000 × g for 30 minutes to clear the lysate. The resultant supernatant (approximately 0.4 mL) was designated as the WCL. Alternatively, the supernatant fraction was subject to RNP collection, as previously described [35]. Briefly, 200 μL of the WCL was layered onto a sucrose solution cushion (6 mL of 17% sucrose, bottom layer, followed by 4 mL of 8.5% sucrose, top layer, overlaid by 200 μL of the WCL) and ultracentrifuged at 178,000 × g for 2 hours at 4°C. Following ultracentrifugation, the supernatant was aspirated and the resultant RNP pellet was suspended in 100 μL of water supplemented with 1X Complete Mini EDTA-free Protease Inhibitor Cocktail (Roche Applied Science). Bradford assays (Bio-Rad Laboratories, Hercules, CA) were used to determine protein concentrations. WCLs generally yielded 15–19 μg/μL of protein. RNP preparations yielded 6–10 μg/μL of protein. The protein samples were stored at −80°C.
Western blot experiments were performed as previously described, with minor modifications [100]. Briefly, protein samples were collected as described above and then were incubated with a 2X solution of NuPAGE reducing buffer (containing 1.75%–3.25% lithium dodecyl sulfate and 50 mM dithiothreitol [DTT]) (ThermoFisher Scientific). An aliquot (20 μg) of the reduced proteins were incubated at 100°C for 10 minutes and then were separated by electrophoresis on 10% precast mini-PROTEAN TGX gels (Bio-Rad Laboratories, Hercules, CA) run at 200 V for 1 hour in 1X Tris/Glycine/SDS (25 mM Tris-HCL, 192 mM glycine, 0.1% SDS, pH 8.3) buffer (Bio-Rad Laboratories). Transfer was performed using the Trans-Blot Turbo Mini PVDF Transfer Packs (BioRad Laboratories) with the Trans-Blot Turbo Transfer System (BioRad Laboratories) at 25 V for 7 minutes. The resultant membranes then were cut at the 75-kDa marker using the Precision Plus Protein Kaleidoscope marker (Bio-Rad Laboratories) as a guide. The membranes then were incubated at room temperature in blocking solution (containing 1X PBS and 5% dry low-fat milk) (Kroger, Cincinnati, OH). The eIF3 antibody (Santa Cruz Biotechnology Inc. [SC-28858]) was used at a 1:1,000 dilution to probe membranes for eIF3 at 110 kDa as a loading control. The α-N-ORF1p [100] antibody (directed against ORF1p amino acids 31–49; EQSWMENDFDELREEGFRR), α-C-ORF1p (directed against ORF1p amino acids 319–338; EALNMERNNRYQPLQNHAKM), and anti-T7 (Merck Millipore 69048 T7•Tag Antibody HRP Conjugate) antibodies were used at 1:10,000, 1:2,000, and 1:5,000 dilutions, respectively, to probe membranes for ORF1p. Antibody hybridizations were carried out overnight at 4°C in blocking solution. The blots were washed three times with 1X PBS, 0.1% Tween-20 (Sigma Aldrich) and then were incubated with a 1:5,000 dilution of secondary Amersham ECL HRP Conjugated Donkey anti-rabbit IgG Antibodies (GE Healthcare Life Sciences) for 60 minutes at room temperature blocking solution. The membranes were washed three times with 1X PBS, 0.1% Tween-20 (Sigma Aldrich). The signals then were visualized using the SuperSignal West Pico Chemiluminescent Substrate reagent (ThermoFisher Scientific) according to the protocol provided by the manufacturer. The membranes were exposed to Amersham Hyperfilm ECL (GE Healtchare) for a time that spanned 5 seconds to 5 minutes and were developed using a JP-33 X-Ray Processor (JPI America Inc.).
|
10.1371/journal.pgen.1002698 | An Essential Role for Katanin p80 and Microtubule Severing in Male Gamete Production | Katanin is an evolutionarily conserved microtubule-severing complex implicated in multiple aspects of microtubule dynamics. Katanin consists of a p60 severing enzyme and a p80 regulatory subunit. The p80 subunit is thought to regulate complex targeting and severing activity, but its precise role remains elusive. In lower-order species, the katanin complex has been shown to modulate mitotic and female meiotic spindle dynamics and flagella development. The in vivo function of katanin p80 in mammals is unknown. Here we show that katanin p80 is essential for male fertility. Specifically, through an analysis of a mouse loss-of-function allele (the Taily line), we demonstrate that katanin p80, most likely in association with p60, has an essential role in male meiotic spindle assembly and dissolution and the removal of midbody microtubules and, thus, cytokinesis. Katanin p80 also controls the formation, function, and dissolution of a microtubule structure intimately involved in defining sperm head shaping and sperm tail formation, the manchette, and plays a role in the formation of axoneme microtubules. Perturbed katanin p80 function, as evidenced in the Taily mouse, results in male sterility characterized by decreased sperm production, sperm with abnormal head shape, and a virtual absence of progressive motility. Collectively these data demonstrate that katanin p80 serves an essential and evolutionarily conserved role in several aspects of male germ cell development.
| Microtubules are critical components of cells, acting as a “scaffold” for the movement of organelles and proteins within the cytoplasm. The control of microtubule length, number, and movement is essential for many cellular processes, including division, architecture, and migration. We have defined the role of the microtubule severing protein katanin p80 in male germ cell development. Male mice carrying a point mutation in the p80 gene are sterile as a consequence of low numbers of sperm, abnormal sperm morphology, and poor motility (ability to “swim”). We show that this mutation is associated with defects in microtubule structures involved in the division of immature sperm cells, in structures that shape the sperm head, and in the sperm tail, which is essential for sperm movement in the female reproductive tract. This study is the first to show that katanin p80, via its effects on microtubule dynamics within the testis, is required for male fertility.
| The regulation of microtubule dynamics is an essential requirement for all cells and in many aspects of their daily function. The ability to precisely regulate microtubule number, the assembly of networks, and the rate of microtubule assembly and disassembly underpins cellular processes including division, differentiation and migration. Male gamete development in particular relies upon the co-ordinated development and rapid remodelling of complex microtubule structures, such as the mitotic (spermatogonia) and meiotic (spermatocyte) spindle; flagella formation needed for sperm motility; and the manchette, which determines sperm head shape and contributes to tail structure. Approximately one in 20 men of reproductive age is sub-fertile or sterile, of which 60% of cases are due to intrinsic defects in spermatogenesis. This heterogeneous disorder manifests clinically as diminished sperm number, or abnormal motility or morphology, or commonly combinations thereof, in the ejaculate [1]. All of these clinical presentations may be underpinned by defective microtubule dynamics.
Microtubule severing is emerging as a key regulator of microtubule dynamics [2], [3], [4], [5], . The most well characterized microtubule severing enzyme is the katanin complex [8], the severing function of which is carried out by an ATPase enzymatic subunit, named p60, encoded by the Katna1 gene. Katanin p60 is a member of the AAA domain (ATPases Associated with diverse cellular Activities) protein family. Upon binding ATP, katanin p60 oligomerizes onto the tail of an individual tubulin subunit within a microtubule to form a 14–16 nm ring structure [9]. ATP hydrolysis confers a conformational change in the oligomer and ‘tugs’ upon the tail of the tubulin subunit. This leads to destabilization, and ultimately severing, of the polymer [2]. Other AAA microtubule-severing proteins include spastin and fidgetin [8]. Mutations in the gene encoding spastin cause progressive axon degeneration and underlie ∼40% of autosomal dominant cases of hereditary spastic paraplegia [10] and deletion of the fidgetin gene in mice results in a severe behavioural and developmental phenotype [11], illustrating the importance of the family in neuronal development.
The regulation and compartmentalization of microtubule severing is essential for normal cell function and survival. Katanin p60-mediated severing can be modulated by a p80 regulatory subunit, encoded by the Katnb1 gene [9]. The p80 subunit of katanin binds to p60 and targets it to the centrosome in transfected mammalian cell lines [9], [12], and generally enhances severing, but can also inhibit it depending on the cellular context [4], [12]. In Tetrahymena thermophila [13] and Caenorhabditis elegans [14], p80 null mutants phenocopy p60 null mutants. However the in vivo role of the p80 subunit in mammals remains enigmatic. Katanin was identified as a heterodimer of p60 and p80 subunits in sea urchins [8], however the ratio of the two subunits shows developmental and regional variation in rat neurons [15] and in mouse testis (the current study), suggesting that the expression of the p80 regulatory subunit may be one way in which p60-mediated severing is controlled in mammals [4]. The Katnb1 gene contains a C-terminal WD40 domain predicted to be involved in protein-protein interactions, and as such is a strong candidate for targeting p60-mediated severing to particular locations within a cell and for targeting p60-mediated severing to post-translationally modified or microtubule-associated protein (MAP)-associated tubulin polymers [16], [17]. Katanin p80 also binds to the molecular motor protein dynein [18] and to dynein-regulating proteins [18], [19] and could thus be involved in the transport of the katanin complex to specific sites.
Katanin function is evolutionarily conserved, with p60 and p80 orthologues identified in species from all 5 kingdoms, including in C. elegans, Drosophila melanogaster, Arabidopsis thaliana, Chlamydomonas reinhardtii, mice and humans. Katanin localizes to mitotic spindle poles in mammalian cell lines, where it regulates spindle structure and chromosome movement [2], [3], [20], [21]. Mutations in C. elegans orthologues of katanin p60 and p80 reveal roles for katanin in oocyte meiotic spindle assembly and chromosome movement [20], [22], and katanin regulates different microtubule populations, including kinetochore-associated bundles, to control oocyte meiotic spindle length in Xenopus [23]. Of note, katanin regulates mitotic chromosome movement in D. melanogaster by participating in the so-called “Pacman-mediated” shortening of spindle microtubule plus-ends which results in the poleward movement of the chromosomes [24]. In addition, mutations in katanin orthologues in two distantly related organisms Tetrahymena and Chlamydomonas result in the absence of the central axoneme microtubule pair and cilia/flagella defects, indicating a role in ciliogenesis [13], [25], [26]. To date there have been no in vivo models of katanin dysfunction in mammals.
Here we show that a missense mutation in the highly conserved WD40 domain of the Katnb1 gene, encoding the p80 regulatory subunit of katanin, causes male sterility in mice characterized by oligoasthenoteratozoospermia. Katnb1 mutant mice, denoted as Taily, show frequent failure of meiotic spindle resolution, defective manchette function and abnormal axoneme development. Our findings highlight the critical role for katanin p80 in the regulation of microtubules dynamics in many aspects of male germ cell development. These data raise the possibility that defective katanin function may also contribute to human infertility, specifically defective sperm number, morphology and/or sperm motility.
Mouse lines carrying ENU-induced mutations causing male sterility were identified using breeding trials as described previously [27]. Lines with G3 male sterility at a frequency of one in four, but with normal mating behaviour, were chosen for further analysis. They included the ‘Taily’ line. The causal mutation was mapped using SNP-based methods and ultimately narrowed to a linkage interval on chromosome 8 (between SNP markers rs3089148 and rs3710112) containing 74 genes. Candidate genes were chosen on the basis of testis expression and proposed function. The protein-coding region and intron-exon boundaries of 30 genes were sequenced. The causative mutation in Taily mice was identified as a recessive G to T substitution in exon 9 of the Katnb1 gene. No other mutations were found. Unaffected males possessed either homozygous wild type alleles (Katnb1WT/WT) or were heterozygous for the wild type and Taily allele (Katnb1WT/Taily). Greater than 50 mice of each genotype were assessed and the genotype-phenotype correlation was absolute. The Taily mutation resulted in the conversion of a valine (GTC) to a phenylalanine (TTC) in the WD40 repeat region of the katanin p80 protein (Figure 1). The presence of an aliphatic amino acid (e.g. V or I) at amino 234, relative to the mouse sequence, is absolutely conserved across all species, and is strongly suggestive of a functionally important role (Figure 1B). Western blot analysis revealed that haploid germ cells from Katnb1Taily/Taily mice contained markedly less p80 protein than those from Katnb1WT/WT mice (Figure 2), demonstrating that the ‘Taily’ allele likely results in a loss-of-function.
Katnb1Taily/Taily males showed no overt behavioural abnormalities, were morphologically identical to wild type littermates, were of normal weight (Figure S1), but were uniformly sterile when mated with wild type females (n≥10, Katnb1Taily/Taily males aged ≥8 weeks of age). Katnb1Taily/Taily females had apparently normal fertility.
Testes from adult (8–12 weeks) Katnb1Taily/Taily mice were 18.7% smaller than those from wild type littermates (p<0.0001, Figure 3A). Seminiferous tubules contained all germ cells types. Two major discordant features were apparent within the seminiferous epithelium from Katnb1Taily/Taily mice: 1) abnormally shaped spermatid heads (Figure 3C and 3D) and 2) abnormal meiotic cells at metaphase-anaphase (Figure 3E and 3F). Stereological analysis revealed that the number of Sertoli cells, spermatogonia and spermatocytes per testis was not different between genotypes (Figure 3B and Table S1), indicating that the initiation of spermatogenesis and entry into meiosis were unaffected by the Taily mutation. The latter finding suggested that the function of the Sertoli cell blood-testis-barrier was normal, a proposition supported by the appearance of normal inter-Sertoli cell junctions by electron microscopy (not shown). By contrast, testes from Katnb1Taily/Taily mice contained ∼30% fewer spermatids (round and elongating) compared to wild type (Figure 3B). The reduction in spermatid populations was due to a decrease in the number of cells exiting meiosis, specifically during the final meiotic division in stage XII (Figure 3B and Table S1). TUNEL-labelling revealed that apoptotic cells were predominantly present in stage XII and stage I tubules which is when the final events of meiosis occur (Figure 3G). Apoptotic spermatocytes in the process of meiotic division were observed (Figure 3H). Collectively these results indicate that katanin p80 function is required for the final phases of male meiotic cell division.
Stereology also showed that additional germ cells were not lost as they progressed through spermiogenesis (Table S1), however there was a 36 fold increase in the number of spermatozoa being phagocytosed by Sertoli cells (Figure 3B) in stages IX–XI tubules (Figure 3I and 3J). These data indicate a failure in spermiation, the process by which sperm are released by the Sertoli cell at the end of their development, prior to their passage to the epididymis [28]. The data show that a significant proportion of spermatozoa failed to be released from the Sertoli cell and were instead phagocytosed, thus leading to a reduced number of sperm entering the epididymis. As a consequence, the epididymides from Katnb1Taily/Taily males contained a lower total number of sperm than would be anticipated from the testicular daily sperm output, i.e. 11% of wild type in the epididymis compared to 57% in the testis (Figure 4A). Of the sperm found in the cauda epididymis of Taily mice, when compared to wildtype (Figure 3K), all had abnormally shaped heads (Figure 3L) and displayed compromised total motility as assessed by computer assisted sperm analysis (80.3% in Katnb1WT/WT versus 37.2% in Katnb1Taily/Taily) (Figure 4B). Very few sperm were capable of forward (progressive) motility (52.4% in Katnb1WT/WT versus 11.1% in Katnb1Taily/Tail) (Figure 4B).
Collectively these data indicate that katanin p80 has a role in germ cell exit from meiosis, in the establishment of structures or pathways within the sperm tail involved in motility and in the shaping of the sperm head. Katnb1Taily/Taily males were sterile as a consequence of decreased sperm production, abnormal sperm morphology and sperm being unable to ascend the female reproductive tract following mating. The analogous human phenotype is referred to as oligoasthenoteratozoospermia (low sperm count, poor motility and abnormal shape).
Both katanin p80 and p60 mRNAs were expressed in the testis during the post-natal establishment of the spermatogenic cycle (Figure 5). The katanin p60 microtubule severing enzyme was expressed at relatively similar levels in all ages examined, suggesting expression in multiple cell types (Figure 5A). Katanin p80 regulatory subunit expression, however, peaked at day 30, suggesting predominant expression in post-meiotic haploid spermatids (Figure 5A). These data are consistent with previous microarray data (germonline.org) indicating that the katanin p60 catalytic subunit is expressed in Sertoli cells and germ cells to a similar degree, whereas the p80 regulatory subunit while detectable in Sertoli cells and spermatogonia, is more highly expressed (5-fold) in spermatocytes and spermatids [29]. In accordance with the mRNA data, katanin p80 protein was most strongly localized within round through to elongating spermatids (Figure 5B). Katanin p60 was also prominent in spermatids and was visible in Sertoli cells (Figure 5B).
In addition, the katanin p60 orthologues Katnal1 (p60-like 1) and Katnal2 (p60-like 2) were also expressed within the developing post-natal testis with a timing similar to that observed for katanin p80 (Figure S2). Absence of KATNAL1 immunolocalization in germ cells (Smith et al, submitted for publication) and immunolocalization of KATNAL2 predominantly to the sperm tail and cytoplasm, but not associated with the sperm head, (Figure S2), suggests that the meiotic and spermatid head-shape phenotypes reported herein were primarily mediated by katanin p80 regulation of the eponymous p60 subunit.
The loss of germ cells during meiotic division prompted an investigation of the microtubule-based meiotic spindle in Katnb1Taily/Taily males. Compared to Katnb1WT/WT littermates, all Katnb1Taily/Taily meiotic spindles were abnormal (Figure 6A and 6B). Specifically, metaphase spindles appeared to be more densely populated with microtubules, and projected from the poles at a wider angle than those from wild type animals (Figure 6A and 6B and Videos S1 and S2). Pole-to-pole measurements in metaphase spindles were longer in Katnb1Taily/Taily compared to Katnb1WT/WT (12.15±0.16, n = 58, versus 10.52±0.17 µm, n = 78, mean ± SEM, p<0.0001).
Within metaphase and anaphase cells, p60 (Figure 6C and Videos S1 and S2) and p80 (Figure 6A and 6B) proteins localized to microtubules of meiotic spindles. The Katnb1Taily/Taily mutation did not overtly alter this localization. Both p60 and p80 were observed along the microtubules of the spindle and at the microtubule-chromosome interface. The latter localization is consistent with katanin involvement in the poleward movement of chromosomes in Drosophila mitotic cells [24]. Specifically within Drosophila, katanin is believed to participate in the depolymerization of microtubule plus-ends in the midzone at anaphase during Drosophila mitosis, effectively “chewing away” the microtubule ends to facilitate spindle shortening via a process known as “Pacman” [24]. A role for katanin p80 in the Pacman-mediated poleward movement of chromosomes in mammalian meiotic anaphase is further supported by the appearance of multiple cells stalled in late anaphase in Katnb1Taily/Taily mice (Figure 3F).
Disordered meiosis is further evidenced by the frequent occurrence of binucleated haploid spermatids in Katnb1Taily/Taily testis sections (Figure 6D). Binucleated spermatids were never observed in wild type animals. These data strongly suggest a role for katanin p80 in cytokinesis and midbody resolution. This hypothesis is supported by the localization of katanin p80 (Figure 6E) and p60 (Figure S3) to the microtubules of the midbody in late telophase cells in both Katnb1WT/WT (Figure 6E) and Katnb1Taily/Taily germ cells (Figures S3 and data not shown). In addition, telophase cells with prominent midbody microtubules were observed in testes from Katnb1Taily/Taily, but not in wild type animals (Figure 6F). Taken together, the results suggest that katanin p80, most likely in association with p60, has a prominent role in midbody dissolution in male meiotic cells, and that this function is disrupted in Taily mice.
Abnormal sperm head shape (Figure 3D and 3L) is frequently associated with defects in the function of the manchette [30], [31], [32]. The manchette is a transient microtubule structure assembled in elongating spermatids with proposed roles in both the sculpting of the sperm head and in the movement of proteins destined for the sperm tail, via a process referred to as intra-manchette transport (IMT) [33]. The manchette is comprised of large, parallel arrays of microtubule bundles that extend from beneath the acrosome/acroplaxome region of the spermatid head and project into the spermatid cytoplasmic lobe containing the growing sperm tail ([34] and Figure S4). The manchette is first seen at step 8 of spermiogenesis, when the round spermatid nucleus polarizes to one side of the cytoplasm, and the spermatid commences elongation. Nucleation of microtubules in the manchette is thought to occur on the perinuclear ring region of the spermatid head (Figure S4), and large parallel bundles are assembled as the spermatid nucleus starts to change shape in step 9.
In order to investigate the hypothesis that head abnormalities in sperm from Katnb1Taily/Taily mice were the consequence of abnormal manchette structure or function, testis sections were examined using electron microscopy. Manchettes in wild type elongating spermatids displayed the characteristic perinuclear ring and microtubule array structure (Figure 3C). Those from Katnb1Taily/Taily elongating spermatids, however, displayed several abnormalities including constricted perinuclear rings, nuclear distortion and abnormally long microtubules extending into the distal cytoplasm (Figure 3D).
A stage-by-stage comparison of sections from wild type and Katnb1Taily/Taily males suggested defective manchette resolution in mutant animals. Although manchettes eventually resolved, the removal of manchettes in Katnb1Taily/Taily males was delayed. In wild type mice, manchettes normally reduce in size and then disappear in step 13 spermatids. In contrast, and when compared with wild type mice (Figure 7A), in step 13 spermatids from Katnb1Taily/Taily mice, abnormally long manchette microtubules extended into the cytoplasm and were associated with tubulin-labelled ‘clouds’ (Figure 7B). The timing and location of the ‘clouds’ is suggestive of abnormal microtubule disassembly.
The above defects were confirmed and more dynamically visualized in elongating spermatids isolated from wild type and Katnb1Taily/Taily males labelled with α-tubulin and TOPRO to visualize microtubules and the nucleus, respectively (Figure 8 and Videos S3 and S4). An analysis of progressively more mature elongating spermatids revealed that while manchettes in Katnb1Taily/Taily mice appeared to form at the correct time and initially began to move distally as spermiogenesis proceeded, movement stalled at approximately step 10 (Figure 8). By contrast, the progressive constriction of the peri-nuclear ring that normally occurs as the manchette moves over the caudal half of the spermatid, continued to occur (Figure 8). This resulted in a bulbous nuclear shape forward of the stalled peri-nuclear ring, and an abnormally elongated nucleus distally, resulting in the unusual ‘knob-like’ head structure also visible at an electron microscopic level (Figure 3D). As observed by electron microscopy (Figure 3D) and in testis sections (Figure 7B) abnormal elongated manchette microtubules were easily observed in isolated late step spermatids (Figure 8).
Consistent with the defects seen in Katnb1Taily/Taily mice, both the katanin p80 regulatory subunit and the p60 catalytic subunit localized to microtubules of the manchette (Figure 7C and 7D). Focal labelling of katanin subunits was observed along the manchette microtubules and particularly at the microtubule ends projecting into the cytoplasm (Figure 7C and 7D). This localization is consistent with a role for katanin-mediated severing in regulating manchette length. Localization was not obviously affected in the manchette of Katnb1Taily/Taily mice (not shown). Both subunits also localized to the acrosome/acroplaxome region, in a manner reminiscent of proteins that undergo trafficking in an acrosome/acroplaxome-manchette-flagella pathway [35].
Collectively these results reveal that katanin p80, most likely in association with p60, has an essential role in both the formation of the manchette and in the dynamics of its movement and resolution. Perturbed katanin p80 function results in the failure of manchette migration, abnormally long and mal-orientated manchette microtubules and abnormal dissolution. Such abnormalities are entirely consistent with the observed defects in sperm head shape (teratozoospermia) in sperm from Katnb1Taily/Taily animals.
The manchette also plays a critical role in the development of sperm flagella via a process known as intra-manchette transport, or IMT [33]. IMT is thought to be involved in sperm tail development in a manner analogous to intra-flagella transport in somatic cilia and in flagella in organisms including Chlamydomas and Trypanosomatids [36]. Defects in manchettes and sperm motility in Katnb1Taily/Taily mice, and a previously demonstrated role for katanin p80-mediated (PF15p) severing in the formation of axoneme central microtubules in Chlamydomonas [26], prompted us to investigate flagella/tail structure in Katnb1WT/WT and Katnb1Taily/Taily sperm. When compared to controls (Figure 7E), electron microscopic analysis revealed a variety of axoneme defects, including missing central microtubules and hemi-axonemes (Figure 7F). Of note, the majority of Katnb1Taily/Taily sperm also contained flimsy or missing outer dense fibers (Figure 7F) consistent with previously proposed roles for the manchette in the transport of proteins into the developing sperm tail and formation of accessory tail structures [33]. The outer dense fibers are rod-like structures running parallel to, and connected to, the microtubules of the axoneme that are believed to protect sperm against shearing forces and to provide directionality to tail bending (reviewed in [37]). Collectively, these defects demonstrate an essential role for katanin p80 in the development of the sperm flagellum, including the axoneme and the formation of the accessory structures.
Analysis of male mice with a mutation in the regulatory subunit of the katanin microtubule severing enzyme complex revealed several roles for katanin p80 in mammalian microtubule dynamics. These studies reveal an essential requirement for katanin p80 in male fertility, and in multiple aspects of mammalian male gamete development, including in meiotic spindle dynamics, cytokinesis, flagella development and sperm head shaping.
Together with in vitro data and evidence of disordered microtubule structure and function in lower order species, this novel mouse mutation reveals that katanin has roles in controlling microtubule dynamics. While severing can result in microtubule destruction, it is also important for the creation of new microtubules, via the severing of longer stable microtubules into shorter segments that can then be used as “seeds” for further microtubule polymerization [4], [7]. For example, katanin severs newly created microtubules at the neuronal centrosome, facilitating their transport to other sites, such as within developing axons [38]. Katanin can sever microtubule lattice defects in a quality control mechanism [39], and in neurons can create branched microtubule networks [4], [40]. Recent data also revealed that katanin can depolymerize microtubules at their plus-ends [5], [16], [39] and finally, the p60-p80 katanin complex has a severing-independent, microtubule cross-linking function at C.elegans oocyte meiotic spindle poles [23].
The p80 protein contains a WD40 domain that likely mediates protein-protein interactions [9], [12]. The C-terminal region interacts with the p60 enzyme [12] and contains binding sites for the molecular motor protein dynein. The C-terminal region also binds to the dynein-associated proteins LIS1 and NDE1 in neurons [18]. While p80 is thought to modulate p60 targeting and activity [3], [9], [12], the precise in vivo roles of p80 are not well understood. The Katnb1Taily/Taily mutation in the WD40 domain results in decreased p80 protein within germ cells and defects in microtubule-based processes. Based on the position of the mutation, and on the fact that less p80 protein is produced in mutant mice, we predict that the mutation influences the ability of p60 to sever, as well as the targeting of this severing activity to specific sites within the cell. This mouse model recapitulates many of the proposed functions of katanin observed in lower order species. It is of note, that the manchette defects observed in Taily mice phenocopy many of the defects observed in a Lis1 null mice [31]. This observation and previous studies showing the localization of LIS1 and NDE1 in the manchette [41], suggests that similar to the proposed role for these proteins in neurons, the katanin complex co-operates with LIS1 and NDE1 during sperm head shaping. This interaction will be the subject of future investigations.
The data demonstrate a role for katanin p80 in mammalian male meiotic cell division. Null mutations in the C.elegans p80 ortholog mei2 are associated with meiotic defects in oocytes, including an inability to assemble a meiotic spindle [14]. Katanin function in male germ cells has not previously been studied to the best of our knowledge. Our observations on male metaphase spindles in Katnb1Taily/Taily mice are consistent with the longer metaphase meiotic spindles produced in C.elegans oocytes with a partial loss-of-function mei2 mutation [20] and with the recent demonstration of a conserved role for katanin in controlling the length of meiotic metaphase spindles in Xenopus oocytes [23].
Within C.elegans, the p80 protein targets the p60 severing enzyme to the spindle poles in meiotic oocytes [20], [42]. We did not observe either p60 or p80 at spindle poles in male meiotic germ cells. All meiotic spindles were, however, abnormal, indicating a role for katanin in the assembly of spindles, as supported by various studies [14], [20], [21], [23]. In meiotic male germ cells, the most obvious localization of katanin subunits was to the microtubule ends near the chromosomes in metaphase-anaphase cells, suggesting a role for katanin in the shortening of microtubule plus ends during anaphase. The “Pacman-mediated” shortening of microtubule plus-ends within the spindle midzone is important for the poleward movement of chromosomes in mitotic anaphase [24], but has not been studied in meiotic cells. Our observation of cells apparently stalled in anaphase, together with the finding that 30% of cells die during the later phases of meiosis, supports the hypothesis that disturbed p80 function causes defects in the poleward movement of chromosomes during anaphase. Such defects result in disturbed spindle resolution and, often, cell death. Finally, the appearance of binucleated spermatids and the localization of katanin subunits to the midbody in meiotic cells in male mice supports the hypothesis that p80, and potentially the katanin complex, has a conserved role in modulating microtubule dynamics at the midbody during meiotic cytokinesis. In support, katanin p80 dysfunction or mislocalization is associated with defective mitotic cytokinesis in Trypanosomes [25] and in sarcoma cells in vitro [43].
Katanin p80 is essential for sperm head shaping via the regulation of the manchette, which is in itself a complex microtubule network. An analysis of manchette position during spermiogenesis indicated that manchette movement is defective in Katnb1Taily/Taily mice, suggesting katanin is involved in the organization and remodelling of this microtubule network as it moves over the nucleus. The localization of p60 and p80 within the manchette, together with the Taily phenotype is consistent with a role for katanin action at multiple sites. These include 1) the severing of microtubules at the perinuclear ring, thereby facilitating the release of microtubules from the nucleating center, and the production of the microtubule lattice, as has been proposed in other systems [4]; 2) the severing of microtubules near the nucleus to permit movement of the manchette perinuclear ring as it shapes the nucleus; 3) within the microtubule lattice to facilitate remodelling of this complex structure; and 4) the severing of microtubules at the caudal end of the manchette to control manchette length and dissolution. Katanin activity also regulates the dynamics of large microtubule-based array structures in neurons [18], [38], [40]. Taken together, the data support the hypothesis that p80, and katanin function, is important for the movement and remodelling of large microtubule arrays in mammalian cells. This role is essential for the normal development and shaping of sperm, which in turn is critical for normal sperm function and male fertility.
Finally, we demonstrate for the first time that katanin p80 is required for mammalian sperm flagella development and subsequent motility. A conserved role for katanin in the assembly and disassembly of cilia and flagella has been revealed in two distantly related lower order species Chlamydomonas [26] and Tetrahymena [13] (reviewed in [6]). Katanin activity controls flagellum length in Trypanosomatids [25] and severs axonemal microtubules during the deflagellation process in Chlamyodmonas [44], [45]. In Chlamydomonas, katanin p80 is also specifically required for the assembly of the central microtubule doublet of the flagellum axoneme [26]. Defects in axonemal structures (including missing central microtubule doublets) and outer-dense fibers in Katnb1Taily/Taily mice suggest that katanin p80 regulates sperm motility by acting at multiple sites in sperm development. Specifically, p80 function is required in the regulation of axonemal assembly and in the delivery of proteins to the developing flagellum via IMT.
Given the high level of expression of katanin p80 in other tissues, and the proven role for katanin in C.elegans oogenesis, it is surprising that other overt phenotypes were not noted in mutant mice. We hypothesize that other phenotypes will be revealed when mice are exposed to environmental insults. Studies into the role of katanin in oocyte function are ongoing.
In conclusion, the p80 subunit of the katanin microtubule severing enzyme complex is required for male fertility in mice. This is the first in vivo mammalian model of katanin p80 dysfunction, and it presents with a phenotype reminiscent of a commonly observed clinical phenotype of male infertility characterized by low sperm counts, poor motility and abnormal sperm morphology (referred to as oligoasthenoteratospermia or OAT). We conclude that p80 katanin is required for male meiotic spindle development and dynamics, and for the shaping of the sperm head via the regulation of manchette development and movement. Katanin p80 also participates in meiotic cytokinesis, likely via the regulation of the microtubules within the midbody, and controls the development and function of sperm flagella.
All animal experimentation was approved by the Australian National University and Monash University Animal Experimentation Ethics Committees and performed in accordance with Australian NHMRC Guidelines on Ethics in Animal Experimentation.
Point mutant mice were generated as described previously on a C57BL/6 background and outbred to CBA [27]. Mouse lines containing sterility causing mutations were identified by breeding trials wherein eight G3 brother-sister pairs per line were co-housed and the presence of pups monitored. Lines where male sterility was observed in a ratio of approximately one in four in the G3 generation with apparently normal mating behaviour were selected for further analysis.
Affymetrix 5K mouse SNP Chip arrays were used to map the sterility causing mutation. Genomic DNA from five affected males was hybridized onto the array at the Australian Genome Research Facility and compared to wild type C57BL/6 and CBA sequences. The linkage interval was subsequently narrowed using additional mice and SNPs (www.well.ox.ac.uk/mouse/INBREDS/) using the Amplifluor SNP Genotyping System (Chemicon). Plates were read in a BMG Fluostar optima fluorescent microplate reader.
Following the identification of the causal mutation, mice were specifically genotyped using the Amplifluor SNPs HT genotyping system using a wild type-specific antisense primer 5′-GAAGGTCGGAGTCAACGGATTAAGAGCACCCGTACCTGAC-3′, a mutant allele antisense primer 5′-GAAGGTGACCAAGTTCATGCTGAAGAGCACCCGTACCTGAA-3′, a sense primer, 5′-GGTGGTGAGCTGCATTGAA-3′ and Platinum Taq DNA Polymerase (Invitrogen). Conditions for amplification were as follows: 4 minute denaturation at 95°C, 35 cycles of denaturation at 95°C for 10 seconds, annealing at 60°C for 20 seconds and elongation at 72°C for 40 seconds, followed by a final 3 minute elongation at 72°C. Following the reaction, plates were read in a BMG Fluostar optima fluorescent plate reader.
Infertility in the Taily mouse line was classified using the scheme outlined in Borg et al [46]. Daily sperm output and total epididymal sperm content were determined as described previously [47]. Sperm motility was assessed using computer assisted sperm analysis [48] and ultra-structure using electron microscopy [49]. Cauda epididymal sperm morphology was assessed following staining with hematoxylin. Cells undergoing apoptosis were visualized using the Apoptag kit (Millipore) as recommended by the manufacturer.
The number of germ cells per Sertoli cell were enumerated in 25 µm thick, periodic acid Schiffs (PAS) stained methacrylate sections using the optical disector as previously described [50]. Retained elongated spermatids were counted in stage XI-XI [50] and expressed as fold wild type.
RNA was extracted from testes at defined periods throughout post-natal development using TRIzol regents (Life Technologies), treated with DNase I (Ambion) and cDNA sythnesized using oligo-dT primers and SuperScript III reverse transcriptase (Life Technologies). The relative expression of Katnb1, Katna1, Katnal1 and Katnal2 were defined using quantitative PCR using TaqMan assays (Applied Biosystems) Mm01244795_m1, Mm00496172_m1, Mm00463780_m1 and Mm00510701_m1 respectively. Expression of these was normalized against peptidylprolyl isomerase A (Mm002342429_g1).
Germ cell sub-populations were purified using the Staput method as previously described [51]. Single cell suspensions were loaded onto a 2–4% continuous BSA gradient and elongated spermatids and round spermatids collected after a 3 hour and 3.5 hour sedimentation period, respectively. For immunofluorescent staining, gradient fractions were pelleted and resuspended in 4% paraformaldehyde (PFA) fixative for 2 hours on ice. Cells were then washed with PBS and spread onto slides.
Protein was extracted from round spermatid fractions (>90% purity) using 20 µL M-PER buffer (Thermo Scientific). 10 µg of protein was separated on a 10% SDS-PAGE gel and probed for rabbit katanin p80 (HPA041165, which recognizes a C-terminal region of p80, Sigma Aldrich) and actin (A2066, Sigma Aldrich). Bound antibody was detected using a goat anti-rabbit IgG HRP (P0488, Dako) secondary antibody and an enhanced chemiluminescence (ECL Plus) detection kit (Amersham Biosciences).
Katanin subunits and α-tubulin were localized in testis sections as described [52]. Primary antibodies included: anti-α-tubulin (T5168, Sigma, diluted 1 in 5000), anti-katanin p60-like 2 (p60AL2, #sc-84855, Santa Cruz, diluted 1 in 100), anti-katanin p80 (diluted 1 in 200) [9] and anti-katanin p60 (diluted 1 in 200) [53]. Both p60 and p80 antibodies were affinity-purified from rabbits immunised against full length recombinant human proteins. These antibodies have been validated extensively and have been shown to recognize a single polypeptide in HeLa cells and a range of human tissues [9], [53]. Given the sequence homology between the p60 and p60L1 subunits, and between p80 and the uncharacterized c15orf29 subunit, there remains the possibility of partial cross-reactivity. Secondary antibodies included: Alexa Fluor 555 donkey anti-rabbit IgG (A-31572) and Alexa Fluor 488 donkey anti-mouse IgG (A-21202) (diluted 1 in 500). DNA was labelled using DAPI (Invitrogen). To define the localization of proteins within isolated elongating spermatids, cells were permeabilized in 0.2% Triton X-100 diluted in 10% normal horse serum (NHS) in PBS for one hour at room temperature. Non-specific labelling was minimized by blocking in 10% NHS in PBS for 30 minutes. Primary antibodies were diluted in 10% NHS in PBS and incubated overnight at 4°C. Secondary antibodies were diluted 1 in 200 and incubated at room temperature for 2 hours. DNA was labelled using TOPRO3 (Invitrogen, 1 in 200) or DAPI. Images were taken with an SP5 5-channel (Leica Microsystems) confocal microscope in the Monash University Microimaging facility. Metaphase spindle lengths were measured on α-tubulin and DAPI-stained sections from Katnb1WT/WT and Katnb1Taily/Taily mice using LAS AF (Leica Application Suite Advanced Fluorescence) software. Z-stacks of spindles and manchettes were collected at 0.5 µm intervals. Images were assembled using Adobe Photoshop. Test and subject images were adjusted uniformly across the image and between groups.
Differences between Katnb1WT/WT and Katnb1Taily/Taily mice were determined using unpaired t tests in GraphPad Prism 5.0.
|
10.1371/journal.pcbi.1003221 | Potential Role of a Bistable Histidine Kinase Switch in the Asymmetric Division Cycle of Caulobacter crescentus | The free-living aquatic bacterium, Caulobacter crescentus, exhibits two different morphologies during its life cycle. The morphological change from swarmer cell to stalked cell is a result of changes of function of two bi-functional histidine kinases, PleC and CckA. Here, we describe a detailed molecular mechanism by which the function of PleC changes between phosphatase and kinase state. By mathematical modeling of our proposed molecular interactions, we derive conditions under which PleC, CckA and its response regulators exhibit bistable behavior, thus providing a scenario for robust switching between swarmer and stalked states. Our simulations are in reasonable agreement with in vitro and in vivo experimental observations of wild type and mutant phenotypes. According to our model, the kinase form of PleC is essential for the swarmer-to-stalked transition and to prevent premature development of the swarmer pole. Based on our results, we reconcile some published experimental observations and suggest novel mutants to test our predictions.
| Recent evidence suggests that the transition of PleC from phosphatase to kinase is induced by its own substrate, DivK. Based on experimental clues, we propose a molecular mechanism to explain this substrate-induced conformational change in PleC. The general principles of thermodynamics, enzyme-substrate reactions and the Monod-Wyman-Changeux model of allostery motivate the elementary chemical reactions proposed in our model. Formulating our hypothesis in terms of nonlinear ordinary differential equations, we show that the PleC transition could function as a bistable switch. Although initial experimental studies have suggested that the primary role of PleC is as a phosphatase, our simulations show that the PleC kinase form is relevant for the correct temporal regulation of the Caulobacter cell cycle.
| The function of the cell division cycle of both prokaryotes and eukaryotes is to produce two nearly identical copies of a progenitor cell. The two progeny cells have identical genomes (modulo unavoidable mutations in the DNA replication process), and they are usually quite similar in all other aspects (called “symmetric” cell division). However, there are many cases of asymmetric cell division, in which the two progeny cells are notably different from each other [1]. An interesting example of asymmetric cell division is the freshwater bacterium, Caulobacter crescentus. Because Caulobacter populations typically live in low-nutrient environments, they have developed a strategy of asymmetric cell division to limit intraspecific competition [2]. During the cell division process, proteins are unequally distributed to the two halves of the cell, giving rise to two morphologically distinct daughter cells. One daughter cell (the stalked cell) is anchored to its place of birth via an appendage called the stalk, while the other daughter cell (the swarmer cell) is equipped with a flagellum and pilus that allows it to swim away from its place of birth. Hence, even though the total number of cells doubles, the number of stalked cells at a specific location stays the same. Another key difference is that, after cell division, the stalked cell immediately initiates a new round of DNA replication and cell division, while the wandering swarmer cell is not competent for DNA replication (it is in a prokaryote version of G1 phase). Once the swarmer cell finds a nutritionally suitable location, it will differentiate into an immobile stalked cell, initiate DNA replication, and establish a new population.
Orchestration of this asymmetric cell division cycle requires proper temporal and spatial regulation of several key proteins (see Figure 1A). The temporal dynamics of these proteins was captured in a pair of papers by Li et al. [3], [4]. At least two of these proteins, PleC and CckA, are bifunctional, capable of acting as either phosphatase or kinase. PleC kinase activity is up-regulated by its own response regulator, DivK. It is unknown how DivK alters the activity of its own phosphorylating enzyme, PleC. DivK is present at roughly constant level throughout the cell cycle [5]. However, PleC is a phosphatase during the swarmer stage of the cell cycle and kinase during the stalked stage (see Figure 1B). It would be interesting to know how this cross-talk between PleC-kinase and its substrate, DivK, is restricted to the stalked stage of the cell cycle.
At the level of physiology, whether a cell has a stalk or a flagellum depends on the phosphorylation status of the proteins DivK, PleD and CtrA. In the swarmer cell, CtrA∼P (the active, phosphorylated form of CtrA) binds to the origin of replication on the Caulobacter chromosome and inhibits initiation of DNA replication [6]. During the transition from swarmer to stalked cell, CtrA gets dephosphorylated and degraded, thereby lifting the block on DNA replication. In addition, CtrA affects the transcription of over 125 genes, so periodic changes in CtrA activity causes widespread changes in the expression profile of Caulobacter genes during the cell division cycle [7], [8]. DivK, on the other hand, is unphosphorylated in the swarmer cell and gets phosphorylated during the transition to the stalked cell. In the phosphorylated state, DivK initiates a pathway for stalk formation [9]. It is also responsible (indirectly) for the dephosphorylation and proteolysis of CtrA [10]–[12].
The phosphorylation states of DivK and CtrA are governed by the bifunctional histidine kinases PleC and CckA, respectively. Both PleC and CckA can switch between two conformations: a kinase conformation and a phosphatase conformation [9], [11] (see Figure 1B). Typically, in bacteria the change in activity of a bifunctional histidine kinase is brought about by an external signal molecule binding to the sensor region of the protein [13]. However, the change in PleC from a phosphatase to a kinase is brought about by its substrate, DivK [9]. In fact, the sensor domain of PleC is not essential for its function [14]. This interaction, where substrate binding to a bifunctional histidine kinase changes its function, has, to our knowledge, been observed only for PleC in Caulobacter. It has been suggested that DivK up-regulates PleC kinase activity preferentially in stalked cells because it is in stalked cells where DivK∼P and PleC are co-localized at the poles [9].
The initial phosphorylation of DivK during the swarmer-to-stalked transition is brought about by a kinase DivJ that localizes to the old pole. Hence, DivJ is considered as the enzyme that initiates the swarmer-to-stalked transition [15], [16]. A second and perhaps more crucial function of PleC kinase is to phosphorylate PleD, a diguanylate cyclase enzyme. On getting phosphorylated, PleD monomers dimerize and localize to the cell pole [17]. Active PleD converts two molecules of GTP into cyclic di-GMP, which signals production of the stalk [9]. Although mutations in divJ and pleC are not lethal, they result in growth and morphological defects in the cell. pleC::Tn5 mutants are stalkless [18], [19], while divJ-null mutants are filamentous and have elevated levels of CtrA-dependent transcription products [20], [21]. DivK∼P level is elevated in pleC::Tn5 mutants and reduced in ΔdivJ background. ΔdivJ pleC::Tn5 double mutants exhibit an even lower level of DivK∼P than ΔdivJ single mutants [21], indicating that PleC has a partial role, at least, as a DivK kinase.
CckA acts as a kinase in the swarmer cell, keeping the level of CtrA∼P high, which in turn blocks DNA replication [22]. In the stalked cell, CckA becomes a phosphatase, and CtrA gets dephosphorylated, allowing initiation of DNA replication [23]. DivL, a tyrosine kinase has been implicated in maintaining CckA in the kinase state [11], [12], [24], [25]. DivL can phosphorylate CtrA in vitro [18]. However, in vivo its role in maintaining a high level of CtrA∼P is indirect [24]. Multiple lines of evidence support the idea that DivL promotes CtrA phosphorylation via activation of CckA kinase. (a) divL mutants show marked reduction not only in CtrA∼P but also in CckA∼P [24], [26] and CpdR∼P [26]. (b) The phenotype of divJ over-expression mutants is alleviated by mutations in divL [20]. (c) DivK∼P is known to bind to DivL and interfere with its ability to activate CckA kinase [11]. Although the mechanism by which DivL influences CckA is unclear, DivL seems to be the intermediate by which the PleC-DivJ-DivK∼P axis regulates the level of CtrA∼P.
CckA's second substrate, CpdR, is phosphorylated and inactive in swarmer cells [23]. When CckA becomes a phosphatase in the stalked cell, active CpdR turns on the ClpXP proteolytic machinery for degrading CtrA [27], [28]. In this manner, CckA governs both dephosphorylation and proteolysis of CtrA.
Taken together, these observations suggest that PleC-DivJ-DivK and DivL-CckA-CtrA are crucial drivers of the swarmer-to-stalked transition, as summarized in Figure 1 and Figure 2.
Here, we propose a mechanism for ligand-dependent modifications of the bifunctional histidine kinase, PleC. The mechanism consists of elementary chemical reactions describing ligands (either DivK or DivK∼P) binding to the histidine kinase dimer in either its phosphatase or kinase form. The binding states determine the rates of the autophosphorylation, phosphotransfer, and phosphatase reactions catalyzed by PleC. If DivK∼P is more efficient than unphosphorylated DivK at promoting the transition of PleC from phosphatase to kinase, then PleC and DivK∼P would be involved in a positive feedback loop. Such positive feedback loops are well-known for their tendency to function as bistable toggle switches [29], and toggle switches are well-known for their roles in cellular decision-making [30]–[32] including critical transitions in the eukaryotic cell cycle [33]–[35].
In the Supplementary Material (Text S1), we show that a detailed model of the interactions between DivK and PleC, under reasonable conditions on the rate constants (or propensities) of these reactions, exhibits robust bistability as a function of DivJ activity. That is, by carrying out the initial, limited phosphorylation of DivK, DivJ can function as the “toggle bar” for flipping the bistable switch from the PleC-phosphatase state to the PleC-kinase state. When DivJ activity is low (swarmer cell), PleC is a phosphatase and DivK is predominantly dephosphorylated. As DivJ activity rises, enough DivK gets phosphorylated to flip the PleC switch to the kinase form (stalked cell). By coupling DivK∼P to DivL, we show that the PleC switch can induce the transition of CckA from kinase to phosphatase form, causing CtrA∼P and CpdR∼P levels to drop in the nascent stalked cell (Figure 1B).
This model of a PleC bistable switch is an intermediate step on the way to a full spatial model of the asymmetric division cycle in Caulobacter cells (in preparation). Using a model based on ordinary differential equations (biochemical kinetics of spatially homogeneous reactions), we address in this paper only certain features of the control system that are independent of the complex spatio-temporal choreography of the cell cycle control system. In particular, we validate our model of the PleC switch against known mutant phenotypes, and then we discuss some predictions of the model: (a) over-expressing DivK should result in a loss of asymmetry and cell cycle arrest in the stalked cell stage, (b) PleC kinase is required to ensure that the nascent swarmer pole will mature only after cytokinesis, and (c) the swarmer-to-stalked transition is robust to fluctuations in nutrients available in the environment.
Our detailed mechanism of substrate-induced conformational changes in PleC is presented in the Supplementary Material (Text S1). The model is based on the following considerations. PleC is a homodimeric, bifunctional histidine kinase. It can bind to either DivK or DivK∼P. As a kinase, it phosphorylates DivK to DivK∼P, and as a phosphatase it hydrolyzes DivK∼P back to DivK. We assume that, when DivK or DivK∼P are bound to both subunits of PleC, the enzyme undergoes a concerted conformational change from its phosphatase form to its kinase form. The conformational change is described in the manner of the Monod-Wyman-Changeux [36] theory of allosteric enzymes. A detailed model of PleC-DivJ-DivK-PleD interactions contains 38 biochemical species (Table S4, Eq. 1–38; Figure S1A and B), many of which are involved in null-cycles. To build a kinetic model of this reaction network, we must assign reasonable values to all the forward and reverse rate constants (kf and kr), respecting the fact that kf/kr = Keq = exp(−ΔG0/RT), where ΔG0 is the standard Gibbs free energy change and Keq is the equilibrium constant for the reaction. In the Supplementary Material (Text S1) we assign reasonable ΔG0 values to every reaction in the network, and then assign kf and kr values consistent with the computed equilibrium constants. In this way, we are assured that our kinetic model satisfies the Principle of Detailed Balance around all null-cycles. (For a null cycle, ΔG0 = 0 and Keq = 1; hence, the product of forward rate constants around the cycle = the product of reverse rate constants around the cycle.) Having built a kinetic model that is consistent with the thermodynamic requirements of the histidine kinase (PleC)—response regulator (DivK) system, we then show (see Figure 3) that the ‘two component’ system does indeed exhibit bistability as a function of DivJ activity.
In the next subsections, we examine biochemically relevant features of this bistable control system.
ΔdivJ cells are filamentous [20], [21], show mislocalized stalks and delocalized DivK [5]. In addition, the level of phosphorylation of DivK in ΔdivJ cells is reported to be only 44% of wild-type level [21]. Not surprisingly, CtrA∼P level is higher in this deletion mutant [20]. Furthermore, mutations in divJ have an adverse effect on cell division rate [20], [37], [38]. Hence, DivJ is considered to be a cell-fate determinant, essential for a smooth swarmer-to-stalked transition [39].
Paul et al. [9] suggested that DivJ initiates the PleC phosphatase-to-kinase transition, by a positive feedback loop: DivK, on being phosphorylated by DivJ, activates PleC autokinase, and PleC kinase makes more DivK∼P. Their experiments, however, indicate that PleC kinase activity is up-regulated by DivK irrespective of DivK's phosphorylation state. Given that the total concentration of DivK remains the same throughout the cell cycle [5], why isn't PleC a kinase at all times?
Presumably, the phosphatase form of PleC has a higher affinity for its substrate DivK∼P than for its product DivK. Therefore, even though the PleC phosphatase-to-kinase transition may be promoted by either DivK∼P or DivK, DivK∼P has a greater propensity than DivK to induce the conformational change. Once PleC becomes a kinase, it produces more DivK∼P, which enhances the rate of change from phosphatase to kinase. This self-reinforcing positive feedback loop between DivK∼P and PleC kinase can turn the PleC transition into a bistable “toggle” switch [29].
As shown in Figure 3A, DivJ can function as the lever of this toggle switch. As the activity of DivJ increases, PleC switches abruptly from a steady state of low kinase activity to a steady state of high kinase activity. DivK also transitions from a mostly-unphosphorylated steady state to a mostly-phosphorylated steady state (Figure 3B and Figure S6B), as does PleD as well (Figure 3C and Figure S6F). We propose that this toggle switch underlies the swarmer-to-stalked transition, where the arrival of DivJ at the old pole triggers PleC to switch to its kinase form, thereby triggering a new stalk end through PleD phosphorylation. It has been shown that upon glucose starvation, DivJ localization is inhibited, and the proportion of swarmer cells in the population doubles [39]. To test the signal-response curves in our model, it would be interesting to see if single cells can toggle between swarmer and stalked morphology upon changing nutrient composition.
According to Paul et al., accumulation of DivK∼P at the poles causes its local concentration to increase beyond a threshold required for the activation of PleC kinase. Our model does not address this possibility because (at present) it does not take space into account. While we cannot rule out the contribution of polar localization, our model shows that it is not essential for the phosphatase-to-kinase transition. Our simulations indicate that a large fraction of PleC kinase is bound to DivK (Figure 3D). Hence, it is possible that localization of DivK∼P is not the cause but the consequence of PleC kinase up-regulation. PleC kinase molecules may serve as docking sites for DivK molecules at the flagellar pole. PleC phosphatase on the other hand need not have any bound DivK. This picture is in agreement with observations that PleC, DivJ and DivL contribute to localization of DivK∼P to the poles [19], [40].
In vitro experiments show that PleC kinase activity increases in response to increasing DivK concentration, even in the absence of DivJ [9]. The specific activity of PleD in forming cyclic di-GMP was used as a proxy to measure PleC kinase activity. Surprisingly, the specific activity of PleC kinase in vitro is two-fold greater in the presence of DivKD53N, a mutant form of DivK that does not get phosphorylated. This indicates that DivK need not be phosphorylated to induce a conformational change in PleC. In vivo, however, PleC remains a phosphatase in the DivK-rich swarmer cell. Another odd result of the assay is that the specific activity of PleC kinase drops sharply at high DivK concentrations.
To reproduce these results in ΔdivJ mutants, we set [DivJ] = 0 in our simulations (Table S8). To simulate the divKD53N mutation, we set the rates of all phosphotransfer reactions to zero (Table S8). In Figure 4 we plot steady-state PleD phosphorylation level against increasing total concentration of DivK (from 0.3 to 30). Our simulations show a qualitative similarity to the experiments [9]. PleD∼P level rises at first and then drops at high [DivK] (Figure 4A–C). PleD∼P levels in ΔdivJ divKD53N simulations (Figure 4A) are comparable to PleD∼P levels in ΔdivJ (Figure 4B) and wild-type (Figure 4C) simulations. These results support the findings by Paul et al. [9] that unphosphorylated DivK is also able to up-regulate PleC kinase. There is a sharp drop in PleD phosphorylation at high [DivK] because PleC shifts predominantly to DivK-bound forms that do not have a free binding site for PleD (Figure 3D) and therefore cannot phosphorylate it. Product inhibition by cyclic di-GMP may also play a significant role [41], but this effect is not included in our model.
Since DivK is capable of activating PleC kinase in the absence of DivJ, we plotted a two-parameter bifurcation diagram to estimate the effect of varying concentrations of DivJ and DivK on PleC activity (Figure 5A). The enclosed bistable region tapers off as we increase either total DivJ or total DivK (ksyndk). This implies that at moderate concentrations of DivK (e.g., ksyndk = 0.015), the PleC phosphatase-to-kinase transition is robust and dependent on the activity and localization of DivJ (Figure 5B). However, increasing DivK in the cell would lead to transitions that are less robust and independent of DivJ. We predict that a 5- to 10-fold increase in DivK concentration will result in PleC being locked in the kinase form, and the cell will be blocked in the stalked stage of the cell cycle. We propose that in vivo the total concentration of DivK is low enough that it needs to be phosphorylated in order to induce PleC to become a kinase. In this case, the bistable PleC switch becomes reliant on the appearance of DivJ activity rather than on the polar accumulation of DivK.
The DivL-CckA-CtrA module bears a striking resemblance to DivJ-PleC-DivK switch. Nonetheless, there are important differences. DivL can phosphorylate CtrA in vitro, but this reaction is of no significance in vivo [22], [42]. Unlike PleC, which directly transfers its phosphoryl group to an aspartate residue on DivK, CckA relies on a series of phospho-transfer events [23]. To this end, it has an additional aspartate-containing domain which first picks up the phosphoryl group from the histidine residue and passes it on to the histidine residue of a downstream histidine phosphotransfer (HPt) protein called ChpT [43]. Finally, ChpT relays the phosphoryl group to the aspartate residue on the response regulator CtrA. In our mathematical equations, we model ChpT and CckA as a single protein, CckA, whose transition from phosphatase to kinase is promoted by binding to substrate, CtrA. The third difference is that CtrA is not known to up-regulate CckA kinase, so there is no reason to expect bistability in the CckA-ChpT-CtrA phospho-relay system.
It is a well-established fact that DivK∼P inhibits CtrA activity, and the mechanistic details of this process have become progressively clear. Initial experiments showed that DivK∼P down-regulates CckA kinase activity [12]. Later experiments indicated that DivL is required for maintaining CckA as a kinase, and that DivK∼P binding to DivL inhibits this effect [11], [44]. Since the mechanistic details regarding how DivL influences CckA activity are currently unknown, we model this process phenomenologically, using a Hill function to describe how DivL promotes CckA kinase. We couple the PleC-DivK∼P bistable switch to the CckA kinase-to-phosphatase transition by having DivK∼P bind to and inactivate DivL. In the swarmer cell, DivJ is absent and the PleC switch is in the phosphatase state (DivK unphosphorylated). Hence, DivL is active and maintains CckA in the kinase state (CtrA phosphorylated). The up-regulation of DivJ is the trigger for the swarmer-to-stalked transition. DivJ activity flips the PleC switch to the kinase state, DivK gets phosphorylated and binds to DivL. DivL activity drops abruptly (Figure 6A), and consequently CckA returns to its default phosphatase form (Figure 6B). As a result, CtrA becomes dephosphorylated and inactive (Figure 6C and Figure S6D), and CpdR becomes dephosphorylated and active (Figure 6D).
The proposed coupling of these switches is supported by experimental evidence that a ΔdivJ mutant can be rescued by point mutations in divL and cckA genes [20]. CtrA activity, which is high in ΔdivJ cells (Figure 7A), is restored to normalcy by point mutations in divL and cckA that interfere with CtrA phosphorylation (Figure 7B–D). As expected, CtrA∼P level in a ΔdivJ mutant can be reduced by decreasing the specific activity of DivL (Figure 7B). Interestingly, our simulations show that decreasing the specific activity of CckA kinase lowers the level of CtrA∼P (Figure 7C), but increasing the specific activity of CckA phosphatase does not restore CtrA∼P level (Figure 7D). Hence, we predict that the point mutations in CckA that rescue ΔdivJ mutants do so by reducing the kinase activity of CckA.
To simulate the consequences of the divKD90G mutation, we make note of the fact that, in vitro, autophosphorylation of PleC is markedly reduced in the presence of DivKD90G [9]. This fact indicates that DivKD90G, unlike its wild-type counterpart, is unable to up-regulate the kinase form of PleC. Since DivKD90G is not an allosteric ligand, we set , and accordingly updated the equilibrium constants and parameters for all the concerned reactions (Table S8). In addition, although DivKD90G is phosphorylated to the same extent as wild type DivK, it is unable to bind to DivL [11]. Hence, we altered the binding equilibrium of DivKD90G to DivL (Table S8).
Using the altered parameter set, we tried to reproduce two known phenotypes of divKD90G cells. Filamentous divKD90G cells initiate swarmer progeny-specific development (SPD) prematurely. SPD defines a range of cell cycle events, including activation of the flagellum, development of pili, release of the flagellum and ultimately development of the stalk [45]. It is important that these events take place in a timely manner and that they are restricted to the newborn swarmer cell. Filamentous divKD90G mutants, however, initiate SPD in the pre-divisional cell. In particular, pilin synthesis (a part of SPD) requires CtrA∼P. Hence, we examined whether CtrA∼P level is increased in simulations of divKD90G mutant cells. Figure 8 compares one-parameter bifurcation diagrams for wild-type (green) and mutant (red) cells. The levels of DivK∼P (Figure 8A), PleD∼P (Figure 8B) and PleC kinase (Figure 8C) are much lower in mutant cells, while CtrA∼P level remains high (Figure 8D). This could potentially lead to initiation of SPD.
The divKD90G mutation is a suppressor of the pleC::Tn5 mutant phenotype. Cells lacking PleC show extended periods of bipolar localization of DivK∼P and also fail to develop stalks. A pleC::Tn5 divKD90G double mutant does not show any of these defects [45]. Our simulations show that DivK∼P level increases and CtrA∼P level drops in pleC::Tn5 background (Figure 8E and F). Since DivK remains phosphorylated in the absence of PleC, it is not dislodged from the poles [19]. DivK∼P binds to DivL and suppresses CtrA phosphorylation (Figure 8F), thus preventing SPD. However, in the pleC::Tn5 divKD90G double mutant, CtrA∼P level remains high in spite of elevated DivK∼P (Figure 8E–F, red line). This result is in accordance with the finding that CckA∼P, CtrA∼P and Cpdr∼P levels are high when the binding of DivK∼P to DivL is weakened [11]. Restoration of CtrA∼P in the double mutant allows flagellar pole development. Hence, the restoration of unipolar localization of DivK in pleC::Tn5 divKD90G double mutant may be a natural consequence of the inability of DivKD90G to bind to DivL.
Although PleC is bifunctional, its designation in the cell has primarily been that of a phosphatase. This view has been fostered by results showing an elevation in DivK∼P in pleC::Tn5 mutants [21]. Furthermore, pleCF778L mutants, which lack autokinase activity, appear to have a normal cell cycle [45]. However, later experiments have shown that, although cells possessing PleCF778L progress through the cell cycle without any problems, they show a marked reduction in holdfast attachment [9]. These cells also show lower c-di-GMP levels, indicating that PleD is not sufficiently phosphorylated and activated in the absence of PleC kinase activity. Another mutant that reduces PleC autokinase activity is divKD90G [9]. In contrast to the pleCF778L mutants, cells possessing the divKD90G mutation do not require cytokinesis to initiate SPD.
If both mutations result in loss of PleC autokinase activity, why does only one of them exhibit premature SPD? One may argue that premature SPD is not due to the loss of PleC kinase activity, but is instead a consequence of inability of DivKD90G to bind to DivL. However, we found that altering the rate constants governing the binding reaction had no effect on the phenotype, because DivKD90G∼P is low at all times and hence does not inhibit DivL. To shed light on this discrepancy, we propose a novel mutant strain of DivK, which we call divKX. The novel mutant deviates from divKD90G in that it retains wild type ability to bind to DivL. By simulations, we compare the phenotypes of divKD90G, divKX and pleCF778L (see Figure 9). To model the pleCF778L mutant, we set the autophosphorylation rates to zero (Table S8).
In comparison to wild type, pleCF778L cells show a reduction in the level of PleD∼P; but DivK∼P and PleC kinase levels show only modest difference (Figure 9A and C). This simulated comparison agrees with experimental observations, which show that pleCF778L cells have reduced surface attachments but otherwise cycle normally. We reason that, although pleCF778L does not have kinase activity, it still retains its ability to switch to the kinase form. Hence, in stalked and pre-divisional cells, the majority of PleC is locked in the inactive kinase conformation. It follows that the PleC phosphatase to DivJ ratio is low and most of the DivK is phosphorylated. In comparison, divKD90G and a divKX show a reduction in the PleC kinase level (Figure 9B and D). Since most PleC is in the phosphatase form, DivK∼P level is low and CtrA∼P level remains high throughout the cell cycle, thereby initiating SPD prematurely. Based on these simulation results, we propose that PleC kinase is important to prevent premature SPD. In the pre-divisional cell prior to compartmentalization, DivJ maintains PleC as a kinase while DivK is phosphorylated and bound to the pole/s. Once cytokinesis occurs, DivJ and PleC find themselves in different compartments, causing PleC to switch back to a phosphatase and allowing SPD.
We propose a model of the Caulobacter swarmer-to-stalked (G1-to-S) transition based on a pair of bifunctional histidine kinases, PleC and CckA. We suggest that the phosphatase-to-kinase transition of the PleC bifunctional enzyme is governed by concerted conformational changes brought about by homotropic interaction with its response regulator, DivK. By formulating a mathematical model based on a set of elementary chemical reactions, we show that the transition from phosphatase to kinase can function as a bistable switch driven by the starter kinase, DivJ. Our simulations reproduce the in vitro experimental observation that DivK and/or DivK∼P up-regulate PleC kinase activity. We hypothesize that even if DivK and DivK∼P have equal potential for causing the conformational change of PleC, DivK∼P is a more efficient inducer as a natural consequence of it being a substrate to the relaxed form, the phosphatase form of PleC. That DivK∼P is a more efficient inducer of the phosphatase-to-kinase transition creates a positive feedback loop and the potential for bistability, and bistability would explain why the swarmer-to-stalked transition is irreversible [35].
The swarmer-to-stalked transition is triggered by a rise in activity of the starter kinase DivJ. Evidence suggests that DivJ accumulates in response to nutritional signals [39]. Compared to well-fed cells, a greater fraction of Caulobacter cells are devoid of DivJ foci and exist as swarmer cells under conditions of glucose exhaustion. Hence, we consider DivJ as a nutritional proxy and use it as a control parameter in our model. As observed in our bifurcation diagrams, as total DivJ accumulates, the proteins that drive the swarmer-to-stalked transition show abrupt and irreversible changes in activity at the boundary of the bistable region. Once the transition has occurred, the control system will not permit a reverse transition (stalked-to-swarmer) in response to a marginal drop in nutritional level (i.e., in total DivJ concentration). In our view, once the PleC flips to the kinase form, the cell is committed to a new round of DNA synthesis before it can make a new motility apparatus in the pre-divisional stage. While bistability is not an essential feature of the morphological transitions in the Caulobacter division cycle, we propose that bistability in the PleC phosphatase-to-kinase transition may ensure that the swarmer-to-stalked transition is robust and does not undergo a reverse transition in response to small fluctuations in nutrient levels.
Our model is able to reproduce phenotypes of known experimental mutants and provide additional insight into the underlying physiology. Mutants overexpressing DivK show a decrease in CckA phosphorylation, in addition to filamentous growth and chromosomal over-replication [43]. Our two-parameter bifurcation diagrams indicate that cells with elevated DivK can no longer be regulated by DivJ. At higher concentrations, DivK can drive the positive feedback even in the absence of DivJ, resulting in PleC being in the kinase form and CtrA∼P being down-regulated. This prediction can be tested by overexpressing DivK in a ΔdivJ background. Conversely ΔdivJ mutants with a normal level of DivK are blocked in G1 phase owing to high CtrA∼P, while point mutations in divL and cckA rescue ΔdivJ mutants [20]. Our simulations suggest that ΔdivJ mutants can be rescued by point mutations that down-regulate CckA kinase activity, but not by mutants that up-regulate CckA's phosphatase activity.
Prior experiments and a mathematical model [46] dealing with the PleC-DivJ-DivK system have focused almost exclusively on the phosphatase form of PleC, while the kinase form has been considered inconsequential. We argue on the contrary that PleC kinase activity is important for proper progression through the Caulobacter cell cycle. To demonstrate this claim, we make an important distinction between two mutants pleCF778L and divKD90G. Our simulations show that while PleCF778L has no autokinase activity, the majority of PleCF778L molecules in stalked cells are in an inactive kinase form. These cells would therefore, appear normal. On the other hand, most PleC molecules remain in the phosphatase form in cells containing DivKD90G. We predict that in wild-type pre-divisional cells, PleC localized at the new pole is in the kinase form. Compartmentalization has the effect of withdrawing DivJ, causing PleC to switch back to the phosphatase form, as seen in our signal-response curves. The PleC-containing compartment, in the absence of DivJ, transitions into a swarmer cell. In mutant divKD90G cells, we predict that PleC at the new pole is always a phosphatase. This, we reason, would cause the premature presence of CtrA∼P in pre-divisional cells resulting in premature swarmer progeny-specific development (SPD). This conclusion is supported by the fact that filamentous divKD90G mutants show SPD in the absence of compartmentalization [45]. We are aware that divKD90G has a pleotropic effect of binding weakly to DivL. Hence, we hypothesize a novel mutant, divKX, which is similar to divKD90G but retains its ability to bind DivL. We simulate such a mutant and find its behavior to be comparable to divKD90G.
In this work, we are focusing on a small window in the Caulobacter cell cycle, the G1-to-S transition. We have not explored here how these coupled switches would function in a spatio-temporal context and whether they play a role in generating asymmetry in the two halves of the cell at a later stage in the division cycle. To explore these questions requires a spatio-temporal model that tracks the location of proteins in the cell and takes into account the effects of protein diffusion through the cytoplasm, as in [12], [46]. Without an accurate spatio-temporal model of these molecular interactions, we are still a long way from understanding the network of molecular interactions that governs the asymmetric life cycle of Caulobacter crescentus.
The complete reaction network (Figure S1) was translated into a system of 52 non-linear ordinary differential equations (Table S4) using the mass-action law of chemical kinetics, with one exception. The mechanism by which DivL promotes the kinase form of CckA is unknown, so we modeled this step phenomenologically with a Hill function. Because there are many closed loops of elementary chemical reactions in Figure S1, we must choose rate constant values that respect the thermodynamic principle of detailed balance, as explained in Text S1. As long as we satisfy these thermodynamic constraints, we find that the reaction network exhibits bistability over a robust range of parameter values. The parameter values that we use for our simulations of the full model (Table S4) are given in Table S5.
The full model can be simplified slightly by reducing the first 28 equations in Table S4 to the first 20 equations in Table S6, as explained in Text S1, section D, and confirmed in Figure S4.
The equations for both the full model and the reduced model were encoded as .ode files (Text S2, S3, S4, S5) and simulated using the freely available software, XPP-AUT. The signal-response curves were drawn using the AUTO facility of XPP-AUT. From the data points generated by XPP-AUT, the plots shown in the figures were generated using the python library, Matplotlib [47]. Figure 3 is a simulation of the full model described in Table S4, while Figures 4–9 are simulations of the reduced model and its corresponding mutants (Table S4 and Table S8).
|
10.1371/journal.pgen.1004928 | Functional Interplay between the 53BP1-Ortholog Rad9 and the Mre11 Complex Regulates Resection, End-Tethering and Repair of a Double-Strand Break | The Mre11-Rad50-Xrs2 nuclease complex, together with Sae2, initiates the 5′-to-3′ resection of Double-Strand DNA Breaks (DSBs). Extended 3′ single stranded DNA filaments can be exposed from a DSB through the redundant activities of the Exo1 nuclease and the Dna2 nuclease with the Sgs1 helicase. In the absence of Sae2, Mre11 binding to a DSB is prolonged, the two DNA ends cannot be kept tethered, and the DSB is not efficiently repaired. Here we show that deletion of the yeast 53BP1-ortholog RAD9 reduces Mre11 binding to a DSB, leading to Rad52 recruitment and efficient DSB end-tethering, through an Sgs1-dependent mechanism. As a consequence, deletion of RAD9 restores DSB repair either in absence of Sae2 or in presence of a nuclease defective MRX complex. We propose that, in cells lacking Sae2, Rad9/53BP1 contributes to keep Mre11 bound to a persistent DSB, protecting it from extensive DNA end resection, which may lead to potentially deleterious DNA deletions and genome rearrangements.
| DNA double strand breaks (DSBs) are among the most deleterious types of damage occurring in the genome, as failure to repair these lesions through either non-homologous-end-joining (NHEJ) or homologous recombination (HR) leads to genetic instability. The 5′ strand of a DSB can be nucleolytically degraded by several nucleases and associated factors, including Mre11, CtIP/Sae2, Exo1 and Dna2 together with Bloom helicase/Sgs1, through a finely regulated process called DSB resection. Once resection is initiated, error-prone NHEJ is prevented. Several findings suggest that DSB resection is a double-edged sword, if not finely regulated, since on one hand it is needed for faithful HR, but on the other it may lead to extensive DNA deletions associated with genome instability. Both in mammals and yeast, 53BP1/Rad9 protein binds near the lesion and counteracts the resection process, limiting the formation of ssDNA. By using S. cerevisiae as a model organism, here we show that Rad9 oligomers block the removal of hypo-active Mre11 protein from a persistent DSB, thus limiting initiation of resection and the recruitment of the recombination factor Rad52, in the absence of Sae2. Altogether, these findings pinpoint a critical role of 53BP1/Rad9 in balancing HR and NHEJ repair events throughout the cell cycle.
| Similarly to what is seen in higher eukaryotes, in S. cerevisiae the ends of a double-strand DNA break (DSB) are recognized and bound by the Mre11-Rad50-Xrs2 (MRX) complex and the Ku70-Ku80 heterodimer, which compete for end binding. Once the MRX complex, together with CDK1-phosphorylated Sae2 (CtIP in human), initiates resection of the DNA ends, Ku70-Ku80 binding and NHEJ (non-homologous end-joining) are prevented [1], [2], [3], [4]. Subsequent 5′–3′ long-range resection can then occur by one of two pathways: the first utilizes the RecQ helicase Sgs1 (BLM in human), in cooperation with the endonuclease Dna2, and the second utilizes the exonuclease Exo1 [5], [6], [7], [8], [9].
The regulation of DSB end resection is very important to choose the right pathway to repair a DSB and avoid chromosomal rearrangements [10], [11]. Whereas classical NHEJ requires little or no resection, HR (homologous recombination) is characterized by extensive exonucleolytic degradation of one strand. Blocking DNA end resection affects the efficiency and accuracy of how a DSB is repaired. For example, inhibiting resection leads to de novo telomere addition, and eventually loss of a portion of a chromosome [12], [13]. On the other end, extensive DNA end resection could lead to accumulation of unstable DNA intermediates and eventually to the highly error-prone microhomology-mediated end joining (MMEJ) and single-strand annealing (SSA) events, which may cause DNA deletions and translocations [14], [15], [16].
It is now clear that the DNA damage checkpoint response (DDR) plays a central role in regulating DSB end resection. In fact, while resection proceeds, the formation of RPA-coated ssDNA activates the upstream kinase Mec1 (ATR in mammals) and the effector kinase Rad53 (Chk2 in mammals), which in turn phosphorylates and inhibits Exo1 [17]. Interestingly, Exo1 is regulated through a DDR pathway in human cells, too [18], [19].
Moreover, studies both in yeast and mammals showed that Exo1 and other DNA end-processing enzymes are inhibited through a physical structural “barrier” formed by Rad9 oligomers (53BP1 in mammals) bound near a DSB [10]. RAD9 was originally identified as the first checkpoint gene in S. cerevisiae and recognized as an “adaptor” protein, linking the upstream kinase Mec1 to the activation of effector kinases Rad53 and Chk1. Rad9 is recruited to chromatin through three different pathways: i) the constitutive interaction with the histone H3 methylated at the K79 residue by Dot1 [20], [21], [22]; ii) the binding to the histone H2A phosphorylated at the S129 residue by Mec1 [23]; iii) the interaction with Dpb11 [24], [25]. In particular, phospho-H2A mediated Rad9 recruitment spreads many kilobases around a DNA lesion [26]; whereas Dpb11 appears to be more specific at the site of lesion, by binding to a damage-induced phosphorylation in the Ddc1 subunit of the 9-1-1 complex [25], [27], [28]. All of these three pathways cooperate for efficient checkpoint arrest and cell survival after genotoxic treatments throughout the cell cycle. Moreover, Rad9 contains motifs that are necessary for its oligomerization and DNA damage checkpoint signalling [24], [29], [30].
Notably, the Rad9-mediated inhibition of DSB resection is a regulatory function conserved throughout evolution. In fact, 53BP1 facilitates NHEJ at the expense of HR, protecting DNA ends from inappropriate 5' resection, in cooperation with the telomere binding protein RIF1 [31], [32], [33], [34], [35].
Here, we show that in the absence of Sae2, or in presence of mutations affecting Mre11 nuclease activity, Rad9 dimers and/or oligomers, recruited near a DSB mainly by Dpb11 interaction, inhibit the short-range DNA end processing, thereby preventing Mre11 removal from the lesion and limiting Rad52 recruitment by an Sgs1-dependent mechanism. As a consequence, DSB ends cannot be kept efficiently tethered to each other, and repair through an SSA process is prevented. We propose a novel molecular role of Rad9/53BP1 to protect genome integrity from extensive DNA degradation and rearrangements during DSB repair, also suggesting important implications for malignant transformation in mammalian cells.
It is known that deletion of the RAD9 gene in yeast leads to faster DSB resection and repair through an SSA process [36], [37]. To further understand the role of Rad9 in DSB processing and repair, we decided to combine the deletion of RAD9 gene with mutations in genes encoding factors either involved in the short-range (SAE2), or the long-range (EXO1, SGS1) DSB resection [38]. We took advantage of the YMV80 background, in which the galactose-induced expression of the HO nuclease causes a single DSB at a specific site on chromosome III. Repair of this DSB occurs mainly through SSA between flanking homologous leu2 repeats one of which is 25kb from the DSB [39]. We deleted RAD9, EXO1, SGS1 and SAE2 to obtain all viable single, double and triple mutant combinations. Although the sae2Δ sgs1Δ double mutant is a synthetic lethal combination [40], [41], rad9Δ interestingly suppresses sae2Δ sgs1Δ lethality (S1A Fig.). Therefore, it was possible to test the sae2Δ sgs1Δ rad9Δ triple mutant cells. After plating the cells in the presence of galactose to induce one DSB, we found that viability of the sae2Δ and sgs1Δ single mutant and sgs1Δ exo1Δ double mutant was severely reduced (Fig. 1A), as expected [6], [7], [42]. We also found that the deletion of RAD9 gene effectively rescued the viability of the sae2Δ, sgs1Δ and sae2Δ exo1Δ mutant strains following one DSB (Fig. 1A). Interestingly, the viability of the sae2Δ sgs1Δ rad9Δ and exo1Δ sgs1Δ rad9Δ triple mutant cells was very low in the presence of one DSB. Moreover, the HO-induced lethality of the sae2Δ sgs1Δ rad9Δ mutant was not rescued by the expression of the Sgs1-K706A protein variant (S1B Fig.), whose helicase activity is severely reduced [43]. While the failure to repair the DSB in the exo1Δ sgs1Δ rad9Δ triple mutant was expected, since at least one of the Exo1 and Sgs1-dependent pathways is necessary to extensively resect a DSB, the result obtained with the sae2Δ sgs1Δ rad9Δ mutant was surprising. We therefore concluded that an Exo1-independent, Sgs1-dependent pathway is necessary for the viability of sae2Δ cells following a DSB in the absence of RAD9.
Since Sae2 stimulates the activity of the MRX complex in the first step of the DSB end processing [44], we considered the possibility that RAD9 deletion may also rescue an Mre11 nuclease defective mutant or the rad50Δ mutant, in which the MRX complex is disassembled. Interestingly, we found that rad9Δ suppresses the nuclease-defective mre11-D56N mutant [45], through an SGS1-dependent pathway, while it does not rescue rad50Δ mutant, as expected [36] (Fig. 1B). These results suggest that the nuclease activity of the MRX complex is dispensable for the DSB repair in rad9Δ cells; however, the MRX complex must be physically present, likely playing an essential structural role. Indeed, rad50Δ mutation does not rescue sae2Δ cell viability following a DSB (Fig. 1B). Of note, deletion of RAD9 also suppresses the double mutant mre11-D56N sae2Δ, further indicating that Mre11 and Sae2 work together in the same pathway (Fig. 1B).
Importantly, the deletion of RAD9 rescues sae2Δ cell viability through an EXO1-independent, SGS1-dependent pathway also in presence of camptothecin (Fig. 1C), a topoisomerase-aborting agent that causes formation of end-blocked DSBs [46].
To further investigate the findings shown in Fig. 1A at the molecular level, we tested the kinetics of DSB repair by Southern blotting in cells blocked in G2/M cell cycle phase by nocodazole. In agreement with the cell lethality reported in Fig. 1A, we found that the efficiency of the DSB repair is reduced in both the sae2Δ and sgs1Δ single mutants, as previously described [6], [7], [42], and it is severely compromised in sae2Δ sgs1Δ rad9Δ (Figs. 2B and 2C). On the contrary, DSB repair is accelerated and very efficient in the rad9Δ, sae2Δ rad9Δ and sgs1Δ rad9Δ mutants (Figs. 2B and 2C). These results indicate that, in the absence of Rad9, an Sgs1-dependent mechanism is necessary to efficiently repair a DSB in sae2Δ cells.
To test if Sgs1 cooperates with Dna2 to repair a DSB in sae2Δ rad9Δ mutant cells, we took advantage of an auxin-based degradable Dna2 protein variant (Dna2-DEG). This is a common genetic strategy to induce the degradation of a protein by the addition of auxin compound to the cell culture medium [47], and it is particularly useful in the case of an essential gene, such as DNA2. By Southern blotting analysis, we found that the sae2Δ rad9Δ double mutant cells do not repair a DSB in the absence of Dna2 (Fig. 2D and 2E). Therefore, taking all the data in Fig. 2 together, we concluded that the deletion of RAD9 rescues sae2Δ cells through a DSB resection mechanism mediated by the Sgs1-Dna2 pathway.
In addition, we ruled out the possibility that in the absence of Rad9, the DSB can be repaired more efficiently through a strand invasion-based mechanism (such as a break-induced replication process [48]). In fact, we observed faster DSB repair and high viability when we analysed the sae2Δ rad9Δ rad51Δ triple mutant, in which break-induced replication is impaired, but SSA is not inhibited (S2 Fig.).
A critical step to repair a DSB through SSA is 5′ to 3′ resection of the DSB end. Therefore, based on our results in Figs. 1 and 2, we hypothesized that in sae2Δ sgs1Δ rad9Δ triple mutant DSB resection may be affected, as it was shown in the sae2Δ single mutant [6], [7], [42], while it should be faster in sae2Δ rad9Δ double mutant. To test the kinetics of DSB processing we used JKM139 background derivatives, where prolonged expression of HO causes an irreparable DSB at MAT locus, because of the absence of HML and HMR homologous cassettes. Therefore, the analysis of the formation of the 3′ single-stranded (ss) DNA is not biased by a repair process [49]. Using Southern blotting of denatured DNA after restriction enzyme digestion [50], we tested the formation of the 3′ ssDNA filament (as depicted in Fig. 3A), after the induction of one DSB in each sister chromatid, in G2/M-blocked cells.
As expected, we found that the formation of a long 3′ ssDNA tail is slightly delayed in the absence of SAE2, EXO1 and SGS1 genes, and it is severely compromised in the exo1Δ sgs1Δ double mutant [6], [7], [51]. Interestingly, we found more extensive 3′ ssDNA in the absence of Rad9 in all the mutants tested, except the exo1Δ sgs1Δ rad9Δ triple mutant (Figs. 3B, 3C and S3). These results support the model that both the Exo1 and the Sgs1-dependent pathways cooperate to resect a DSB, and rule out the hypothesis that additional nuclease(s) may take over to process a DSB in the absence of Rad9. However, we noticed that in the sae2Δ sgs1Δ rad9Δ triple mutant strain the appearance of ssDNA is slightly delayed compared to wild type and sae2Δ rad9Δ strains (Figs. 3B and 3C). This result may suggest that the initiation of DSB resection is affected in sae2Δ sgs1Δ rad9Δ cells.
To test more precisely DNA processing near a DSB we employed a quantitative PCR-based method [52]. In particular, by this procedure we determined if the RsaI restriction enzyme can cut the DNA at a specific site 150 bp from the HO-cut site, thus indicating whether DSB resection has already passed beyond this site, since, as resection proceeds, the RsaI site becomes single stranded and resistant to digestion, which results in a PCR fragment amplification (see scheme in Fig. 3D). Thus, the rate of PCR fragment amplification, normalized to the efficiency of HO cutting, corresponds to the rate of resection [52]. We also tested with the same procedure another RsaI site 4800 bp from the HO cut site, as a control. Interestingly, we noticed a higher amount of un-resected DNA at 150 bp proximal the DSB site, between 60 and 180 minutes after the cut in nocodazole blocked sae2Δ and sae2Δ sgs1Δ rad9Δ triple mutant cells with respect to the wild type and sae2Δ rad9Δ mutant (Fig. 3E). However, at later time points resection has efficiently passed beyond the RsaI site 4800 bp far from the HO cut site (Fig. 3F), not only in the wild type and sae2Δ rad9Δ cells, but also in the sae2Δ sgs1Δ rad9Δ triple mutant cells, according to the visualization of the 3′ ssDNA formation by denaturing Southern blotting (Figs. 3B and 3C).
These studies revealed one striking unexpected result: although sae2Δ sgs1Δ rad9Δ triple mutant cells resect a DSB and expose an extended 3′ ssDNA (Figs. 3B, 3E and 3F), they are severely compromised in DSB repair through SSA (Figs. 2B and 2C), suggesting that the long-range resection is not the limiting step to repair a DSB in these cells, rather the defect is different from simply creating enough ssDNA to allow SSA to take place. Therefore, we hypothesize that an Sgs1-dependent mechanism contributes to efficiently initiate DSB processing in the absence of both Rad9 and Sae2, and the kinetics of the initial step of resection would become somehow critical to complete the subsequent steps of the SSA repair.
We then investigated whether the faster DSB end processing that we observed in sae2Δ rad9Δ cells would be associated with reduced NHEJ events, which are significantly elevated in the sae2Δ cells [53]. To this aim, we treated cells of JKM139 strains with nocodazole to block cell cycle in G2/M phase and we added galactose to induce one persistent DSB in each sister chromatid. Cells were kept in nocodazole for 2 hours to avoid potential interference caused by cell cycle transition, before plating in the presence of galactose. In this condition, the continued expression of HO leads to a recurrent cut of the MAT locus and precludes precise religation, until the sequence of the HO site is corrupted by deletion/addition of few bases and the ends are joined by imprecise NHEJ [54]. This is a relatively inefficient process in yeast, with a frequency of about 1-3×10−3 in wild type cells [54]. We found that the frequency of imprecise NHEJ events is increased in sae2Δ cells, in agreement with previous finding [53], while it is slightly reduced in the absence of Rad9. Interestingly, deletion of RAD9 reduces NHEJ events to wild type value in sae2Δ cells (Fig. 3G).
These results suggest that Rad9 plays a critical role to balance NHEJ and HR events in G2/M phase, likely acting at an early step of DSB processing, leading to increased NHEJ events in the absence of Sae2.
The delay in DSB resection in sae2Δ cells has been correlated with a prolonged Mre11 binding at the DSB site [42], [55]. More recently, it was also shown that an Sgs1-dependent process can contribute to remove Mre11 from a DSB in sae2Δ cells, promoting DSB resection and repair through homologous recombination [56]. Therefore, we decided to investigate Mre11 binding near a DSB by a chromatin immunoprecipitation-after-crosslinking-protocol (ChIP), followed by quantitative PCR (qPCR), with primers specific for the DSB site. Contrary to wild type, rad9Δ or sgs1Δ cells, we found greater and persistent levels of Mre11 bound near DSB ends in sae2Δ cells (Fig. 4A), supporting previous analysis of the Mre11 foci by microscopy [51], [56], and by ChIP [55]. Importantly, we found a decrease in fold enrichment of Mre11 binding to the DSB site in sae2Δ rad9Δ cells, but not in the sae2Δ sgs1Δ rad9Δ triple mutant cells (Fig. 4B). These results suggest that the deletion of RAD9 gene promotes an Sgs1-dependent process to remove Mre11 from DSB ends in the absence of Sae2, supporting and expanding recent findings [56], and it may explain the high efficiency of SSA repair and viability of the sae2Δ rad9Δ that we showed in Figs. 1 and 2. Moreover, the prolonged binding of Mre11 near the DSB further supports previous results in Fig. 3, showing that short-range resection in the sae2Δ and sae2Δ sgs1Δ rad9Δ triple mutant cells is delayed.
Since it is known that Mre11 persistence at a DSB limits the recruitment of Rad52 [4], [57], which is necessary to establish DNA end-tethering and HR pathways [58], [59], we investigated by immunofluorescence Rad52 loading onto one DSB in all the mutants described. We found that deletion of RAD9 totally restores Rad52 binding in sae2Δ cells through an Sgs1-dependent mechanism (Fig. 4C). These results correlate with the analysis of Mre11 binding in these mutants (Fig. 4B), and suggest that the limiting step to efficiently complete an SSA process in nocodazole-blocked sae2Δ and sae2Δ sgs1Δ rad9Δ cells is not the delay in DSB resection per se (Figs. 3B and 3C), but rather the reduced binding of Rad52.
Rad52 is a critical factor to maintain DSB ends tethered to each other, which was suggested to be a relevant event in HR [42], [58], [59], [60], [61]. As we showed that the deletion of RAD9 allows Rad52 binding in sae2Δ cells (Fig. 4C), we investigated whether it may also contribute to rescue DSB end-tethering defect in these cells. To this end, we took advantage of a specific yeast background in which the DNA proximal to the irreparable HO break could be visualized by binding of a LacI-GFP (green fluorescent protein) fusion protein to multiple repeats of the LacI repressor binding site, LacO. These arrays are integrated at a distance of 50 kb on either side of the HO cleavage site on chromosome VII [58]. Cultures of the original wild type and isogenic sae2Δ, sae2Δ rad9Δ and sae2Δ sgs1Δ rad9Δ derivative strains were arrested in mitosis and kept blocked by nocodazole treatment during break induction by galactose addition. After 2 hours to ensure HO cut formation, we observed two LacI-GFP spots in only 12.5%±2.1% of the wild type cells, and 11.0%±3.1% in sae2Δ rad9Δ mutant cells, thus indicating their ability to hold the broken DNA ends together. In contrast, 42.3%±3.8% of sae2Δ and 42.5%±4.8% of sae2Δ sgs1Δ rad9Δ cells showed two LacI-GFP spots, indicating a failure in DSB end-tethering (Fig. 4D, and see also [42], [62]).
Therefore, we conclude that the deletion of RAD9 rescues both the Rad52 binding and DSB end-tethering in sae2Δ cells, contributing to efficiently repair a DSB through an SSA process that requires the resection of 25 kb of DNA between the repeats (Fig. 2A).
It was previously suggested that Rad9 limits DSB resection acting as a physical barrier toward the actions of nucleases, through a function distinct from its role in DNA damage checkpoint signalling [10]. Therefore, we sought to address if a checkpoint-independent function of Rad9 was involved to limit sae2Δ cells viability following one DSB. To this aim, we tested the chk1Δ rad53-K227A double mutant in the YMV80 background, in which the Rad53 kinase activity is dead and both the two checkpoint-signaling pathways acting downstream Rad9 are abrogated. By plating the cells in the presence of galactose to induce one HO cut, we found that the viability of the sae2Δ chk1Δ rad53-K227A triple mutant cells is reduced, similarly to sae2Δ cells (Fig. 5A). This result indicates that signaling through Rad53 and/or Chk1 is not involved into the mechanism by which Rad9 limits SSA repair in sae2Δ cells.
In order to further understand how Rad9 inhibits SSA repair in sae2Δ cells, we then investigated specific mutations that affect Rad9 binding to a DSB. It is known that Rad9 constitutively binds chromatin through the interaction between its TUDOR domain and the histone H3 methylated at the K79 by Dot1 [20], [21], [22]. In addition, Rad9 binds chromatin around a DSB site through the interaction of its BRCT domain with the histone H2A phosphorylated at the S129 (γ-H2AX) by upstream kinase Mec1 and Tel1 [23]. Further, Rad9 is recruited near a DNA lesion through the interaction with Dpb11 protein. In particular, Dpb11 binds the CDK1-dependent phosphorylated S462 and T474 Rad9 residues, reinforcing the Rad9 binding to damaged DNA and promoting Rad9 phosphorylation by Mec1 [25].
To test the contribution of the different pathways that mediate Rad9 binding to chromatin, we analysed the viability in the presence of HO-induced DSB of specific mutations that abrogate each of them in the YMV80 background. The deletion of DOT1 gene eliminates the H3K79 methyl transferase Dot1 protein, and greatly reduces the constitutive binding of Rad9 to chromatin [21], [24]. As expected [36], deletion of DOT1 leads to a faster long-range DSB resection in sae2Δ cells (S4A and S4B Figs.). However, by the qPCR-based method, we found that the initial short-range resection is still delayed in these double mutant cells (S4C Fig.), suggesting that the Dot1-dependent resection barrier may have a role only at distal region from the cut site. Indeed, by plating the YMV80 derivative cells in the presence of galactose to induce one DSB, we found that deletion of DOT1 gene does not rescue sae2Δ lethality (Fig. 5A). Further, we deleted SAE2 gene in a strain that expresses the H2A-S129A histone variant, which is not phosphorylatable by Mec1 and Tel1 kinases and leads to a faster DSB resection [63]. We also deleted SAE2 gene in a strain that expresses the Rad9-S462A-T474A (hereafter we refer to rad9-S462A-T474A as rad9-2A) protein variant, which does not interact with Dpb11 [25]. Interestingly, both the failure to phosphorylate the H2A-S129 site and the rad9-2A mutation increase the viability of sae2Δ cells after one DSB, with the major contribution done by the mutation that abrogates the Rad9-Dpb11 interaction (Fig. 5A). Taking all these genetic results together, we concluded that the recruitment of Rad9 near the DSB site, mediated by its interaction with Dpb11 and partially with γ-H2AX, limits sae2Δ cells viability when a DSB must be repaired by SSA.
Consistently with our genetic evidence, we found an increased binding of Rad9 close to an irreparable DSB in sae2Δ cells by ChIP analysis (Fig. 5B), which correlates with the increased binding of Mre11 (Figs. 4A and 4B). Of note, the Rad9-2A protein variant does not bind near a break (Fig. 5B), supporting the viability data of the sae2Δ rad9-2A double mutant cells following one DSB (Fig. 5A). Moreover, Rad9 binding close to the break is only partially dependent on γ-H2AX and not by Dot1 (S5 Fig.), in agreement with cell viability of the sae2Δ h2a-S129A and sae2Δ dot1Δ double mutants (Fig. 5A).
Then we tested if the capability of Rad9 to form oligomers at the DNA damage site [29], [30], [64] was involved in inhibiting sae2Δ cells viability following a DSB. To this aim, we introduced a plasmid vector that expresses either the rad9-7xA allele or the RAD9 gene as a control, by transformation into rad9Δ and sae2Δ rad9Δ YMV80 derivatives. The Rad9-7xA protein variant cannot be phosphorylated at critical sites by upstream Mec1 and Tel1 kinases (see also Fig. 5C), and is unable to oligomerize [29], [64]. After plating cells in the presence of galactose to induce one DSB, we found that the expression of the Rad9-7xA protein variant rescues the lethality of sae2Δ cells, contrary to the wild type Rad9 (Fig. 5D). This result suggests that the oligomerization of Rad9 molecules is implicated in limiting SSA repair in sae2Δ cells. To further support this conclusion, we took advantage of the rad9-ΔBRCT-FKBP chimeric allele, which leads to the production of a truncated variant of Rad9 protein, in which the C-terminal BRCT domains are replaced with a FKBP tag [24]. It was shown that the Rad9-ΔBRCT-FKBP protein variant, which cannot form oligomers due to the absence of the BRCT domains, can dimerize in the presence of the small inducing molecule AP20187, binds chromatin and partially transduces the checkpoint signal (S6B Fig. and see also [24]). Consistent with our hypothesis, we found that the rad9-ΔBRCT-FKBP mutation does not rescue sae2Δ lethality in the presence of AP20187, while the viability in the sae2Δ rad9-ΔBRCT-FKBP double mutant cells is almost identical to the wild type value (Fig. 5E), further suggesting that the dimerization/oligomerization of Rad9 affects SSA repair.
It is now clear that DSB processing is a finely regulated process, which acts at the crossroad between HR and NHEJ recombination pathways. Indeed, as soon as a DSB is resected, homologous recombination pathways can be used to repair the break in lieu of NHEJ, with important implications for chromosome rearrangements and genome integrity.
Similarly to what seen in higher eukaryotes, three distinct nucleases cooperate to resect a DSB in S. cerevisiae. According to a model recently proposed for meiotic DSBs [65], Mre11, activated by Sae2 [44], introduces a nick near a DSB, triggering a bidirectional nucleolytic degradation of the 5′ strand: Exo1 and Dna2-Sgs1 resect the DNA in the 5′-to-3′ direction from the nick, while the Mre11 complex resects the DNA in the 3′-to-5′ direction toward the DSB ends. In G2/M blocked cells, it appears that the Exo1 and Dna2-Sgs1 pathways cannot actively resect a DSB starting from its ends, which are occupied by Ku70-Ku80 complex [1]. Indeed, it was suggested that the Mre11 activity might contribute to the removal of Ku complex, clearing the ends [2], [3], [11], [65], [66]. Importantly, in the absence of a functional Sae2, the Mre11-dependent DSB processing is compromised, and Ku-dependent NHEJ events and translocations increased [62]. In addition, Mre11 and Rad52 binding are, respectively, increased and reduced in sae2Δ cells (Fig. 4, and see [4], [57]), which are severely defective in repairing a DSB through SSA (Fig. 2, and see also [6], [42]). Moreover, sae2Δ cells cannot keep the DSB ends tethered, which was shown to be relevant for DSB repair (Fig. 4, and see [42], [58], [60]). Here, we show that the deletion of the RAD9 gene suppresses all these phenotypes of sae2Δ cells. Indeed, we found that deletion of RAD9 leads to a faster 5′–3′ resection both through the Exo1 and Dna2-Sgs1 pathways, but the Dna2-Sgs1 pathway becomes essential, in the absence of Sae2, to efficiently initiate DSB processing and repair through an SSA process that requires 25 kb DNA resection (Figs. 2 and 3). We also found elevated levels of Mre11 bound near an HO-induced break both in sae2Δ and sae2Δ sgs1Δ rad9Δ mutants, accordingly with a defect in Rad52 binding and DNA end-tethering (Fig. 4). The requirement of DSB end-tethering for SSA repair has never been explored before, however it is relevant to underline that Rad52 is important for end-tethering [58], and also our results indicate that a defect in end-tethering is linked with a failure to accomplish SSA repair. Further investigation will be required to fully understand the interplay between SSA and end-tethering. Interestingly, recent findings underlined a role of exonuclease processing of a DSB in maintaining broken chromosome ends in close proximity [61].
Taken all these findings together, we suggest that the prolonged binding of Mre11 near the break site may represent the critical barrier to efficiently initiate DSB resection, load Rad52 and establish end-tethering in the absence of Sae2, and it can be by-passed by a resection-based mechanism mediated by Sgs1-Dna2 in the absence of Rad9.
A similar role to remove Mre11 from a DSB site in sae2Δ cells was recently shown for Sgs1, in the absence of Ku70-Ku80 complex [56]. Indeed, deletion of KU70 suppresses sae2Δ cells sensitivity to low doses of CPT and other DSB inducing agents [1], [3]. Surprisingly, we did not see a rescue of sae2Δ cells lethality by deleting KU70 after a DSB that can be repaired through an SSA process between two homologous leu2 repeats 25kb far from each other, although deletion of RAD9 suppresses the sae2Δ ku70Δ double mutant (S7 Fig.). One possibility is that Rad9, bound near a DSB site, may limit the Sgs1-Dna2 activity starting from the break ends, leading to prolonged Mre11 binding. This might occur in cooperation with Ku complex, bound to the DSB ends, or rather it might represent a second distinct mechanism to limit DSB ends resection and DNA end-tethering. Alternatively, or in addition, Ku and Rad9 may limit DSB processing in different cell cycle phases. Indeed, the Ku complex acts on a DSB mainly in G1, while Rad9 acts predominantly in G2/M phase [36], [67], [68].
Genetic and biochemical evidence in Fig. 5 suggest that Rad9 protein dimerization and/or oligomerization, together with Rad9 interactions with Dpb11 and partially with γ-H2AX, are important to limit short-range resection and repair in sae2Δ cells. Indeed, Dpb11 is recruited on to the DNA lesion through the interaction with the 9-1-1 complex [28], and both the 9-1-1 complex and Dpb11 are recruited rapidly near a DSB site [69], likely at the ssDNA-dsDNA junction [70]. It is possible that the interactions with γ-H2AX, as well as with the histone H3 methylated at Lys79 by Dot1, become more important to recruit Rad9 in a distal region from the DSB site, contributing to slow down the long-range resection, which is not the limiting step in sae2Δ cells. This hypothesis is supported by the fact that DNA damage sensitivity of fun30Δ cells, that resect slower a DSB because of their inefficient Rad9 removal from chromatin flanking a DSB [37], is partially rescued in the absence of γ-H2AX or Dot1 [37], [63]. Of importance, deletion of DOT1 gene does not rescue sae2Δ cells (Fig. 5A). Notably, although Rad9 binding close to the break is not particularly elevated in wild type cells, it is enriched in sae2Δ cells (Fig. 5C). Consistent with our genetic evidence, Rad9 binding close to DNA ends depends on Dpb11, partially on the histone γ-H2AX, but not on the histone H3 methylated at Lys79 by Dot1 (Figs. 5B and S5). Possibly, these data are in agreement with the low amount of modified histones detected in chromatin within 1–2 kb of the break [22], [26], [71], [72], [73].
Overall, our genetic and molecular results suggest a model shown in Fig. 6, in which Rad9, in addition to its known role in inhibiting long-range resection, may affect the initial short-range processing of an HO-induced DSB. In fact, Rad9, once recruited close to a DSB end in G2 phase mainly through the interaction with Dpb11, limits the Sgs1 dependent resection starting from DNA ends, whenever Mre11 is blocked near the DNA ends. In the future it will be interesting to investigate whether Rad9 plays a similar role in limiting rapid and coincident resection of dirty radiation-induced DSBs, in cells lacking Sae2 and/or Mre11 [74].
We believe that our findings might have important implications for understanding how the genome stability is preserved, especially in higher eukaryotes, whose genomes are enriched of repeats and SSA events can be particularly frequent. In fact, it becomes clear that too-efficient DSB resection can lead to an excessive initiation of homologous recombination and accumulation of toxic DNA intermediates and rearrangements between repeats [16]. Moreover, DSB resection may lead to highly error-prone alternative ends joining (A-EJ) and MMEJ events [14], [16]. In this view, our results in yeast might help to understand recent finding in human cells at the molecular level, showing a role for 53BP1 in protecting from BLM and CtIP-Mre11 dependent A-EJ events and genome rearrangements [75].
Furthermore, our findings suggest that the functional interplay between 53BP1/Rad9 and Mre11 may also have a physiological relevance to protect from error-prone imprecise NHEJ events in genomic regions containing no repeats. It is also worth mentioning that the inactivation of 53BP1 was shown to potentiate homologous recombination and increase DNA damage tolerance of cancer-prone BRCA1 -/- cells [32], [76], [77], [78], with severe implications for therapeutic treatments.
In conclusion, we show novel insights on the structural barrier induced by Rad9, together with Dpb11 and γ-H2AX, to limit DSB processing and repair. The Sgs1-Dna2 pathway becomes essential to efficiently remove hypo-active Mre11 from a DSB site, in the absence of Sae2 and Rad9, triggering DSB resection and repair. The efficient removal of Mre11 from the DSB site is essential not only to switch to the more processive long-range resection, but also to allow an efficient recruitment of the recombination factor Rad52. This allows the maintenance of DSB end-tethering, which is an important prerequisite to complete repair, especially for those lesions that require extensive resection. These events increase in the absence of Rad9 and might contribute to accumulation of toxic HR events, leading to genome rearrangements and genetic instability.
All the strains listed in S1 Table are derivative of JKM139, YMV80 and yJK40.6. To construct strains standard genetic procedures of transformation and tetrad analysis were followed. Deletions and tag fusions were generated by the one-step PCR system [79]. For the indicated experiments, cells were grown in YP medium enriched with 2% glucose (YEP+glu), raffinose 3% (YEP+raf) or raffinose 3% and galactose 2% (YEP+raf+gal). All the synchronization experiments were performed at 28°C.
DSB end resection in JKM139 derivative strains was analyzed on alkaline agarose gels using a single-stranded RNA probe as described previously [36], [50].
TCA protein extract was prepared [80] and separated by SDS-PAGE. Western blotting was performed with anti-Rad53 (EL7), anti-HA (12CA5), anti-Rad9 (generously provided by N. F. Lowndes), and anti-actin using standard techniques.
Repair of an HO-induced DSB in YMV80 background was analyzed by a Southern blotting procedure described previously [39].
YMV80 derivative strains were inoculated in YEP+raf, grown O/N at 28°C. The following day, cells were normalized and plated on YEP+raf and YEP+raf+gal. Plates were incubated at 28°C for three days. Viability results were obtained from the ratio between number of colonies on YEP+raf+gal and YEP+raf. Standard deviation was calculated on three independent experiments.
JKM139 derivative strains were inoculated in YEP+raf, grown O/N at 28°C. The following day, after cell cycle block in G2/M by nocodazole, 2% galactose was added to one part of the culture to induce HO cut. After 2 hours of HO induction, cells were normalized and plated on YEP+raf and YEP+raf+gal. Plates were incubated at 28°C for three days. Viability results were obtained from the ratio between number of colonies on YEP+raf+gal and YEP+raf. Standard deviation was calculated on three independent experiments.
ChIP analysis was performed as described previously [69]. Input and immunoprecipitated DNA were analysed by quantitative PCR using a Biorad MyIQ2 system or a Biorad CFX connect. The oligonucleotides used are listed in S2Table. Data are presented as fold enrichment at the HO cut site (0.15 or 4.8 kb from the DSB) over that at the PRE1 locus on chromosome V, then normalized to the corresponding input sample. The obtained fold enrichment values were normalized to the fold enrichment of the t0 sample. Standard mean error (SEM) was calculated on three independent experiments.
Quantitative PCR (qPCR) analysis of DSB resection was performed accordingly to [52]. The oligonucleotides used are listed in S2 Table. The DNA was digested with the RsaI restriction enzime (NEB) that cuts inside the amplicons at 0.15 kb and 4.8 kb from the DSB, but not in the PRE1 control region on chromosome V. qPCR was performed on both digested and undigested templates using StoS Quantitative Master Mix 2X SYBR Green (Genespin) with the Biorad MyIQ2 PCR system. The ssDNA percentage over total DNA was calculated using the following formula: % ssDNA = {100/[(1+2ΔCt)/2]}/f, in which ΔCt values are the difference in average cycles between digested template and undigested template of a given time point and f is the HO cut efficiency measured by Southern blot analysis.
Cells of strains derivative from yJK40.6 background were grown in YEP+raf and blocked 3 hours in G2 with nocodazole. 160 µM CuSO4 was added one hour before inducing HO cut with galactose, accordingly to [58]. Samples taken at the indicated time were analysed with a fluorescence microscope. Cells with 2 LacI-GFP foci separated by more than 0.5 µm were considered defective in DSB end-tethering.
|
10.1371/journal.ppat.1002739 | HIV-Specific Antibodies Capable of ADCC Are Common in Breastmilk and Are Associated with Reduced Risk of Transmission in Women with High Viral Loads | There are limited data describing the functional characteristics of HIV-1 specific antibodies in breast milk (BM) and their role in breastfeeding transmission. The ability of BM antibodies to bind HIV-1 envelope, neutralize heterologous and autologous viruses and direct antibody-dependent cell cytotoxicity (ADCC) were analyzed in BM and plasma obtained soon after delivery from 10 non-transmitting and 9 transmitting women with high systemic viral loads and plasma neutralizing antibodies (NAbs). Because subtype A is the dominant subtype in this cohort, a subtype A envelope variant that was sensitive to plasma NAbs was used to assess the different antibody activities. We found that NAbs against the subtype A heterologous virus and/or the woman's autologous viruses were rare in IgG and IgA purified from breast milk supernatant (BMS) – only 4 of 19 women had any detectable NAb activity against either virus. Detected NAbs were of low potency (median IC50 value of 10 versus 647 for the corresponding plasma) and were not associated with infant infection (p = 0.58). The low NAb activity in BMS versus plasma was reflected in binding antibody levels: HIV-1 envelope specific IgG titers were 2.2 log10 lower (compared to 0.59 log10 lower for IgA) in BMS versus plasma. In contrast, antibodies capable of ADCC were common and could be detected in the BMS from all 19 women. BMS envelope-specific IgG titers were associated with both detection of IgG NAbs (p = 0.0001)and BMS ADCC activity (p = 0.014). Importantly, BMS ADCC capacity was inversely associated with infant infection risk (p = 0.039). Our findings indicate that BMS has low levels of envelope specific IgG and IgA with limited neutralizing activity. However, this small study of women with high plasma viral loads suggests that breastmilk ADCC activity is a correlate of transmission that may impact infant infection risk.
| In the absence of intervention, only about one third of infants born to HIV-1 infected mothers who are continuously exposed to maternal breast milk over prolonged periods get infected. This observation raises the possibility that immune factors in infected women play a role in limiting HIV-1 transmission. Identifying factors associated with reduced HIV-1 transmission risk will improve our understanding on the potential correlates of protection that should be the focus of generating effective immunogens and vaccination protocols. Here we assessed the functional role of breast milk antibodies in a group of women with high plasma viral loads and systemic NAbs and determined that overall, breast milk contains low levels of neutralizing antibodies when compared to plasma. In contrast, we observed a robust non-neutralizing activity in breast milk that was associated with infant infection status. Our study adds to the growing evidence of a potential role of non-neutralizing antibodies in limiting HIV-1 transmission and calls for more attention to this arm of the HIV-1 response.
| Breast milk (BM) can be a vehicle for transmission of various pathogens, but the risk of infant infection is balanced by the potential clinical benefit of BM, which provides significant passive immunity and protection against many infectious agents [1]–[4]. In the case of HIV-1, exposure to virus through breastfeeding accounts for almost half of the 30–40% of vertical transmissions that occur in untreated, breastfed infants of HIV-1 positive women [5]–[7]. Replacement feeding, avoidance of breastfeeding and reduced BM exposure by early weaning can significantly reduce BM transmission, however, these interventions have been associated with significant increase in infant morbidity and mortality [8]–[13]. Additionally, HIV-1 infected as well as exposed uninfected infants who do not breast feed have been shown to exhibit stunted growth [14], [15]. These observations highlight the challenges facing HIV-1 infected women in sub- Saharan Africa where prolonged breastfeeding could lead to HIV-1 transmission but no breast feeding could increase the risk of morbidity and mortality resulting in a diluted benefit of HIV-1 free survival [16]–[18]. Consequently, greater understanding of BM protective factors in HIV-1 infection may open promising new ways to make breastfeeding safe for infants born toHIV-1 infected women.
Approximately 15–20% of infants born to all HIV-1+ mothers in chronic infection acquireHIV-1 through BM [6], [7], [19], [20]. This relatively low infection rate despite continued exposure suggests that either BM infectivity is low or that antiviral factors in BM may play a role in modulating transmission and/or acquisition of HIV-1 via the oral mucosa. Indeed, antiviral innate immune factors present in BM such as alpha defensins, bile salt-stimulated lipase, lactoferrin, and mucins have all been associated with modulating the risk of BM transmission [21]–[23]. BM is also composed of both innate and activated adaptive immune cells, presumably derived from other mucosal sites such as the gut associated lymphoid tissue. Indeed, HIV-1 specific CD8 T cells and B cells have been reported in BM [24]–[26], but to date there have been no published studies that have explored the association between the functional immune responses in BM and risk ofHIV-1 transmission through breastfeeding.
Vertical transmission, including BM transmission, is characterized by a transmission bottleneck [27]–[39]. In mother- to-child transmission (MTCT), it has been suggested that this bottleneck is in part a result of selection pressure from Nabs because the viruses that are transmitted tend to be relatively insensitive to neutralization by maternal autologous antibodies (Abs), even in mothers who harbor viruses with a range of neutralization sensitivities[32], [39]. Consistent with the hypothesis that adaptive immunity plays a role in MTCT, several studies comparing levels of maternal plasma neutralizing antibody (NAb) titers reported that transmitting (T) mothers have lower levels of NAb in plasma compared to non-transmitting (NT) mothers [27], [32], [36]–[42] suggesting that maternal NAb may contribute to protection of the infant. However, the results of these studies are not consistent, particularly with respect to a role for NAb in protection by different routes of transmission [43]–[45]. Moreover, a recent study of passive Absin 100 HIV-1 exposed infants did not find evidence for a protective effect of broadly NAb on infant infection [46].
Until recently, most studies of BM HIV-1 Abs focused primarily on determining the association between the levels or presence of binding Abs to envelope (env) proteins and transmission. Several studies that have focused on BM IgG and IgA have showed no association between levels of these antibodies and transmission [47], [48]. Notably, infant infection status in these early studies was determined by serology and/or clinical manifestation of AIDS, a situation that could result in misclassification of infant infection status. A more recent study that determined infant infection by DNA PCR showed increased levels of BM IgA in T compared to NT women suggesting that, rather than providing protection, BM HIV-1 env specific soluable IgA, is associated with increased risk of transmission [49]. However, all these studies used subtype B env proteins, in some cases from lab adapted viruses to detect HIV-1 binding Abs despite being conducted in sub-Saharan Africa where such variants are not typical of transmitted strains of HIV-1 [49], [50]. Taken together, the results from BM binding studies have not provided clear evidence of a role of BM Abs in vertical transmission.
BM Abs could provide benefit by directly neutralizing the virus within the milk or by non-neutralizing mechanisms such as antibody dependent cellular cytotoxicity (ADCC)that target infected cells. This could result in reduced levels of infectious cell-free virus and BM infected cells, which are both correlates of BM transmission [51]–[54]. The potential of Abs in BMto neutralize HIV-1 and/or mediate ADCC has only very recently been examined, and in this study of ARV-exposed, subtype C-infected women in Malawi, NAbs were detected in about half of the BM samples while ADCC activity was present in all BM samples obtained at 1 month after delivery [55]. There have been no studies to-date looking at BMS samples obtained from untreated T and NT women, particularly in colostrum and early milk, which is relevant given that virus levels are highest in colostrum [51]and the majority of BM transmissions occur early in life [6], [20]. There has also been no study looking at how these BM Abs function in relation to MTCT.
We evaluated neutralizing, binding and ADCC activity in BMS or BMS-derived IgG and IgA and matched plasma from antiretroviral (ARV) naïve T and NT mothers with high plasma viral loads and systemic NAbs. Our data shows that BM Nabs are rare and their levels are significantly lower than in plasma. However, we report a high frequency of ADCC activity in BMS that was significantly higher in NT women compared to T women. These data suggest that BMADCC mediating Abs but not Nabs may play a role in modulating HIV-1 transmission.
Women enrolled in a randomized clinical trial comparing breastfeeding to formula feeding in Nairobi Kenya provided BMS samples used in this study [6]. Subjects received coded identification numbers at the clinic and therefore BMS samples were anonymous to laboratory personnel. The ethical review committees of the University of Nairobi, the University of Washington and the Fred Hutchinson Cancer Research Center approved this study and the Kenyan ministry of health gave permission for the original study to be conducted. The methods for enrollment, counseling and follow up have been described elsewhere [6], [51]. Briefly, HIV-1 positive women were enrolled at 32 weeks gestation and blood samples were taken for viral load and CD4 count testing. Maternal blood, breast milk samples, and infant blood samples were collected within the first week post-delivery, at 6 weeks, 14 weeks, 6 months and quarterly thereafter until 2 years. Infant HIV-1 status was determined using DNA PCR [56]. Breast milk samples were centrifuged to remove the lipid layer and the supernatant was stored at −70°C before being shipped either on dry ice or in liquid nitrogen to Seattle, Washington for long term storage at −70°C until use. Plasma and BM viral loads were determined using the Gen-Probe HIV-1 RNA assay (Gen-Probe, La Jolla, Calf) [51], [57]. Breastmilk samples used in this study were chosen as the first available breastmilk sample after delivery for each woman and the reported breastmilk viral loads are contemporaneous.
BMS IgG was purified using NAb Protein G spin columns (Pierce, Biotech, Rockford, IL), with minimal changes to the manufactures instructions. Briefly, 250 ul of heat-inactivated BMS was added to 250 ul of binding buffer and the mixture was added to a protein G column followed by incubation at room temperature (RT) with end over end mixing for 30 min. Thereafter, the column was centrifuged to obtain the IgG flow through (IgG stepFT) which was saved for subsequent IgA purification. The column with bound Ab was washed 3 times with 400 ul of binding buffer. Bound Ab was eluted with 1 ml of elution buffer (pH 2.8) and the eluate was neutralized by adding 100 ul of 1 M Tris. HCl (pH 8.5). Thus, the final purified IgG Ab was diluted 4-fold relative to the original BMS. The final eluted IgG and IgA was retained at a 1∶4 dilution of the original BMS and this was used undiluted in further neutralization assays. Coomassie blue staining (Simply Blue, Invitrogen) and ELISAs using Human IgG ELISA kit (E-80G) and human IgA ELISA kit (E-80A) (Immunology Consultants laboratory, Newberg, OR) were used to confirm the purity of Ab fractions.
BMS IgA was purified from the IgG step FT using the method outlined by Hirbod et.al with some modifications [58]. Spin columns (Thermo) were packed with 400 ul of immobilized jacalin (Pierce biotech, Rockford, IL) and washed 3 times with 400 ul of PBS to equilibrate. The column was then loaded with 500 ul of the IgG step FT and incubated on an end over end roller for 2 hours at RT. After incubation, the column was centrifuged and a final flow through (FT- fraction lacking IgG and IgA) was collected and stored for analysis. The column was washed 3 times with PBS followed by a 3-hour incubation with 500 ul of 1 M Melibiose to elute bound IgA. The column was further washed with another 500 ul of elution buffer to maximize recovery and bring the final dilution of purified IgA fraction to 1∶4 relative to the original BMS, similar to the IgG fraction. As before, coomassie staining and ELISA were used to confirm the purity of Ab fractions.
The subtype A HIV-1 envQ461.d1 was cloned directly from peripheral blood mononuclear cells (PBMCs) of a recently infected Kenyan woman as described previously [59]. Autologous PBMC and BM cell derived clones have either been previously described or were obtained using the same protocol [39], in some cases with modification of primers to allow amplification of the HIV-1 variant in that particular sample (primers are available upon request). Plasmid DNA encoding the env of interest and a plasmid encoding an env-deficient HIV-1 subtype A proviral DNA, Q23Δenv [60], were co-transfected into 293T cells at a 1∶2 molar ratio to generate pseudotyped viral particles as described [39], [61]. Virus was harvested 48 hrs post-transfection and the infectivity was determined by single round infection of TZM-bl cells as described [39]. Pseudoviruses were also generated using Q23Δenv and simian immunodeficiency virus clone 8 (SIV) [62]oramphotropic murine leukimia virus (MuLV)envelope clones [63].
Neutralization was assessed by determining infection of a reporter cell line, TZM-bl, as previously described [39]. Briefly, 500 infectious particles were incubated with 2-fold serial dilutions of heat inactivated plasma or BMS, purified BMSIgG or IgA fraction, FT fraction or media only in a total volume of 50 ul at 37°C for 1 hour. TZM-bl cells in 100 ul of growth medium containing 30 ug/ml of diethylaminoethyl-dextran were then added. After 48 hours, neutralization was determined by measuring β-galactosidase activity present in the TZM-bl cell lysate. For each virus/Ab combination, at least two independent experiments were performed. Each experiment was performed intriplicate for plasma and BMS or duplicate for purified BMSAb fractions. Median inhibitory concentrations (IC50s) were defined as the reciprocal dilution of plasma, BMS or purified antibody that resulted in 50% inhibition, calculated by interpolation of the linear portion of the neutralization curve on the log2 scale as previously described [39], [61]. Plasma and BMS samples were tested at 1∶100 and 1∶20 dilution respectively, while purified BMSAb fractions were tested at 1∶8 dilution (a 2-fold dilution of the recovered purified fractions that were diluted 4 fold during processing). For the purposes of analysis, in cases in which the IC50s were less than the lowest dilutions tested, the midpoint value between the lowest dilution and zero was assigned. IC50s from replicate experiments were averaged by the geometric mean. Here IC50s indicate the geometric mean IC50 estimates [64].
Human IgG ELISA kit (E-80G) and human IgA ELISA kits (E-80A) (Immunology Consultants laboratory, Newberg, OR) were used to determine the levels of total IgG and IgA in un-purified BMS and plasma samples according to the manufacturer's instructions.
HIV-1env specific ELISAs were performed using the protocol outlined by Sather et.al with minimal modifications [65]. Briefly, Immulon 2HB ELISA plates were coated with 25 ng/well of a HIV-1 subtype A Q461.d1 soluble trimeric gp140 protein purified as described in [66] in 0.1 M NaHCO3, pH 9.4 overnight at room temperature. Plates were blocked in phosphate buffered saline (PBS), supplemented with 10% dry milk and 0.3% Tween-20 for 1 hr at 37°C. Unpurified BMS and plasma samples were diluted in 10% dry milk, 0.03% Tween in PBS. For detection of HIV-1env specific IgG and IgA, BMS samples were diluted at 1∶100 and were titrated 2-fold up to a maximum dilution of 12,800. In cases where an end point titer could not be determined at this dilution, samples were diluted further up to a final dilution of 104,200. For HIV-1 env specific plasma IgG, samples were diluted at 1∶100,000 followed by a 2-fold titration up to a maximum dilution of 12,800,000 while for IgA samples were initially diluted 1∶200 followed by a 2-fold dilution up to 25,600. Samples were loaded in duplicate wells and incubated for 1 hr at 37°C. Plates were washed in a plate washer and bound IgG Ab was detected at 37°C for 1 hr with goat anti-human IgG- horseradish peroxidase (HRP) (Bio-Rad, Hercules, CA) diluted 1∶3000 while IgA was detected by goat anti human IgA HRP(Invivogen, San Diego, CA) diluted 1∶4000. Plates were developed with 50 ul of 1-Step Ultra TMB-ELISA solution (Pierce Biotech, Rockford. IL) and stopped with 50 ul 1 N H2SO4. Absorption at 450 nm was read on an EL808 Ultra Microplate Reader (Bio-TEK Instruments.inc). In this study, end point titer (EPT) was defined as the BMS or plasma reciprocal dilution at which the average OD value was greater than or equal to two times the average OD value of background.
The ability of BMS and their matched plasma to mediate ADCC activity was determined as described by Gomez-Roman et.al with a few modifications [67]. Briefly, CEM. NKr cells, a natural killer resistant cell line (AIDS Research and Reference Reagent Program, NIAID,NIH) were double stained with a membrane dye, PKH-26(Sigma, St. Louis, MO, USA) and a viability dye, carboxyfluorescein diacetate, succinimidyl ester (CFSE) (Molecular Probes, Eugene, OR, USA) as recommended by the manufactures. After staining, 1×105 cells were coated for 1 hr at RT with 1.5 ug HIV subtype Agp120 protein obtained from an infant in the Nairobicohortat 6 weeks post-infection (BL035) [39]. Coated cells were then washed once and resuspended in 1 ml of RPMI with 10%FBS. Five thousand coated or uncoated CEM. NKr cells were added to the appropriate duplicate wells containing 100 ul of 1∶100 or 1∶1000 heat inactivated BMS or plasma respectively. Similar experiments were performed using media only or HIV IgG (NIH AIDS Research, Germantown, MD, USA)as negative and positive controls, respectively. The antibody-target cell mixture was incubated at RT for 10 min to allow the antibody to interact with the antigen on the surface of target cells. Following incubation, 50 ul of effector cells (HIV negative donor PBMCs) were added to the mixture at an effector to target cell (E/T) ratio of 50∶1 and incubated for 4 hours at 37°C. For all 19 BMS and plasma samples, PBMCs from the same donor were used in parallel assays. Cells were then washed and fixed in 150 ul of 1% paraformaldehyde-PBS and stored at 4°C overnight. Fixed cells were analyzed within 24 hours of the ADCC assay using a BD LSRII instrument (Becton Dickinson, San Jose, CA, USA). Flow cytometry data was analyzed using Flojo version 9.4.6(Tree Star Inc, Ashland, OR, USA). ADCC percent killing was defined as the percentage of membrane labeled cells (PKH-26+) that had lost their viability dye (CFSE−) after subtracting two times the level of killing in the media only wells (background), as described in (67).
Odds ratios (OR) for assessment of associations between detection of HIV-1 specific and non-specific activity in BMS and transmission were estimated by Fisher's Exact Test. IC50sfor HIV positive and HIV negative controls were compared by one-sided t-test on the log2 scale. All comparisons of Ab total concentrations and HIV-1 env specific titers were based on paired t-tests on the log10 scale, noting that differences on the log scale were approximately normally distributed, and corresponding multivariate adjustments were by linear regression. HIV-1 specific titers among those with detected virus neutralization by BMS IgG and IgA were each compared to titers among those with undetected neutralization using Welch's t-test on the log10 scale. All correlations were measured by Pearson's product moment correlation coefficient (PPMCC), denoted r, with p-values based on the Student's t approximation for the distribution of the corresponding standardized test statistic. The relationship between maternal clinical correlates and BMS Ab neutralization, HIV-env specific binding titers and ADCC activity were each individually assessed by Welch's t-test with corresponding adjusted estimates by linear regression. Statistical analysis was performed using R 2.13 ISBN 3-900051-07-0 and STATA version 11 edition, (College Station, TX).
The goal of this study was to determine the presence and functional capacity of BM HIV-specific antibodies and to determine if they impact MTCT. Therefore, we selected women who had high plasma viral loads (greater than the cohort median of 4.6 log10)and thus were at increased risk of transmission. Among these women, we identified those who exhibited potent plasma NAb responses (Majiwa and Overbaugh, unpublished data) to maximize the chances of detecting BM NAbs. From this subset of women, we selected those that breast-fed for greater than 3 months to capture cases of BM HIV exposure to the infant. Women whose infants were HIV-1 positive before 6 weeks of life were excluded to ensure that transmission was as a result of BM and not late in-utero, or intra-partum exposure. An additional criteria was that women had available BMS samples collected at less than14 weeks after delivery because this early period is the window within which the majority of BM transmissions occur [6] and protein concentrations are highest [68], [69]. Nineteen women with a median CD4 count of 360 cells/uL met these criteria. The median plasma and BM viral loads were5.22 and 2.44 log10 respectively, an ∼2-log difference that was also observed in the larger cohort [51]. Nine of these women transmitted HIV-1 to their infants via BM at various time-points postpartum (Table 1).
The ability of heat inactivated BMS to neutralize virus bearing a highly sensitive env variant isolated from a Kenyan woman soon after her infection was determined. This heterologous HIV-1 subtype A env variant, Q461.d1, was chosen because >90% of plasma from individuals in the region showed detectable neutralization of this virus at a 1∶100 plasma dilution [70]. The results with plasma from 4 representative women are shown in Figure 1A. All4 plasma samples neutralized Q461.d1 with IC50 values of ∼500 or greater. Importantly, 50% inhibitory activity was not achieved when testing plasma samples against SIV suggesting that the neutralization response was specific to HIV-1. Overall, virtually all19 plasmas displayed potent HIV-1 specific neutralization, with IC50s ranging from 185 to 3144 (Table 1).
We could not detect HIV-1 neutralization in any of the BMS at a similar starting dilution as plasma (1∶100 data not shown). At a very low starting dilution (1∶4) there was substantial non-specific inhibition of SIV and MuLV and preliminary assays suggested potential cytotoxic effect of more concentrated BMS, as reported previously [71]. BMS was therefore tested at a starting dilution of 1∶20, hence 5× more concentrated compared to plasma. Results from BMS of 4 representative women against Q461.d1 and SIV are shown in Figure 1B. While a low level of inhibition of HIV-1 was observed with some BMS such as MJ776 and MP199, there was little difference in the magnitude of BMS neutralization of Q461.d1 and SIV in all 4 cases. Among all 19 women, 9 BMSs - 6 from T and 3 from NT women - showed HIV-1 inhibition with IC50 values ranging from 21–85; there was no detectable inhibition by BMS from 3 T and 7 NT women. BMS from the majority of women also inhibited SIV and MuLV pseudoviruses, with IC50 values ranging from 20–95 (Table 1). A paired comparison of BMS HIV-1 IC50s with the geometric mean of IC50s for corresponding negative control viruses (SIV and MuLV)showed that HIV-1 IC50s were not statistically greater than those of the negative controls (p = 0.44). This observation suggested that the majority of inhibition we observed with BMS was likely not due to HIV-1 specific Abs.
The presence of a non-specific inhibitor of HIV-1 in BMS could nonetheless be relevant to transmission risk. We thus examined the association between detection of non-specific activity and transmission and found that this relationship was not statistically significant (OR = 4.77; 95% CI: 0.51, 71.53; p = 0.17).
To determine what portion of the non-specific inhibition observed with unfractionated BMS was due to Abs versus other factors, we separately purified IgG and IgA Abs from BMS for use in the neutralization assays. Bands of the expected sizes for IgG and IgA were observed in the respective purified fractions by coomassie staining and cross contamination between Ab isotype fractions by total Ig ELISA was below detection (data not shown). Purified Ab fractions were tested at a starting dilution of 1∶8, which translated to a dilution 2.5 times higher than the most concentrated BMS we tested (1∶20 dilution). Using the purified BMS IgG fractions, neutralization of greater than 50%was detected in only 2 (subjects MJ776 and MP199) of 19 purified BMS IgG tested, with IC50s of 9.4 and 9.9 respectively. (These two examples are shown in figure 2A and a summary of the 19 in Table S1). Of these women MJ776 transmitted HIV-1 to the infant while MP199 did not. In contrast, there was no detection of neutralization by purified BMS IgA fractions tested (Results from 4 representative women are shown in figure 2B and a summary of the 19 in Table S1). Importantly, purified BMS IgG and IgA fractions did not inhibit viruses pseudotyped with SIV env including the two BMS IgG fractions from subjects MJ776 and MP199, which had detectable neutralization of virus pseudotyped with Q461.d1env (Figure 2A, B, and Table S1). The FT fraction, which contained undetectable levels of BM IgG and IgA both by ELISA and coomassie staining, (data not shown) retained the non-specific activity displayed by BMS (Table S1).
To ensure that we were not missing NAb responses by using a heterologous virus, we examined the ability of BMSAb fractions to neutralize autologous virus in a subset of the 19 women. BMS IgG and IgA Ab fractions and FT from a total of 8 women were each tested against 2 pseudoviruses bearing autologous env variants from blood [39]. Of the 8 women, 2 women both NTs, showed low potency neutralization of the blood-derived autologous virus to one of the two viruses tested. MM471 displayed low neutralization potency with anIC50 of 15against one of her autologous viruses when using IgG but not the IgA fraction (representative experiment is shown in figure 3A). In contrast, MA411 displayed low neutralization potency with an IC50 of 9 against one of the autologous virus with IgA but not with IgG fractions (a representative experiment is shown in figure 3B). BMS IgG and IgA fractions from the remaining six women, all Ts did not neutralize their respective autologous viruses above 50%. Autologous viruses for MJ776 and MP199 were not available for testing
The ability of plasma and BMS purified Ab to neutralize variants obtained from BM was also determined for two subjects MF535 (T) and ML055 (NT). Autologous plasma from MF535 and ML055 diluted at 1∶100 neutralized the respective BM viruses withIC50s of 152 and 718, respectively. In contrast, there was no detectable neutralization by BMAb fractions against these autologous BM viruses (data not shown).
To determine if low NAbs in BMS reflected lower total BM Ab levels, we measured the levels of total and HIV-1envspecific IgG and IgA Abs in BMS and compared them to plasma (Figure 4). The levels of total BMS IgG were 0.88 log10 lower than BMS IgA(p<0.0001) (Figure 4A, black symbols). This is in contrast to plasma, where the IgG levels were found to be 1.02 log10 higher than IgA (p<0.0001) (Figure 4A, grey symbols). There was a pronounced difference between the magnitude of total IgG in BMS and plasma with BMS total IgG being2.25 log10 lower than plasma IgG (p<0.0001). In contrast, the total IgA levels in plasma were only slightly higher than in BMS, with a modest 0.39log10 difference between BMS and plasma (p = 0.004). We found statistically significant correlation between total BMS IgG and plasma IgG (r = 0.67; p = 0.0034)while the levels of BMS total IgA correlated with total plasma IgA (r = 0.78; p = 0.0003). There was no significant correlation between BMS total IgG and BMS total IgA (r = 0.39; p = 0.10) (Table S2).
Next, we determined HIV-1 env specific IgG and IgA titers in unfractionated BMS and plasma against soluble gp140 protein derived from the subtype A variant, Q461.d1, that was used for the neutralization studies (Figure 4B). HIV-1 env specific IgG titers were obtained in 100% of BMS and plasma samples. In contrast, HIV-1 env specific IgA titers were obtained in 50% of BMS and 90% of plasma samples; the rest were below the cut off value for EPT as defined in this study. BMS HIV-1 env specific IgG titers were 1.96 log10 higher compared to env specific IgA (p<0.0001) (Figure 4B, black symbols). Similarly, HIV-1 env specific IgG titers in plasma were higher by 3.63 log10 when compared to the env specific IgA titers (p<0.0001) (Figure 4B, grey symbols). Overall, similar to what we found for total IgG levels, BMS HIV-1 env specific responses were 2.22 log10 lower compared to that in plasma (p<0.0001) (Figure 4B). For HIV-1 env specific IgA, the log10 difference between BMS and plasma was 0.59 (p = 0.0004) (Figure 4B). BMS HIV-1 env-specific IgG titers were correlated with plasma HIV-1 env specific IgG titers (r = 0.81; p<0.0001)and BMS total IgG (r = 0.76; p = 0.0003). There was no statistically significant correlation between BMS HIV-1 env specific IgG titers and BMS HIV-1 env specific IgA (Table S2). Similar to BMS HIV-1 env specific IgG titers and BMS total IgG, BMS HIV-1 env specific IgA titers and BMS total IgA levels were also positively correlated (r = 0.69; p = 0.015) (Table S2.)
We examined the relationship between the levels of HIV-1 env specific titers in BMS and detection of neutralizing activity. The three women with IgG neutralizing activity had a log10 IgG titer of 4.41 as compared to a mean of 3.83 among non-IgG-neutralizers (p = 0.0001). The one woman with IgA NAbs also had the highest IgA env specific titer, which was1.10log10 greater than the group median. (Figure. S1).
We determined the capacity of BMS binding antibodies and their matched plasma to mediate ADCC. The appropriate BMS and plasma dilution for the ADCC assay was determined by testing serial 10-fold dilutions of 4 representative BMS and plasma in the ADCC assay. The dilution that permitted detection of HIV-specific ADCC activity above background levels, but did not yield inhibition of ADCC activity that can occur with more concentrated samples [72] was chosen for testing (1∶100 for BMS and 1∶1000 for plasma). Using a single dilution also allowed us to test all 19 BMS and plasma samples with effector cells obtained from a single PBMC donor, which is critical for avoiding bias due to differences in effector cell activity observed from donor to donor. Overall, ADCC activity was detected in all BMS and plasma samples tested (Figures 5 A and B). BMS ADCC mediated killing ranged from 1–27% (median,15%) while that of plasma ranged from 16–36% (median, 24%). BMS ADCC activity was correlated with gp140 env specific IgG titers (r = 0.56, p = 0.014) (Figure 6). A log10 increase in gp140 titers was associated with an absolute increase of 9.3 in % ADCC mediated killing by BMS (95% CI: 2.18, 16.41; p = 0.013).
The relationship between maternal clinical correlates and BMS Ab neutralization, HIV-env specific binding titers and ADCC activity were each individually assessed. There was no statistically significant association between antibody titers and any of the clinical parameters examined. (Table S3).
There was no statistically significant association between detection of NAbs and infant infection (OR = 0.31; 95% CI: 0.0050, 4.94; p = 0.58). We observed a trend for statistical significance between infant infection and reduced BMSgp140 HIV-1 env specific IgG titers but not plasma titers (estimated mean log10 difference 0.35 95% CI: −0.07, 0.77; p = 0.098) in a univariate analysis (Figure 7A). This association was in similar direction after controlling for plasma viral load(p = 0.038). Importantly, NT women were more likely to have higher BM ADCC activity compared to T women (estimated mean % killing difference 6.89; 95% CI: 0.41, 13.37; p = 0.039) (Figure 7B). This relationship remained significant in a multivariate analysis controlling for plasma viral load (p = 0.011) and both plasma and BM viral load (P = 0.012). There was no association between BM RNA viral load and BM ADCC activity (p = 0.520) in these 19 women. There was also no significant difference between plasma ADCC in T and NT women (Figure 7B).
The potential of HIV-1 specific Absin BM to inhibit HIV-1 or impact transmission risk has not been well defined. Despite the fact that the levels of both IgG and IgA were low in BM compared to plasma, we observed a trend for inverse correlation between the levels of HIV-1 specific IgG and risk of infant infection in the 19 women examined here. The effect of these antibodies did not appear to be through neutralization, as only 4 of 19 women had any detectable neutralizing IgG or IgA Abs and there was no correlation between detection of NAb and risk of infant infection. Rather, the important functional activity of these antibodies was linked to ADCC activity, as there was a statistically significant inverse correlation between the levels of ADCC activity and risk of infant infection. These data suggests that antibodies capable of mediating ADCC may be one factor that impacts the risk of BM HIV-1 transmission.
We found that BM HIV-1 env-specific IgG titers were significantly higher than those of IgA but significantly lower when compared to IgG from matched plasma samples. A reduced IgA response at mucosal sites in HIV-1 infection is contrary to what is observed with mucosal responses to other pathogens but consistent with previous reports of a low HIV-1 specific binding IgA response in favor of IgG at various mucosal sites [73]–[76]. In general, low mucosal BM IgA might reflect an ability of HIV-1 to impair local immune responses as a means of evading the humoral immune system at the mucosal site. However, the observation that BM HIV-1 env specific IgG titers were correlated with total plasma IgG levels suggests that some of the BM IgG may originate from systemic circulation, a process that could help fight infection at the mucosal site.
Despite low HIV-specific antibody levels in BMS compared to plasma, antibodies capable of ADCC were detected in all BMS samples. We found that the capacity to mediate ADCC was associated with the levels of HIV-1 env specific IgG titers, which is in agreement with data from previous studies [55], [77]–[79]. This is perhaps not surprising given that envelope binding is a required step for ADCC activity measured in the assay used here. Using purified BMS antibodies from a subset of these women, we further confirmed that ADCC activity in BM was exclusively mediated by IgG (data not shown). Thus, IgG mediated ADCC can be detected in unfractionated breastmilk, which includes IgA and other factors, as well as with purified antibody. ADCC titers have previously been shown to be generally higher compared to NAbs titers in the same individual possibly due to the specificity required to overcome the constraints posed by env protein in a bid to escape neutralization and also the fact that virus neutralization requires that all of the functional trimers be occupied by at least one antibody [80], [81]. Thus it may be possible to elicit high levels of antibodies capable of ADCC using an HIV-specific immunogen even in cases where neutralizing responses are limited.
BMS ADCC activity was significantly greater in NT compared to T women, suggesting a possible role in impacting infant infection. The mechanism by which BM ADCC might reduce transmission remains to be determined. ADCC would be expected to lead to effective clearance of infected cells. Given that the levels of HIV-infected cells in BM are correlated with transmission risk [52], it is plausible that HIV-specific ADCC responses within BM may act through reducing cell-associated viral transmission.
Other studies have implicated antibodies capable of ADCC in providing protection from infection and/or controlling an established infection. Several studies have shown that de novo ADCC responses to HIV and SIV infection are correlated with better viral control in chronic infection and/or clinical outcome. [77], [78], [82]–[85]. Vaccine-induced ADCC responses have also been correlated with reduced viral loads following SIV challenge [78], [79], [86]–[88], supporting a potential role of Fc-mediated antibody responses in blunting a new infection in SIV-infected macaques. A study by Forthal et al. also provided evidence that antibody-dependent cell-mediated virus inhibition, which is a measure of ADCC in combination with other antiviral activities, was correlated with infection rate in the Vax004 vaccine trial, although ADCC alone was not directly examined in this study [89]. In addition, studies of passive immunization using HIV monoclonal antibodies in macaques suggest that FcγR binding is required for optimal protective efficacy [90]. These findings support a potential role for antibodies that act through ADCC in providing protection from infection in the non-human primate model. The current study is the first that reports an association between HIV-specific ADCC activity and risk of HIV infection in humans.
This is the first study to examine BMS HIV-1 specific IgG and NAbs in relation to transmission risk using a relevant HIV-1 env representing recently transmitted virus from the dominant subtype in the population. This may explain our ability to detect a trend in association between binding antibodies and transmission, which was not seen in prior studies using other env proteins less representative of viruses in the study population to measure binding [47], [48].
We used the same highly neutralization sensitive (tier 1B) subtype A HIV-1 env representing the dominant subtype in the population under study to optimize our chances of detecting NAbs in BMS. Importantly, plasma from all subjects had a potent NAb response against this virus, indicating that all subjects had generated NAbs capable of specifically recognizing this test virus. Only 4 BMS had Abs that could neutralize >50% of either heterologous or autologous blood-derived viruses and the presence of HIV-1 specific NAbs was not associated with infant infection. The neutralizing activity was observed in women with higher levels of total IgG Abs in BMS. Therefore, it is possible that generally low IgG and IgA titers in BM might explain the limited neutralization capacity displayed by BM Abs.
The results of our study, showing low levels of HIV-1 env specific NAbs in BMS, are consistent with another recent study of BM HIV-1 NAbs [55]. In this study of a NVP-treated, clade C infected cohort, the levels of NAbs and HIV-1 env specific IgG were low in BM collected at 4 weeks post-delivery compared to plasma. We observed similarly low NAb levels in the breastmilk of ARV naïve women in a cohort that was enrolled prior to the availability of ARVs for prevention of MTCT [6]. Thus, collectively these studies indicate that the level of HIV-1 specific NAb are low in both early and mature milk, in both treated and untreated women and this is true no matter the infecting HIV-1 subtype.
We detected non-specific inhibition of HIV-1 and unrelated viruses (SIV, MLV) with several unfractionated BMS samples. This observation is perhaps not surprising because innate factors in BM such as defensins, lipids and lactofferin have documented activity against many viruses including enveloped retroviruses [4]. The ability of unfractionated BMS to inhibit HIV-1 in the in vitro TZM-bl assay used here did not correlate with risk of infant infection.
There are several limitations to our study, most notably the fact that we focused on a select group of women with high viral load and systemic NAbs in order to optimize our chances of detecting NAbs and to examine antibody levels in relation to transmission risk. Thus it is unknown if these findings are applicable to women with low viral loads or low systemic NAbs levels. Interestingly, a correlation between ADCC activity and viral control in SIV- infected macaques was only observed when animals with low viral load were excluded [86]. These authors suggested that a threshold of antigen may be needed to elicit robust ADCC. Certainly, larger studies using relevant env antigens to examine HIV-1 specific BM antibody responses in other populations will be needed to verify these findings and determine if the findings apply to women with lower viral levels and/or systemic NAb responses. In addition, while we focused on breastmilk antibodies in relation to post-partum transmission, there could be some misclassification of time of infection in this study. Specifically, the cases of transmission examined here were all cases of relatively early post-partum transmission and we cannot exclude that some were the result of intrapartum transmission, where BM antibody levels would be less relevant. Finally, while we did not see an association between BM viral RNA levels in this small study, but this does not rule out a relationship between ADCC and the cellular viral reservoir. Larger studies that include cell-associated virus levels and ADCC activity will be needed to clarify this issue.
In conclusion, we found that the capacity of BM to neutralize heterologous and autologous viruses obtained from blood and BM is limited. This observation can be explained in part by the low titers of Abs in BM compared to plasma in general, particularly IgG. It is unclear if such low NAb levels could play a role in protection, but no association was observed in this small study. However, the association between HIV-1 env specific IgG titers and ADCC activity with infant infection suggest that BM Ab could be playing some role in modulating infection through non-neutralizing mechanisms. To the best of our knowledge, this is the first study to report a positive association between BM transmission and ADCC capacity in BM. If these results are verified in a larger study of MTCT, then it would suggest that immunogens tailored at enhancing BM Abs capable of ADCC might be of potential benefit, particularly to HIV-1 infected women with high viral loads, who are at the greatest risk of transmission.
|
10.1371/journal.pntd.0007442 | Linear growth in preschool children treated with mass azithromycin distributions for trachoma: A cluster-randomized trial | Mass azithromycin distributions have been shown to reduce mortality among pre-school children in sub-Saharan Africa. It is unclear what mediates this mortality reduction, but one possibility is that antibiotics function as growth promoters for young children.
24 rural Ethiopian communities that had received biannual mass azithromycin distributions over the previous four years were enrolled in a parallel-group, cluster-randomized trial. Communities were randomized in a 1:1 ratio to either continuation of biannual oral azithromycin (20mg/kg for children, 1 g for adults) or to no programmatic antibiotics over the 36 months of the study period. All community members 6 months and older were eligible for the intervention. The primary outcome was ocular chlamydia; height and weight were measured as secondary outcomes on children less than 60 months of age at months 12 and 36. Study participants were not masked; anthropometrists were not informed of the treatment allocation. Anthropometric measurements were collected for 282 children aged 0–36 months at the month 12 assessment and 455 children aged 0–59 months at the month 36 assessment, including 207 children who had measurements at both time points. After adjusting for age and sex, children were slightly but not significantly taller in the biannually treated communities (84.0 cm, 95%CI 83.2–84.8, in the azithromycin-treated communities vs. 83.7 cm, 95%CI 82.9–84.5, in the untreated communities; mean difference 0.31 cm, 95%CI -0.85 to 1.47, P = 0.60). No adverse events were reported.
Periodic mass azithromycin distributions for trachoma did not demonstrate a strong impact on childhood growth.
The TANA II trial was registered on clinicaltrials.gov #NCT01202331.
| Mass distribution of a single dose of the broad-spectrum antibiotic azithromycin twice per year to pre-school children in Sub-Saharan Africa has been shown to reduce childhood mortality. The mechanism by which azithromycin reduces mortality is currently not clear, especially since the antibiotic is not targeted to sick children but rather given to all children in the community whether or not they have an infectious disease. In this study, we report the height and weight of children enrolled in a trial in Ethiopia in which communities were randomized either to twice annual mass azithromycin distributions for blinding trachoma or to no treatments. After accounting for age and sex, children from azithromycin-treated communities were on average slightly taller at the 12- and 36-month study visits than those from untreated communities, but the difference was not statistically significant.
| Undernutrition is thought to contribute more to the global burden of disease than any other risk factor.[1] Poor nutrition potentiates the effects of infections such as diarrhea, respiratory infections, and malaria, leading to worse outcomes and higher mortality.[2, 3] Infectious diseases in turn lead to poor growth.[4] It is conceivable that antibiotics could have an important role for breaking this cycle of malnutrition and infection. For example, a randomized trial demonstrated less stunting and underweight in HIV-infected children who took daily co-trimoxazole compared to those taking placebo.[5] This idea is not new; antibiotics have long been thought to be effective growth promoters for animal husbandry.[6]
Mass azithromycin treatments have recently been shown to reduce childhood mortality in sub-Saharan Africa, although the causal pathway for the mortality reduction is unclear.[7, 8] Azithromycin is a broad-spectrum antibiotic with efficacy against a wide array of pathogens that cause respiratory disease, diarrhea, and malaria, so it is possible that azithromycin prevents mortality by directly clearing these infections.[9–11] Alternatively, if azithromycin caused height and weight gain in children, it is possible that the improved growth could be partly responsible for the survival benefit.
In a recent cluster-randomized trial, we randomized communities that had been treated with four years of mass azithromycin for trachoma to either continued antibiotics or cessation of treatment. Cluster-randomization, which was chosen for the primary trachoma outcomes, allowed assessment of both the direct and spillover effects of antibiotics.[12] We recognized that this trial provided an opportunity to assess whether mass azithromycin distributions had an effect on childhood growth, and consequently added height and weight as secondary outcomes. We compared anthropometric measurements at the individual level to determine whether mass azithromycin increased childhood growth.
The study had approval from the University of California, San Francisco; Emory University; the Ethiopian Ministry of Science and Technology; and the Food, Medicine, and Health Care Administration and Control Authority of Ethiopia. The study was carried out in accordance with the Declaration of Helsinki and overseen by a Data Safety and Monitoring Committee appointed by the National Institutes of Health-National Eye Institute. Verbal informed consent was obtained in Amharic from community leaders before randomization and from the guardians of all children post-randomization; verbal consent was approved by all institutional review boards and was used due to the high levels of illiteracy in the study area.
This study describes a secondary outcome from a parallel-group cluster-randomized clinical trial for trachoma (TANA II; clinicaltrials.gov #NCT01202331) performed in the Goncha Siso Enese woreda (district) of the Amhara region in Ethiopia.[13] The trial was conducted in communities that had been treated with 8 biannual mass azithromycin distributions for trachoma as part of an earlier trial (TANA I; clinicaltrials.gov #NCT00322972).[14] In the initial TANA I trial (June 2006 until November 2009), 72 contiguous subkebeles (sub-districts) were randomized to 1 of 6 different trachoma treatment strategies.[14] Rural subkebeles were eligible for enrollment if they were located within a 3-hour walk from the farthest point accessible to a four-wheel drive automobile. As part of the TANA I trial, 12 subkebeles were randomized to receive biannual mass azithromycin distributions (every 6 months, ±1 month). Each subkebele consisted of approximately 4–6 government-defined demographic units known as state teams; all state teams in the subkebele were treated identically. In TANA II (November 2010 until May 2013), two randomly selected state teams from each biannually treated subkebele were randomized to either continued biannual mass azithromycin (N = 12 state teams) or cessation of antibiotics (N = 12 state teams). Anthropometric indices were measured as a secondary outcome in these 24 state teams 12 and 36 months after randomization, providing a randomized controlled trial comparison of childhood growth in communities treated with mass azithromycin distributions versus no treatment. No changes were made to the anthropometry portion of the study after its addition to the main trial. The trial was reported according to CONSORT guidelines (S1 Checklist). Details of the study design were pre-specified in a trial protocol (S1 Protocol).
This study employed pair-matched cluster-randomization with matching based on the TANA I subkebele. A biostatistician (TCP) used the statistical package R (R Foundation for Statistical Computing, Vienna, Austria) to randomize 12 pairs of state teams from each of 12 subkebeles, with one member of the pair allocated to intervention and the other to control. A study coordinator enrolled the communities and participants and assigned the intervention. Allocation was concealed by enrolling all communities before randomization and administering the intervention to the entire community population. Although study participants were not masked to their cluster’s treatment assignment, the study personnel responsible for anthropometric assessments were not informed of the treatment allocation or study hypothesis.
Local health extension workers performed a population census each year to enumerate all community members eligible for treatment. Because birthdates are imprecise in this part of Ethiopia, census-takers recorded age in years. During scheduled mass treatment visits, these same health workers offered a single dose of directly observed oral azithromycin (20mg/kg height-based approximation for children, 1g for adults) to all community members except for children under 6 months of age, self-reported pregnant women, and those known to be allergic to macrolide antibiotics, each of whom were instead offered a six-week course of twice daily tetracycline ophthalmic ointment. Each community was treated twice per year (Fig 1). Antibiotic coverage during mass treatments was assessed from the preceding census. Although study participants from all communities continued to receive routine government health services during this time, no other studies or interventions were performed during the study period. No adverse events were reported through routine passive surveillance activities; active monitoring of adverse events was performed during the trial and is reported elsewhere.[15]
We performed anthropometric assessments after the month 12 and month 36 treatments. All children aged 0–59 months identified from the most recent census were offered height and weight measurements, conducted in a central area in each state team. We recruited anthropometrists from the local communities, and conducted a 2-day training session using methods recommended by the World Health Organization (WHO). Anthropometry teams demonstrated high levels of reproducibility for all measurements.[16] We used a portable stadiometer (Shorr Productions, LLC, Olney, MD, USA) to measure height or length and a Seca 874 floor scale (Seca GmbH & Co. KG, Hamburg, Germany) to measure weight. Each measurement was taken in triplicate, with the median used as the official measurement.
The pre-specified primary analysis for the anthropometric outcomes was a post-test only comparison of height adjusted for age and sex in children not old enough to be eligible for treatment during the first trial (i.e., aged 0–36 months at the 12-month visit and 0–59 months at the 36-month visit). Repeated measures for height were modeled as a function of treatment arm, time, age in years (as a continuous variable rounded to whole years), and sex in a mixed effects linear regression, with state team and individual modeled as random effects. Heteroskedasticity was modeled by allowing independent residual errors over each year of age. The p-value was estimated with a Monte Carlo permutation test stratified by subkebele to account for the randomization strategy (10,000 permutations of the likelihood ratio comparing models with and without the treatment term). Intraclass correlation coefficients (ICCs) were calculated from the model to assess cluster-correlation of the primary outcome. Similar models were constructed for the secondary underweight analysis (weight adjusted for age and sex). We purposefully did not use z-scores in the primary analysis given the imprecision of reported ages in the study area. For all outcomes, we performed intention-to-treat analyses.
Sample size calculations for the trial were based on the primary trachoma outcomes and set at 12 communities per arm; including 18 children per community would provide 80% power to detect a 1 cm difference between the two arms assuming an average height of 90 cm (standard deviation 10 cm), ICC of 0.02, correlation coefficient of 0.9 for the relationship of height with age and sex, equal cluster sizes, and a two-sided alpha of 0.05.[17] All analyses were performed using Stata 14.2 (Statacorp, College Station, TX) except for permutation tests, which were done in R 3.4.0 (R Foundation for Statistical Computing, Vienna, Austria).
The study was conducted from November 2010 until May 2013. Characteristics of the two treatment arms were balanced at baseline (Table 1). The biannual treatment group received 7 mass azithromycin treatments during the continuation trial; the control group received none. Table 2 shows the median antibiotic coverage at each distribution for the under-5 age group.
We obtained anthropometric measurements on 282 of 387 (73%) eligible children aged <3 years at the 12-month visit and 455 of 591 (77%) eligible children aged <5 years at the 36-month visit, with 207 children having measurements at both visits. As shown in Table 3, children who were eligible but did not participate in the anthropometric assessment tended to be younger than those who participated.
The mean height and weight in each treatment arm at the two time points are shown in Table 4, stratified by age. After adjusting for age, sex, and study visit, the mean height was 84.0 cm (95%CI 83.2–84.8) in the azithromycin-treated communities and 83.7 (95%CI 82.9–84.5) in the untreated communities. In the primary pre-specified analysis adjusted for age, sex, and time point, children in the azithromycin-treated communities were on average 0.31 cm taller (95%CI -0.85 to 1.47) than those in untreated communities (P = 0.60). The ICC derived from the main statistical model suggested mild clustering of height measurements within communities (ICC 0.05, 95%CI 0.02 to 0.13).
The weight outcome is summarized in Table 4. After adjusting for age, sex, and study visit, the average weight was 10.8 kg in the azithromycin-treated communities and 10.7 in the untreated communities—a non-significant difference (mean weight 0.09 kg heavier in the azithromycin arm, 95%CI -0.20 to 0.39; P = 0.54; ICC 0.07, 95%CI 0.03 to 0.15). Additional sensitivity analyses using all <5 year-old children at both outcome visits (i.e., including children from the 12-month visit who had been treated in the previous trial) had similar results (S1 File).
We failed to find an association between mass azithromycin distributions and childhood growth. Although on average children from communities treated with biannual mass azithromycin distributions over a 3-year period were slightly taller than children from untreated communities, the differences between the two treatment groups were not consistently statistically significant across different regression model parameterizations and of small magnitude.
This trial assessed the hypothesis that azithromycin functions as a growth promoting agent for preschool children. The specific mechanisms through which antibiotics may promote growth are unknown, though several theories have been proposed.[18] Antibiotics could treat or prevent infections that would otherwise divert metabolic resources. Antibiotics may also alter intestinal absorption by clearing pathogenic organisms and changing the composition of the gut microbiome, both of which may ultimately mitigate the role of bacterial infection in environmental enteropathy.[19, 20] Alternatively, antibiotics like azithromycin are known to have direct anti-inflammatory properties, which could potentially affect growth.
It is conceivable that improvements in growth could translate into improved childhood survival. It is thought that growth and nutrition are important modulators of infectious diseases. Better nutrition may enhance the ability of the immune system to clear infections, and thus make a child less susceptible to severe complications of infection. Indeed, growth promotion is one theory for the improved survival observed following mass azithromycin distributions.[7, 8] While this study does not provide evidence that mass azithromycin impacts childhood growth, it is still possible that azithromycin treatments affect growth but at a magnitude smaller than could be detected by the present study. For example, the results of the present study are consistent with a meta-analysis that estimated antibiotics to confer an additional 0.04 cm of linear growth per month.[21]
Previous randomized trials have studied the impact of antibiotics on childhood growth in the context of a specific disease, such as severe acute malnutrition, HIV, diarrhea, cystic fibrosis, or vesicoureteral reflux.[22–25] Notable among these is a placebo-controlled trial of daily prophylactic co-trimoxazole therapy for HIV-infected children in Zambia that found improved linear growth and decreased mortality in antibiotic-treated children.[5] While many of the other studies failed to find an effect of antibiotic therapy on height measurements, they were of relatively short duration and focused on changes in weight instead of height.[22–25]
Other trials have assessed whether community antibiotic distributions affect growth parameters. Two placebo-controlled studies performed in Guatemala, one of which treated school-aged children with aureomycin and another which treated preschool children with metronidazole, found improved height and weight metrics in children treated with antibiotics. However, these studies may be subject to type I error since they allocated a single cluster of children to the antibiotic intervention but performed analyses at the individual level.[26, 27] Several recent cluster-randomized trials testing more frequent vs. less frequent mass azithromycin distribution strategies for trachoma have failed to detect a difference in height or weight between the different dosing groups.[17, 28, 29] The control groups in those studies were not ideal in that they too received antibiotics—albeit at lower frequencies—and thus may have received a benefit of antibiotic therapy. The present study is an improvement over these previous reports in that no programmatic antibiotics were given in the control group, providing a better assessment of the effectiveness of antibiotics alone.
Mass antibiotics provide selection pressure for antibiotic-resistant organisms, which increases the community prevalence of resistance.[30] Although the clinical importance of such resistance is unclear given the low usage of macrolide antibiotics in developing countries, the potential for antibiotic resistance should be taken into account wherever implementation of mass azithromycin distributions is considered.[31]
Several limitations to this study should be noted. We conducted a post-test only analysis with two follow-up visits. Baseline and more frequent anthropometric measurements would have increased the precision and added to the statistical power of the study. Ages were not known with great accuracy in this study, since families in this region of Ethiopia do not keep health cards and do not record dates of birth. Imprecision in age could have biased the study if age was differentially recorded between treatment arms or if cluster-randomization left an imbalance of ages in the two groups. This study was cluster-randomized since it was performed in the context of community antibiotic treatments for trachoma. An individually randomized trial could have been powered to detect an even smaller difference between children treated with and without antibiotics. On the other hand, the cluster-randomization better approximates the effectiveness of programmatic azithromycin and allows for the possibility of both direct and spillover effects of antibiotics.[12] Finally, this study was conducted in a region of Ethiopia with hyperendemic trachoma that had already received several rounds of mass azithromycin distributions. The generalizability of the findings to an antibiotic-naïve population or a population with milder or no trachoma is unclear.
In conclusion, mass azithromycin distributions have proven extremely effective at reducing the burden of ocular chlamydia and may also have important ancillary benefits, like reducing the prevalence of respiratory infections, diarrhea, malaria, and skin infections—as well as childhood deaths.[9–11, 32] Although we hypothesized that childhood growth promotion might be an additional benefit of mass azithromycin distributions, the present study was unable to provide strong evidence of such an association.
|
10.1371/journal.pgen.1007043 | Combinatorial action of Grainyhead, Extradenticle and Notch in regulating Hox mediated apoptosis in Drosophila larval CNS | Hox mediated neuroblast apoptosis is a prevalent way to pattern larval central nervous system (CNS) by different Hox genes, but the mechanism of this apoptosis is not understood. Our studies with Abdominal-A (Abd-A) mediated larval neuroblast (pNB) apoptosis suggests that AbdA, its cofactor Extradenticle (Exd), a helix-loop-helix transcription factor Grainyhead (Grh), and Notch signaling transcriptionally contribute to expression of RHG family of apoptotic genes. We find that Grh, AbdA, and Exd function together at multiple motifs on the apoptotic enhancer. In vivo mutagenesis of these motifs suggest that they are important for the maintenance of the activity of the enhancer rather than its initiation. We also find that Exd function is independent of its known partner homothorax in this apoptosis. We extend some of our findings to Deformed expressing region of sub-esophageal ganglia where pNBs undergo a similar Hox dependent apoptosis. We propose a mechanism where common players like Exd-Grh-Notch work with different Hox genes through region specific enhancers to pattern respective segments of larval central nervous system.
| Specification of the head to tail axis of the developing central nervous system is carried out by Hox genes. Hox mediated programmed cell death of the neural progenitor cells plays an important role in specification of this axis, but the molecular mechanism of this phenomenon is not well understood. We have studied this phenomenon in abdominal and subesophageal regions of larval central nervous system of Drosophila. We find that different Hox genes use a combination of common players (Extradenticle, Grainyhead and Notch) but employ region specific enhancers to cause progenitor cell death in different segments of developing central nervous system.
| Apoptosis is used to eliminate defective and/or dispensable cell types during development of an organism. Removal of excess cells from the developing tissue is critical for its final size, shape and functionality in the whole organismal context [1–3]. The central nervous system (CNS) also relies heavily on apoptosis for its patterning and development [4–7].
Equally important and one of the earliest steps in development of CNS is specification of anterior posterior axis (AP axis). AP axis specification is carried out by Hox genes [8–11] and their TALE homeodomain containing cofactors; Extradenticle (Exd) and Homothorax (Hth) in Drosophila [12,13]; and Pbx and Meis in vertebrates [14]. Hox genes pattern CNS by regulating proliferation, differentiation and apoptosis of different cell types [15–22] [23,24]. Their role in developmental apoptosis of CNS has been reported earlier, both in Drosophila [16–18] as well as in vertebrates [21,22], but the molecular mechanism of the cell death in neural stem cells as well as their progeny are not known.
In Drosophila, Hox mediated apoptosis of neural stem cells (neuroblast-NB) and their progeny happens both in embryonic and post-embryonic (larval) stages of development [16–18,23–30]. This apoptosis is mediated through activation of RHG family of genes (reaper, hid, grim and sickle) [31–34], but the precise molecular mechanism tying a particular Hox gene to death of a specific cell type is not known. In larval stages, Hox mediated NB apoptosis has been reported for Labial (Lab), Deformed (Dfd), Sex combs reduced (Scr) and Abdominal-A (AbdA) expressing regions of CNS [16,23,24]. AbdA mediated larval NB (postembryonic NB-pNB) apoptosis is so far the best characterized of all [16,35], yet the precise molecular details of the same are lacking.
The segments A3-A7 of embryonic ventral nerve cord has 60 NBs each (30 per hemi-segment). Following an embryonic wave of AbdA mediated NB apoptosis [17], majority of NBs undergo cell death, leaving behind only 3 NBs per hemisegment. Following embryonic phase of apoptosis, abdominal NBs stop expressing AbdA and enter quiescence by the end of embryogenesis. All the 3 NBs have specific locations and developmental potential and are designated as NB5-2 (Ventromedial-Vm), NB5-3 (Ventrolateral-Vl) and NB3-5 (Dorsolateral-Dl) in embryonic stages, and Lineage-6, Lineage-5 and Lineage-9 in larval stages respectively [36,37]. These 6 NBs (3 per hemisegment) exit quiescence in early third instar larval (L3) stage (66–72 hours after egg laying-AEL) and divide for different durations, following which an asynchronous increase in AbdA expression in these cells causes their apoptosis and removal from CNS over a course of next 48 hours [16,35]. The pNBs mutant for either abdA, or grainyhead (grh-a basic helix-loop-helix (bHLH) transcription factor) or RHG genes [as seen in genomic deletion-Df(3L)H99] escape apoptosis, underlying their individual importance in cell death [16,35]. The mechanistic details of how Grh and AbdA mediate pNB apoptosis through RHG genes is not known.
Similarly, in subesophageal ganglia (SEG) of larval CNS (which expresses Dfd, Scr and Antennapedia), 36 NBs (18 segmental pairs) are reported in second instar larval (L2) stage. Out of these 36 pNBs (18 pairs), 10 pNBs (5 pairs) are found in Dfd expressing region of SEG [24], here on referred to as Dfd-SEG. Four out of these 10 pNBs undergo Dfd mediated apoptosis as larva progresses from L2 to L3 stage (illustrated later in a Figure and detailed in S1 Text) [24]. The molecular mechanism of this apoptosis and role of Grh in this phenomenon is yet to be investigated.
A genomic deletion analysis had identified a 22kb region referred to as NBRR (NeuroBlast Regulatory Region), which contains NB specific enhancer for apoptotic genes [38]. A 5 Kb subfragment (enh-1) of NBRR, has been suggested to be the apoptotic enhancer required for embryonic NBs. However, it is not clear whether the same enhancer functions to cause larval NB apoptosis [25]. This study also suggests that pulse of AbdA expression responsible for larval NB apoptosis, is initiated in response to activation of Notch signaling in these cells [25].
In this report, we have investigated the molecular basis of Hox mediated larval pNB apoptosis. We analyzed 22 Kb NBRR and have narrowed down the larval abdominal apoptotic enhancer to a 1Kb region of the genome. Our experiments suggest that AbdA, Exd, Grh and Notch transcriptionally contribute to regulation of RHG genes, and Exd has a Hth independent role in this apoptosis. In vitro experiments suggest that AbdA and Grh physically interact with each other, and Grh-AbdA-Exd assemble a tetrameric complex with DNA on some of the binding motifs in the apoptotic enhancer. In vivo mutagenesis of all the motifs suggest, that they are important not for the initiation but for the maintenance of the enhancer activity, and consequently expression of RHG genes. Our analysis of the enhancer mutant for Su(H) binding sites reveal that Notch signaling also has a direct input in the maintenance of the enhancer activity and hence RHG genes in abdominal pNBs.
Subsequently we show that Dfd mediated pNB apoptosis in SEG use same players (Hox-Exd-Grh-Notch) but employ a different enhancer located outside NBRR.
Taken together, this study describes a common mechanism of RHG gene regulation in pNBs undergoing Hox dependent apoptosis. Wherein combination of Exd-Grh-Notch are employed by specific Hox genes, to carry out apoptosis in different regions of developing CNS through separate spatial enhancers.
23 Kb genomic region (including 22Kb NBRR and additional 500bps on either side; Fig 1A) was divided into 5 over lapping fragments (Fig 1A). LacZ reporter lines were generated for these fragments and analyzed for their expression in larval CNS (Fig 1, S1 Table and S1 Fig). Since AbdA pulse doesn’t come on simultaneously in all the pNBs, therefore these cells die asynchronously. NBs start dying from early L3 stage and over a period of next 48 hrs different pNBs activate AbdA at different times and undergo apoptosis with majority of death happening between mid to late L3 stages. Owing to this, we chose to look at the larval ventral nerve cords (VNC) in time range of 84–90 hrs AEL. At this time, we expected majority of abdominal pNBs to be lacZ+. The reporter lines for NBRR fragments F1 (8 Kb), F2A (6 Kb) and F2B (6 Kb) failed to show any lacZ expression in abdominal pNBs (S1A–S1C Fig). However, NBRR fragment-3 (8 Kb) and fragment-4 (8 Kb) reporter lines (here on referred to as F3-lacZ and F4-lacZ) expressed in pNBs of abdominal CNS (Fig 1B and 1C). This indicated that the apoptotic enhancer lies in 3 Kb overlapping region of F3 and F4 fragments. This was further confirmed by analyzing the expression of the reporter line made from last 4.5 Kb region of F3 (referred to as F3B-lacZ, S1D Fig). Subsequent reporter lines were made by subfragmenting 3Kb overlapping region, of which a 1Kb reporter lacZ line (referred to as F3B3-lacZ) recapitulated larval pNB expression in abdominal pNBs at 84-90hrs AEL (Fig 1D).
In order to isolate the smallest modular enhancer, a 717bp subfragment was further selected from 1 Kb based on its sequence conservation across multiple Drosophila species, chromatin accessibility and the presence of multiple transcription factor (TF) binding sites, as assessed by UCSC genome browser (S2 Fig) [39]. The transgenic line for 717bp subfragment (referred to as 717-lacZ, Fig 1E) was generated by site specific insertion [40]. We found that this reporter line expressed in pNB at 84–90 hrs AEL, but the expression of the reporter was limited to Vl pNBs at late L3 stage. Our analysis suggests that this was a consequence of the insertion of the construct at the specific site (attP40-25C6), since insertion of 1Kb F3B3-lacZ at the same site also restricted its expression to Vl pNBs (S1E Fig). Owing to this, even though 717-lacZ expressed in pNBs (Fig 1E) and exhibited other features of genuine apoptotic enhancer (S3 Fig), specific insertion site seems to have altered its activity. Therefore, majority of following experiments were done with 1Kb F3B3-lacZ, and 717 bp enhancer was mainly considered for in vivo binding site mutant analysis.
In order to assess the temporal regulation of these genomic sub-fragments, we analyzed the expression of 8kb F3, 1kb F3B3 and 717-lacZ at early L3 stage (66-72hrs AEL), which was a few hours prior to initiation of pNB apoptosis (S3 Fig). We found all the reporter lines expressed weakly in abdominal pNBs and their expression was limited only to a few abdominal pNBs. However, as larvae progress to mid L3 stage, the expression is extended into more pNBs (Fig 1B–1E), suggesting that these enhancer-lacZ lines reflect the temporal control of RHG gene expression.
In order to genetically isolate the enhancer, we also generated a 14.6 Kb genomic deletion called M22 (detailed in S1 Text) which deletes the entire F3 fragments (Fig 1A and S4D Fig). We observed that heteroallelic combination of M22/MM3 (MM3 is a 54 Kb deletion used earlier to isolate NBRR [38]) showed ectopic NBs in the abdominal region of CNS (161.9+/-12.7, n = 20 VNCs, S4B and S4C Fig). The number of ectopic pNBs were comparable to those observed in MM3/MM3 deletion (167.6+/-10.8, for n = 12 VNCs), suggesting that 14.6Kb M22 deletion uncovers the enhancer for abdominal pNB apoptosis.
The reporter line expression and M22 deletion analysis strongly suggested that enhancer for abdominal pNB lies within 1Kb F3B3 region of NBRR.
Since, we could capture lacZ+ abdominal pNBs with reasonable frequency, this indicated to us that lacZ expression in these cells was not immediately followed by cell death. We also observed that intensity of lacZ expression in pNBs in early L3 stage was weak and became stronger in mid L3 stage (Compare S3 Fig and Fig 1B–1E). This suggested that lacZ reporter expression (and RHG genes) comes on and then sustain itself, till these cells undergo apoptosis. This is congruent to what is reported earlier in case of grim deletion where NB death is delayed till late L3 stages, when rpr finally executes the cell death [38].
Considering these observations, we expected that the apoptotic enhancer should be capable of maintaining the expression of the lacZ reporter (and RHG genes) in abdominal pNBs even till late L3 stage of development. To this end, we tested different reporter lines (F3, F4, F3B, F3B3, Fig 2A–2C, and 717-lacZ shown with later results) for their sustained expression in pNBs destined for apoptosis. This was achieved by testing the expression of lacZ lines in cell death blocked background by either using genetic deletions (for NBRR) or by expression of apoptosis blocker p35.
We observed that both F3 and F4-lacZ lines expressed in abdominal pNBs as late as 114–120 hrs AEL (Fig 2A and 2B) in M22/MM3 transheterozygotic background.
In order to conclusively confirm that larval NBs which undergo AbdA mediated apoptosis express the reporter line, and this expression sustain till late L3 stage we used tub-GAL80ts; insc-GAL4 driven UAS-p35 expression. This was used to temporally block NB apoptosis specifically from first instar larval stage (L1) (t-shift as shown in S8A Fig). We found that reporter lines F3B3-lacZ (Fig 2C) and 717-lacZ (shown with later results) expressed in the surviving pNBs as late as 114–120 hrs AEL.
Collectively these observations suggest that apoptotic enhancer expression once initiated in a pNB is maintained till it undergoes death.
Grh has been reported to be expressed in CNS from embryonic stage 11. Its expression in larvae is limited to NBs and is excluded from neurons. Since CNS specific grh mutants show a block in abdominal pNB apoptosis [35,41,42], we decided to investigate its role in AbdA mediated pNB apoptosis.
To this end, we used RNA interference (RNAi) to knock down grh, abdA and Notch (discussed in later section) in pNBs and score for their effect on 1Kb F3B3-lacZ reporter line expression. tub-gal80ts was used to temporally induce the knockdown from late embryonic stage and larvae were dissected in late L3 stage (114–120 hrs AEL; t-shift as shown in S8B Fig). The effect of these knockdowns on F3B3-lacZ in abdominal pNBs was quantitated and compared to lacZ levels from pNBs blocked for apoptosis by expression of p35. In order to maintain uniformity of comparison, pNBs of ventromedial (Vm) lineages have been quantitated across all genotypes (Fig 3E). We found F3B3-lacZ expression to be consistently downregulated in pNBs when grh (n = 26 pNBs, Average intensity = 2.3+/-0.4), abdA (n = 23 pNBs, Average intensity = 4.1+/-0.9) and Notch (n = 23 pNBs, Average intensity = 4.9+/- 3.4) were knocked down (Fig 3B-3B””, 3C–3C”” and 3D–3D””), as compared to pNBs expressing p35 (n = 34 pNBs, Average intensity = 19.9+/-10.2) (Fig 3A–3A””). The trend and significance of the data was unchanged across multiple experimental sets. One such set has been presented in the Fig 3. To rule out sample variations, Dpn staining for the NBs across different genotypes was quantified and found to be comparable (S5I Fig).
Ectopic expression of AbdA in thoracic pNBs is known to cause their apoptosis. Therefore, we expected that enhancer-lacZ expression should also get induced in these cells in response to ectopic AbdA (Grh is already present in these cells). Ectopic expression of AbdA was induced from early L3 stage and larvae were dissected 7 and 12 hrs later in early and mid L3 stage for F3 and F4-lacZ respectively (t-shift as shown in S8F Fig). We observed that F3-lacZ expressed in very few thoracic pNBs (Fig 3F) in control VNCs, whereas the ectopic expression of AbdA induced F3-lacZ in many thoracic pNBs (Fig 3G). Consistent with this observation, F4-lacZ expression was also ectopically induced (S5A and S5B Fig). We also found that induction of lacZ happens primarily in pNBs as indicated by co-staining for AbdA, Dpn and lacZ (Fig 3H). The lacZ expression seen in some of the progeny is a consequence of these cells inheriting lacZ from their progenitors. Similarly, we observed 1Kb F3B3 and 717-lacZ also get induced in response to AbdA over expression (S5E–S5H and S5G and S5H Fig). Since these smaller subfragments were more promiscuous in their expression in thoracic region, induction in response to AbdA was scored by increase in intensity of lacZ expression in addition to ectopic expression in thoracic NBs. Ectopic expression of lacZ could also be detected in central brain as well (S5F and S5H Fig).
The inducible expression of F3, F4, F3B3 and 717-lacZ further suggested that apoptotic enhancer is responsive to ectopic induction of AbdA. Further, the knockdown suggest that AbdA and Grh transcriptionally regulate RHG genes in pNBs through 1Kb F3B3 enhancer.
Out of ten pNBs in Dfd-SEG in L2 stage, four undergo Dfd dependent apoptosis as animal progresses to L3 stage of development (Fig 4A) [24]. In order to investigate molecular basis of this apoptosis (Fig 4A), we first checked whether Grh was expressed in pNBs found in L2 stage in Dfd-SEG region. These pNB lineages were identified by their location in Dfd stained region of SEG. The pNBs were marked by Dpn and the whole lineage was marked by inscGAL4 driven UAS-mCD8-GFP expression (Fig 4B and 4C). We consistently found 10 pNBs (9.92+/-0.27, for n = 13 L2 VNCs) which were Dpn+/Grh+ (Fig 4F) in early L2 stage (48–54 hrs AEL). Only few of these early L2 pNBs showed very low but detectable levels of Dfd (0.38+/- 0.65 for n = 13 L2 VNCs) shown as Dpn+/Grh+/Dfd+ (in Fig 4F). By late L3 stage (114–120 hrs AEL) only 6 pNBs and associated lineages were observed. We found that pNBs in these lineages always expressed Grh (6.0+-/-0, for more than 20 L3 VNCs) (Fig 4C and 4F) and showed very low or no expression of Dfd. As in L2 stage, pNBs were consistently Grh+/Dpn+ (5.3+/-0.48, for n = L3 VNCs) (Fig 4C” and 4F) and Dfd negative (Fig 4C and 4F). On a closer observation of the lineages in both abdominal and Dfd-SEG segments of VNCs, we found that Hox-Grh code of pNBs and their progeny was same: pNBs were always Grh+/Hox-, progeny were always Grh-/Hox+ (checked for more than 20 L3 VNCs) (Fig 4D–4D’ and 4E–4E’).
This implied that apoptosis of pNBs in Dfd-SEG may also be Grh dependent, and is possibly triggered by change in Hox-/Grh+ state of pNB to Hox+/Grh+ state. This prompted us to test the functional role of Grh in this apoptosis by knocking down its expression using genetic mutant combination and RNAi.
To this end, we counted and compared the number of pNBs in Dfd-SEG region of grh370/B37 (CNS specific null allelic combination for grh) with wild type controls. In wild type late L3 stage VNC (insc>mCD8-GFP, 114–120 hrs AEL, Fig 5A–5A”‘), we counted 6 pNBs (6.0+/-0.6, for more than 20 L3 VNCs, Fig 5C, bar-1 of the graph) as expected [24]. For the ease of representation Fig 5A shows a single confocal section with 4 of these 6 pNBs (marked by yellow arrowheads, Fig 5A–5A”‘). However, in case of grh mutants (grh370/B37; wr>mCD8-GFP, in late L3 stage Fig 5B–5B”‘) an average 13 pNBs could be detected in Dfd-SEG region (13.6+/-0.5, for n = 10 L3 VNCs, Fig 5C, bar-3 of the graph). Representation show a single confocal section of the mutant VNC where seven out of total 14 NBs found in Dfd-SEG region are shown (Fig 5B–5B”‘). Remaining ectopic pNBs were in other confocal planes and hence are not visible here. Since the number of NBs in case of grh mutant was more than ten we expected some of NBs to be embryonic in origin.
In order to conclusively establish the contribution of Grh in larval NB apoptosis we induced RNAi mediated grh knockdown specifically from late embryonic stages (by this time embryonic NB death has already taken place) and dissected the larvae in late L3 stage at 114–120 hrs AEL (t-shift as shown in S8B Fig). Ectopic pNBs in Dfd-SEG were identified and counted based on their position and Dfd staining. The detailed method for identification of ectopic pNBs in Dfd-SEG are given in S1 Text.
We found 4 ectopic pNB lineages (total of 10 pNB lineages) in late L3 VNCs (9.56+/-0.51, for n = 15 L3 VNCs) (Fig 5C, bar-4 of the graph). These numbers were in agreement with the fact that there are 10 pNBs reported in Dfd-SEG of wild type CNS in L2 stage, as well as when anti-apoptotic gene p35 was specifically expressed from late embryonic stage (Fig 5C, bar-5 of the graph) or early L1 stage (10+/-0.0, for 7 VNCs).
These results show that similar to its role in abdominal segments, Grh also plays an important role in pNB apoptosis in Dfd-SEG region.
In order to identify the genomic location of the enhancer responsible for apoptosis of 4 pNBs in Dfd-SEG, we counted the number of ectopic pNBs in this region in late L3 stage for various deletion combinations. We could recover only 6 pNBs in Dfd-SEG region in M22/MM3 (Fig 5C, bar-6 of the graph), MM3/MM3 (Fig 5C, bar-7 of the graph). This implied that enhancer responsible for activation of apoptosis of pNBs in Dfd-SEG is different from abdominal apoptotic enhancer and lies outside 22Kb NBRR and 54Kb genomic region deleted in MM3 allele.
Notch signaling is often utilized to make decisions in multiple developmental contexts [43–46]. More recently, it has been suggested to play a role in abdominal pNB apoptosis, where it was implicated in activating the pulse of AbdA triggering pNB apoptosis [25]. We first checked whether Notch has a lineage autonomous function in pNBs of abdominal segments by making Notch loss of function MARCM clones (N55e11). We recovered two kind of clones in abdominal region of CNS; the first class of clones were recovered in A3-A7 segments of CNS where the AbdA mediated abdominal pNB apoptosis is mainly reported (discussed below). A second class of clones were recovered in A1-A2 segments where no AbdA mediated apoptosis occurs (detailed in discussion). The clones recovered in A3-A7 segments had a surviving pNB at 114–120 hrs AEL indicating that pNB had failed to undergo apoptosis (Fig 6A and 6B). These clones were small and showed no consistent and significant downregulation of AbdA in the pNBs (compared to the levels of AbdA in progeny of the same lineage, Fig 6B). Therefore, we checked for the levels of Grh in these cells and found them to be unaffected in surviving pNBs (Fig 6A). Out of 23 clones examined across 10 VNCs, only 3 clones showed complete AbdA downregulation, 15 clones showed partial downregulation, and 5 clones showed no downregulation of AbdA. We also employed RNAi to verify these observations. Notch knockdown was initiated from late embryonic stage and its effect was examined in late L3 stage (t-shift as shown in S8B Fig). Three types of ectopic NBs were recovered; the first two type of NBs showed AbdA levels comparable or less than neuronal progeny of the lineage, these were designated as NBs with no or partial AbdA downregulation. The third type of NBs showed no AbdA expression (categorized as NBs showing complete downregulation). We report on an average 20 pNBs (20.0+/-2.7, n = 12 VNCs) surviving per VNC, of which only 2 pNBs (2.5+/-1.44, n = 12 VNCs) could show us complete downregulation, while out of remaining 17, 10 pNB (10+/-2.92, n = 12 VNCs) showed partial and 7 pNBs (7.67+/-3.5, n = 12 VNCs) showed no AbdA downregulation (Fig 6E). Thus, both RNAi and mutant data for Notch demonstrate that Notch probably does not induce AbdA expression in pNB of A3-A7 segment. Moreover, if Notch was indeed capable of inducing AbdA expression in pNBs, in that case overexpression of NICD (Notch Intracellular domain) in abdominal pNBs should induce AbdA and hence F3-lacZ in majority of these cells. We overexpressed NICD from mid L2 stage, and larvae were dissected after 14hrs at 30°C (approximately in early L3 stage, 70Hrs AEL, t-shift as shown in S8G Fig). However, we did not see ectopic induction of F3-lacZ in response to NICD overexpression in majority of the abdominal pNBs (S5L Fig).
These results suggested to us that Notch signaling may not be the trigger for AbdA induction in pNBs of A3-A7 segments but instead may have a direct role in pNB apoptosis. In order to validate this, we induced Notch knockdown and AbdA over expression simultaneously from early L1 stage (gal80ts; inscGAL4>UAS-Notch-RNAi, UAS-abdA, UAS-mCD8-GFP) and dissected larvae at late L3 stage of development (t-shift as shown in S8A Fig). If Notch signaling was working only through AbdA activation, over expression of AbdA in Notch knockdown background should have caused the apoptosis of abdominal pNBs. On the contrary, in our experiments, we found that Notch knockdown blocked apoptosis of abdominal pNBs even when AbdA was over expressed in these cells (16.8+/-3.2 surviving pNBs, in 11 VNCs, Fig 6D). Even though AbdA was expressed in surviving abdominal pNBs in late L3 stage (Fig 6D”), we wanted to ensure that sufficient levels of AbdA were expressed in pNBs in earlier stages, for this we analyzed VNCs in mid L3 stages as well. We found sufficient levels of AbdA were expressed in abdominal pNBs in mid L3 stage (S6A” and S6B” Fig). These observation and downregulation of apoptotic enhancer-lacZ in response to RNAi mediated knockdown of Notch (Fig 3D–3D””), further reinstate our proposition of a direct role of Notch in abdominal pNB apoptosis.
We also observed that the number of thoracic pNBs found in these VNCs were comparatively less in late L3 stages, which suggested that AbdA was able to cause apoptosis in thoracic region despite Notch knockdown in these cells. Noticeably, some of the surviving pNBs in thoracic segment expressed AbdA (S6D Fig). We think these cell types are refractory to AbdA mediated cell death. This possibly points towards a segment specific role of Notch in pNB apoptosis.
Next, we checked for the role of Notch signaling in pNB apoptosis in Dfd-SEG. For this RNAi mediated knockdown for Notch was induced from late embryonic stage and its effect was assessed in late L3 stage (t-shift as shown in S8B Fig). We could consistently recover 2–3 ectopic pNBs (8.61+/- 0.75, for n = 26 L3 VNCs) in Dfd-SEG as against 6 pNB seen in control larvae (Fig 6C and 6E). A closer observation of the ectopic pNBs show that they have normal levels of Grh (Fig 6C) and also expressed low levels of Dfd (Fig 6C). A comparative quantitation of pNBs recovered for both abdominal segments and Dfd-SEG region, upon Notch knockdown is shown in Fig 6E.
These results indicate a role of Notch signaling along with Hox and Grh in mediating pNB apoptosis in both abdominal and Dfd-SEG regions. In both cases, we found that Notch signaling does not impact the levels of Grh. Our results with abdominal pNBs further suggests that Notch knockdown is epistatic to AbdA overexpression in pNBs. Therefore, we believe that Notch does not regulate Hox expression and instead plays a direct role in pNB apoptosis.
Hox genes have been known to function with two other TALE-HD containing transcription factors, Hth and Exd. We tested the role of Exd in pNB apoptosis in abdominal region by making MARCM clones. We recovered NB containing exd mutant clones in abdominal region (n = 16 clones in A3-A7 segment of 13 larval VNCs, Fig 7A–7A””). These pNB also showed a normal expression of Grh (n = 7 clones scored in 6 VNCs) and AbdA (Fig 7A). Exd is known to function with Hth, which helps in its nuclear localization [47]. Surprisingly, we found that hthP2 mutant clones in abdominal region did not block pNB apoptosis (Fig 7B–7B””). We noticed that clones were marked with GFP (Fig 7B’), suggesting that pNBs divided normally following the exit from quiescence but then underwent apoptosis (n = 23 clones scored in 10 VNCs). Since hthP2 is a strong hypomorph, we tested 6 hth RNAi lines. Even when RNAi mediated knockdown was induced from early embryonic stages for hth gene, we could not recover any ectopic pNBs in abdominal region (t-shift as shown in S8E Fig). Though we could see a visible down regulation of Hth expression in thoracic pNB lineages, suggesting that these lines were capable of inducing potent knockdown (S9 Fig).
These results led us to investigate a similar role for Exd and Hth in Dfd-SEG region as well. None of the exd-RNAi lines tested could give us ectopic pNBs in abdominal region. However, knockdown of exd from early embryonic stages (t-shift as shown in S8E Fig) resulted in approximately 2 ectopic pNBs in Dfd-SEG region in late L3 stage VNCs (1.75+/-0.61 for 19 VNCs) (Fig 7C). The ectopic NBs obtained were at the characteristic location where ect1Dfd ectopic pNB lineage is normally observed and these lineages were Dfd+ as well. Similarly, we tested the role of Hth by 6 RNAi lines (as mentioned earlier) by inducing knockdown from early embryonic stages but we could not find any significant difference between control VNCs and knockdown VNCs in late L3 stage (t-shift as shown in S8E Fig).
These results suggested, that Exd has a role in Hox mediated pNBs apoptosis in both abdominal and Dfd-SEG regions, and perhaps this role is independent of Hth.
Next, we checked for potential Hox, Exd and Grh binding sites in entire 1Kb F3B3 genomic region. Since Grh plays an important role in this apoptosis and Hox protein bind to AT rich sequences occurring at a high frequency in the genome, we decided to narrow our search for Hox sites by scanning for potential Grh binding sites in the vicinity. We identified 14 such sites conforming to variation of the known Grh binding consensus sequence (WCHGGTT) [48], these sites also had AT rich sequences (potential AbdA and Exd binding sites) in 20bp flanking region [49]. These 14 Grh sites and surrounding AT rich sequences were categorized into two type of motifs. First type only had one Grh binding site (designated as type-I), and 6 such motifs were identified (shown as green rectangles in Fig 8A). The type-II motifs had 2 closely located Grh binding sites, and 4 such motifs were identified (shown as green squares in Fig 8A). We could find only one Hox-Exd consensus site (A/TGATNNATNN) in the entire F3B3 region referred to as motif-29 (grey rectangles, Fig 8A). We tested all these motifs for binding by EMSA. We found that 5 out of 6 Type-I motifs showed binding to Grh (motifs- 23, 25, 27, 31, 32, lanes-2, 30, 43, 82, 95, S7 Fig), motif-24 didn’t show any Grh binding (lane15-16; S7B Fig). Three out of four Type-II motifs showed Grh binding these were motif-28, 33 (lane-56 and 108, S7 Fig) and motif-30 (lane-2, Fig 8B). Motif-34 did not bind Grh. All these motifs were checked for Exd and AbdA binding as well. Motif-29 which had consensus Hox-Exd binding site showed AbdA-Exd binding but no Grh binding (lane-69, S7F Fig). The details of individual protein binding to these sites are given in S2 Table. Amongst all the motifs that were tested, we found motif-27, 30 and 32 assembled a tetracomplex (DNA-AbdA-Exd-Grh). We decided to analyze motif-30 in details since it showed a good tetracomplex formation with Grh, AbdA and Exd, as well as strong binding for each of the individual proteins.
To gain insights into the tetracomplex assembly, we started out by testing motif-30 for AbdA, Grh and Exd binding. We used increasing concentration of Hox (100 and 200ng, lane-4 and 5, Fig 8C) and fixed concentration of Grh and Exd (200ng, lane-2 and 3, Fig 8C) and found that Grh (lane 2, Fig 8C) and Hox (lane 4 and 5, Fig 8C) bound to DNA on their own (binding indicated by green and red arrowheads), while Exd failed to bind on the DNA (lane 3, Fig 8C). On increasing AbdA in presence of fixed concentration of Grh (200ng), a band of mobility slightly lower than Grh band alone was observed (lane 6 and 7, Fig 8C). As expected increasing concentration of AbdA in presence of fixed concentration of Exd (200ng) showed AbdA-Exd complex formation on the DNA (lane 8 and 9, black arrowhead, Fig 8C). Similarly, Grh in the presence of a fixed concentration of Exd (200ng) together showed only a slight increase in Grh binding on DNA (lane 10 and 11, Fig 8C). Importantly, when Grh and Exd concentrations were kept constant (200ng each), addition of AbdA led to a band of lowest mobility. This tetracomplex (DNA-AbdA-Exd-Grh) is shown by a white arrowhead for lane 12 and 13 (Fig 8C). All subsequent EMSA experiments were done with fixed concentration of all the proteins (200ng).
Next, in order to test the specificity of Grh binding, oligos mutant for potential Grh1 and Grh2 binding site were analyzed. We observed dramatic decrease in Grh binding in oligo with mutation for Grh1 binding site when compared to wild type oligo (lane 15 vs 20, Fig 8D). It was also noticed that AbdA-Grh complex formation on DNA was compromised (lane 16 vs 23, Fig 8D). We found that AbdA-Exd complex was unaffected (lane 24, Fig 8D) while tetracomplex could not be visualized in case of Grh1 binding site mutant oligo (lane 18 vs 26, white arrowhead, Fig 8D). This indicates that Grh1 site is critical for tetra-complex formation and plays an important role in assembly of Hox-Grh complex.
In Grh2 binding site mutant oligo, we could not find a significant decrease in Grh binding (lane 28 vs 33, green arrow head, Fig 8E). We also found that AbdA-Grh complex is slightly reduced but is still present (lane 29 vs 36). This could be attributed to two reasons, one possibility is that Grh2 binding site plays a role in AbdA-Grh complex formation; this seems unlikely since mutation of Grh1 binding site abolished the AbdA-Grh binding completely. The more likely explanation could be that since Grh2 binding site overlaps with both Hox1 and Hox2 binding sites, and therefore the mutation of Grh2 site affects the assembly of AbdA-Grh complex.
We also found that AbdA-Exd binding on oligo mutant for Grh2 binding site was reduced but was still present (lane 37, black arrowhead, Fig 8E). This could be attributed to the fact that Grh2 site also overlaps with Hox2 site and hence could affect AbdA-Exd binding which happens on Hox2-Exd site (confirmed in later analysis in next section). Most importantly we found that tetracomplex was still intact on the oligo mutant for Grh2 binding site (lane 31 and 39, white arrowhead, Fig 8E) suggesting that Grh1 site is more important for the assembly of tetracomplex. These results suggested to us that AbdA and Grh might interact with each other physically.
To test this idea, we performed an in vitro GST-pull down assay, wherein bacterially expressed GST and GST-AbdA protein were bound to GST beads and incubated with His-Grh bacterial lysate (input). The proteins pulled down (from His-Grh lysate) by GST-AbdA and GST alone were separated on SDS-PAGE and probed with anti-His antibody. We found that while GST alone showed no band, GST-AbdA could successfully pull down His-Grh (approximately 90Kda- Fig 8B).
These results indicate that AbdA and Grh are not only important for tetracomplex formation, they also physically interact with each other.
Next, we examined oligos mutants for AT rich sequences to identify Hox and Exd binding sites. Motif-30 oligo mutant for potential Hox1+2 binding sites showed no AbdA binding (lane 122 vs 128, S7J Fig), suggesting that these were Hox binding sites. Subsequently, we tested AbdA and Exd binding on oligo mutant for potential Exd binding site. We found that a lower mobility complex was formed by AbdA in presence of Exd protein on wild type oligo (lane 41 vs 43, Fig 9B) which was abolished in oligo mutant for Exd binding site (lane-50, Fig 9B, Exd site is shown in blue). This suggested that AbdA-Exd complex most likely assembles on Hox2-Exd sites, more so considering the proximity of the two sites. Mutation of Exd site also abrogated tetracomplex (lane 44 vs 52, white arrowhead, Fig 9B), while Grh protein still bound to DNA (lanes 46, 51–52, Fig 9B).
In case of oligo with Hox1 binding site mutation, we noticed a decrease in AbdA binding on DNA (lane 54 vs 61, red arrowhead, Fig 9C). We also found that Grh binding (lane 55 vs 59, green arrowhead, Fig 9C) was reduced but still present. This could be due to the fact that Hox1 mutation is in middle of the Grh binding sites and therefore affected Grh binding onto DNA. Moreover, we found that AbdA-Exd complex binding (lane 63, black arrowhead, Fig 9C) was intact but tetracomplex binding was dramatically diminished (lane 57 vs 65, white arrowhead). We believe this also could be due to effect of Hox1 mutation affecting nearby Grh1 binding site (as discussed above), and hence the tetracomplex formation.
Next, we tested oligo mutant for Hox2 binding site. We found that both AbdA binding (lane 68 vs 74, Fig 9D) and AbdA-Exd binding (lane 70 vs 76, Fig 9D) were abolished in this case. Though the Grh binding could still be detected (lane 73) the tetracomplex formation was completely abolished (lane 71 vs 77, Fig 9D). Since we could not observe any AbdA-Grh complex, this suggests that AbdA-Grh complex uses Hox2-Grh1 binding site. In corroboration to this, we found that in Hox1+2 double mutant oligo, AbdA-Exd, (lane 122 vs 129, S7J Fig), AbdA-Grh (lane 124 vs 130, S7J Fig) and tetracomplex binding is completely abolished (lane 125 vs 132, S7J Fig). Also the effect on complex formation was much stronger in this case compared to individual mutants for Hox1 and Hox2 sites.
The above experiments suggest that AbdA-Exd complex is critical for the tetracomplex formation. AbdA-Exd along with Grh most likely assembles a tetracomplex on Hox2-Exd and Grh1 site on DNA. We believe that this tetracomplex could contribute to regulation of RHG genes through F3B3 enhancer.
In order to test the in vivo relevance of various motif tested for AbdA, Exd and Grh binding in vitro, enhancer mutagenesis was carried out. All the mutagenesis studies were carried on the 717 bp subfragment (Fig 10A). Three kind of mutant constructs were made. In first construct, Grh binding sites in all 8 motifs (present in 717 bp) were mutagenized leaving Hox-Exd binding sites mostly intact (717-Grhmutant-lacZ). In the second construct Hox-Exd and Grh binding sites across all the 8 motifs were mutagenized (717-Hox-Exd-Grhmutant-lacZ) (Fig 10A). A third construct was designed to test direct role of Notch signaling in abdominal pNB apoptosis. Since Notch intracellular domain goes into the nucleus and activates gene through its executive TF Suppressor of Hairless (Su(H)), we identified and mutagenized all recognizable Su(H) binding sites in 717 bp enhancer (717-Su(H)mutant-lacZ) (Fig 10A). We could identify seven such binding sites which were variations of known consensus binding sequence for Su(H) (RTGRGAR) [50]. All the transgenic lines were crossed into the background of UAS-p35 and were subsequently checked for the expression of reporter lacZ in abdominal pNBs in late L3 stage. For comparison of lacZ levels tubulin-GAL80ts; inscGAL4 was used to drive the expression of p35 to block the apoptosis of the pNBs. This helped us to visualize the sustained expression of lacZ in later stages, which serves as a hallmark for identification of abdominal apoptotic enhancer. Since the expression of 717-lacZ was restricted only to Vl lineage in late L3 stage, we compared wild type (Fig 10B) and mutant versions (Fig 10C–10E) of the enhancer for their capacity to drive the expression of lacZ reporter in these cells. We found that reporter lacZ expression was completely missing in abdominal Vl pNBs of all the three mutant versions of the enhancer in late L3 stage (Fig 10C–10E). Next, we decided to visualize the expression of the reporter in early L3 stage of development. Interestingly, we found that the mutant reporter lines for 717-Grhmutant-lacZ and 717-Hox-Exd-Grhmutant-lacZ expressed normally in abdominal pNBs (S10B–S10D Fig). These results suggested that motifs being analyzed here play a crucial role in sustenance of the expression of the apoptotic genes and are not critical for initiation of their expression in early stages.
In our analysis with 717-Su(H)mutant-lacZ, we found that its expression in pNBs was slightly delayed (S10E–S10F Fig) in early stages, but in late L3 stage, like other mutant enhancer-lacZ lines, its expression was completely missing from Vl pNBs.
These results suggested that Notch signaling has a direct role in pNB apoptosis. The enhancer-lacZ analysis suggest that it may have a temporal role in apoptosis initiation but more importantly it seems to have a role in maintenance of the enhancer activity and hence RHG genes during apoptosis.
A large fraction of cell death in developing organism happens in CNS, which underlines its importance in CNS morphogenesis. The coupling of death in CNS with spatial developmental cues like Hox genes is a convenient strategy evolved by nature for patterning of neural tissues to coordinate developmental apoptosis with spatial regionalization of the organism. Therefore, it is of interest to understand the molecular details of this mechanism. We have investigated this in abdominal and Dfd-SEG region of larval CNS. We find that Hox mediated pNBs apoptosis happens through a battery of common players (Hox-Exd-Grh-Notch) perhaps using a similar mechanism, albeit through a different enhancer.
Previous report suggest that RHG genes (mainly grim and reaper) express and function in a combinatorial manner in dying pNBs [38]. Wherein rpr deletion alone shows no block of apoptosis, grim deletion alone shows a delay till late L3 stage, while double deletion completely block this cell death. This indicates that grim is the major player and rpr probably takes over in absence of grim. Since abdominal pNBs are destined to die, therefore, regulation of these genes is designed to ensure that once their expression is switched on, it should be maintained in these cells till they undergo apoptosis. It is also expected that their coordinated expression in a cells of a specific region may be regulated by a single shared enhancer [25,38]. Similarly, their expression in different regions of developing CNS may be controlled by multiple region-specific enhancers. Our data support these ideas in a limited context of Drosophila NBs (Fig 11A). We find that the larval abdominal pNB apoptosis is regulated by an enhancer lying within 1kb region (F3B3) of NBRR. This 1 Kb region is a subfragment of 5 Kb embryonic NB specific apoptotic enhancer (also known as enh-1 [25]) and is deleted in M22 (14.5 kb deletion). Our failure to recover ectopic pNBs in Dfd-SEG in M22/MM3, MM3/XR38 and MM3/MM3 combinations (Fig 5C, bars 6 & 7 of the graph) indicate that enhancer responsible for Dfd mediated pNB apoptosis (in Dfd-SEG) lies outside 22 Kb NBRR and 54 kb MM3 deletion. Thus, apoptosis of larval pNBs in abdominal and Dfd-SEG region are controlled through two distinct enhancers (as shown in Fig 11A).
Therefore, while a single Hox gene like AbdA can activate pNB apoptosis by using same enhancer in embryonic and larval stages of development (as it happens in abdominal segments for enh-1 [25] its subfragment F3B3), different regions of the developing CNS (abdominal and Dfd-SEG regions) employ different enhancers to activate RHG genes.
It is known that abdominal pNBs do not die in a synchronous manner. They start dying from early L3 stage and over a period of next 48 hrs different pNBs activate AbdA at different times and undergo apoptosis. We observed the same from our analysis of different enhancer-lacZ lines, wherein some cells show lacZ expression just prior to early L3 stage, while others express lacZ later on. We also observed that intensity of lacZ reporter in pNBs becomes stronger from early to late stages. This suggests that lacZ reporter expression in pNB can be categorized into two phases, initiation phase followed by maintenance phase of expression. In agreement to this, we find that different enhancer-lacZ lines (F3, F4, F3B, F3B3, and 717-lacZ) show a sustained lacZ expression in the pNBs till late larval stages in a cell death blocked background (Fig 2, Fig 10B).
We identified 8 motifs with composite AbdA-Exd-Grh binding sites within 717bp enhancer. In vitro binding assay suggested that Hox-Exd and Grh form a tetracomplex on 3 out of 8 motifs analyzed by us (motif-27, 30 and 32). Of these 3 motifs we used motif-30 as a model to understand the complex assembly and found that all the three proteins (Hox, Exd and Grh) are critical for tetracomplex formation (Figs 8 and 9). In order to test the in vivo relevance of the composite Grh-AbdA-Exd binding sites found within different motifs of apoptotic enhancer, we mutagenized these binding sites. We tested the capacity of the resulting mutagenized enhancer to drive lacZ in abdominal pNBs. We found that, in both 717-Grhmutant-lacZ and 717-HEGmutant-lacZ, enhancers were normal for the initiation of the lacZ expression in pNBs (S10B–S10D Fig), but interestingly the mutant enhancers were incapable of sustaining the expression of lacZ reporter in these cells till later stages (Fig 10). This implies that these motifs play an important role in maintenance rather than initiation of RHG gene expression. In our experiment we mutagenized AbdA, Exd and Grh binding sites in all 8 motifs found in 717bp enhancer (717-HEGmutant-lacZ; Fig 10D). Since our analysis cannot discriminate whether 3 tetracomplex forming motifs are more critical compared to rest of the 5, therefore the results does not imply that tetracomplex is central for the maintenance activity of the apoptotic enhancer. But considering the direct physical interaction of Grh with AbdA (Fig 8B) and binding assays wherein majority of sites show Abd-Exd and AbdA-Grh complex formation (Figs 8 and 9 and S7 Fig), we believe that AbdA, Exd and Grh proteins together play a role in maintenance activity of the enhancer and a part of the same may be contributed by tetracomplexes assembled on the enhancer. Interaction of helix-loop-helix (HLH) protein (Grh in this study) and HD containing TFs (AbdA in this study) have been reported earlier. It has been shown that HLH and HD transcription factor (Meis/Prep and Pitx family) interact with each other to synergize the transcriptional response [51,52]. However, Hox per se had not been shown to interact with HLH proteins so far.
In our analysis, we have focused on Grh binding sites with nearby AT rich sequences. Some of these sequences turned out to be Hox-Exd binding sites (Fig 8C–8E, motif-30 and S7 Fig), which did not fit conventional consensus sequence A/TGATNNATNN. Therefore, it will be of interest to find out which of the domains known to be important for AbdA-Exd interaction (like YPWM and UbdA motifs [53,54]) play an important role for the complex formation on motif-30. Also, whether any of these domains will contribute to AbdA’s interaction with Grh as well [53,55]. Since our results suggest that above mentioned motifs are important for maintenance but not initiation of the enhancer activity, DNA motifs necessary for enhancer initiation are yet to be identified. We observed multiple standalone Hox binding sites in 717bp enhancer with no recognizable Grh and Exd binding sites in vicinity. These individual Hox binding sites were intact in all the three mutant versions of the enhancer tested by lacZ reporter assay. We believe that these standalone Hox sites could be the first ones to be occupied in response to increasing levels of AbdA in pNBs, and help in initiating the expression of RHG genes. Subsequently, AbdA-Exd could get recruited on the composite sites (Grh-AbdA-Exd sites) and helps to maintain the levels of RHG genes which eventually lead to death of pNBs. Since Grh is proposed to be responsible for installation of apoptotic competence [35], it could also be possible that it occupies its binding sites prior to AbdA pulse coming on. The idea of standalone Hox binding sites being the first responders to increasing Hox protein expression fits well with the fact that less regulation may be required at the at initial stage of enhancer firing. Therefore it is possible that at this stage individual Hox sites on the enhancer may get bound by any Hox protein. This is supported by the fact that overexpression of Abdominal-A or Antennapedia (Antp) or Ultrabithorax (Ubx) in thoracic pNBs resulted in their apoptosis [16]. The occupation of composite site of Grh-Hox-Exd come next and are important for maintenance of gene expression and eventual cell death.
So far, role of Grh has been reported in cell proliferation and in installing the competence to undergo AbdA mediated apoptosis. In this study we show that Grh along with AbdA contributes to transcriptional regulation of RHG genes in causing larval pNB apoptosis. This is based on the observation that Grh and AbdA knockdown downregulated apoptotic enhancer-lacZ (1Kb F3B3-lacZ reporter line, Fig 3E). Moreover, the enhancer mutagenized for Grh binding sites could not maintain its expression in late L3 stages in cell-death blocked background. This observation further supports a transcriptional role for Grh along with AbdA in RHG regulation in pNBs.
We also observe that pNBs in abdominal and Dfd-SEG express Grh but have a very low or no Hox expression (Grh+/Hox-), while their progeny show opposite expression code (Grh-/Hox+). It is observed in abdominal region (during L3 stages), that changing expression code of pNB from Hox-/Grh+ to Hox+/Grh+ (AbdA+/Grh+) results in its apoptosis. We believe that a similar theme for pNB apoptosis is being to be followed in Dfd-SEG as well, where pNBs are known to undergo Hox dependent apoptosis [24]. We find that this apoptosis is also dependent on Grh and like in abdominal segments expression of Dfd in L2 stage may change Dfd-/Grh+ state of pNBs to Dfd+/Grh+ state and cause their apoptosis. It is interesting to note that common TF code of pNBs within different region of VNCs (abdominal and Dfd-SEG) may help them to respond to similar signals (like Hox expression in pNBs) resulting in common outcome (apoptosis in this case).
We tried testing sufficiency of Hox and Grh in causing the apoptosis in different regions, by expressing Grh in Hox positive neurons. We did not see any increase in apoptosis in abdominal neurons or neuron in any other region (Dfd-SEG and thorax) of CNS. Similarly, in Dfd-SEG region overexpression of Dfd even from early embryonic stages could not cause the death of remaining 6 pNBs. This is different from the abdominal pNB which die precociously on AbdA expression or thoracic pNBs which die on expression of AbdA or Ubx or Antp [16]. This indicated that there are other molecular players in addition to Grh which are important for Hox mediated pNB apoptosis. These factors may not be same across different regions (abdominal and Dfd-SEG), but they are likely to function with Grh and contribute to apoptosis. Moreover within a set of pNBs, these factors may be expressed differentially, i.e. they may be expressed in 4 dying pNBs but not in rest of 6 pNBs within Dfd-SEG region. Hence identifying potential partners of Grh may be useful to understand how heterogeneity is generated within a population of pNBs.
Notch mutant clones generated in A1 and A2 segments of VNC very consistently downregulated AbdA (S10A Fig). This supported the earlier claim that Notch signaling could regulate the expression of AbdA in pNBs [25], but interestingly, no AbdA mediated pNB apoptosis is reported in these two segments [16,25,35]. AbdA mediated pNB apoptosis is a hallmark of A3-A7 segments [16,35], where even though we recovered ectopic pNBs (both by Notch MARCM and by RNA interference) we could not observe significant and consistent downregulation of AbdA (Fig 6E). The levels of Grh were unaffected in these pNBs (Fig 6A–6C). Simultaneous Notch knockdown and AbdA overexpression from early L1 stage (t-shift as shown in S8A Fig) blocked abdominal pNB cell death. This suggests that Notch signaling is epistatic to AbdA apoptosis happening in abdominal segments. On the other hand in thoracic segments, Notch knockdown failed to rescue AbdA induced death of some of the thoracic pNBs. Thus, we think that Notch signaling has a direct role in abdominal pNB apoptosis which seems specific for abdominal segments.
In case of 717-Su(H)mutant-lacZ, we found that the initiation of the reporter lacZ was slightly delayed but more importantly the maintenance of the expression in late L3 stage was completely crippled. This led us to suggest that perhaps Notch signaling does have a direct role in apoptosis unlike what has been reported earlier [25]. Whether Notch plays a role only in maintenance or both in initiation and maintenance of the enhancer is currently not clear.
It is to be noted that all recognizable Su(H) binding sites found in 717 bp enhancer are not close (in range of 20bp) to Grh-AbdA-Exd motifs. Therefore, how does Notch signaling play a role regulation of apoptotic enhancer remains to be investigated. One possibility is that Notch signaling is involved in initiation of the expression in collaboration with individual Hox binding sites, but this still doesn’t explain its role in enhancer maintenance.
Notch knockdown results in blocking of cell death of pNBs in Dfd-SEG, similar to abdominal pNBs (Fig 6A and 6C). Considering this we think that Notch perhaps plays similar roles in apoptosis in both these regions. Since the enhancer required for the activation RHG genes in Dfd-SEG is different from abdominal enhancer, and is yet to be identified, therefore it is difficult to currently test this idea at the moment.
We also tested role of known Hox cofactors Exd and Hth. Interestingly, we found that while Exd plays an important role in AbdA mediated apoptosis, Hth was not critical for this function (Fig 7A and 7B). Similarly in Dfd-SEG region knockdown of Exd but not Hth resulted in ectopic pNBs (Fig 7C).
Therefore, we are inclined to believe that while Exd is important for Hox mediated pNB apoptosis in both abdominal and Dfd-SEG region, Hth is not required in this process. Since Hth function as a nuclear transporter of Exd [47], our results suggest a Hth independent mechanism to transport Exd into the nucleus in pNBs. We were unable to test this idea directly, since we could not detect Exd protein in abdominal pNBs with the available antibodies and hence localization of Exd in hth mutant could not be assessed.
Hox mediated apoptosis of pNB is a mechanism which is used in multiple regions of developing CNS in flies [16,23,24]. We propose that Notch-Hox-Exd-Grh are a part of common machinery employed in pNBs, involved in activation of RHG genes in developing CNS. Understanding how Notch signaling coordinate this apoptosis with Hox-Exd-Grh (in abdominal and other regions), and characterization of the assembly of this multi-protein complex on DNA will be of interest in future. Grh is not expressed in post embryonic neurons, therefore whether these neurons employ overlapping players and same enhancer or have an independent mechanism for apoptosis needs further investigations.
The transgenic lines for 717bp enhancer and its mutagenized forms were generated by site specific insertion [40] of the constructs at attP40-25C6.
All other transgenic reporter lines were generated by classical P-element based transgenesis. Multiple reporter lines (at least 3 independent insertions for each fragment) were tested for their expression, representative images for all the lines are shown in the figures.
The deletion line M22 was generated by mobilization of the MiMIC element [56,57] inserted 9 Kb from 5’ region of NBRR (BDSC-30966) (S4 Fig). Multiple RNA interference lines from BDSC NIG and VDRC [58] were used to confirm the results wherever possible, the quantitative data presented is from RNAi stocks which are underlined: grhRNAi (VDRC-101428/KK; BDSC-28820), abdARNAi (VDRC-106155/KK), hthRNAi (NIG-17117-R4 and R2, VDRC-100630/KK, 12763/GD, 12764/GD, BDSC-27655), exdRNAi (VDRC-7802/GD, 7803/GD, 100687/KK), NotchRNAi (BDSC-27988 and 28981). UAS-dcr2; inscGAL4 UASmCD8-GFP; tub-GAL80ts (J. Knoblich, [59]), MM3 (K.White, [38]), worniuGAL4 (C. Doe), grh370, grhB37 (S. Bray [42]), FRT80B-Df(3L)H99 (R. Mann), hthP2 (R. Mann), exd1 (BDSC, 3293), FRT19A-N55e11 (BDSC, 28813), UAS-N-intra (BDSC 52008), UAS-p35 (DGRC, Kyoto, 108019), UAS-AbdA on III (R. Mann, [53]), UAS-AbdA on II (K. VijayRaghavan), elav[C155]-GAL4, UASmCD8::GFP, hsFLP1, w (BDSC, 5146), y w; tub-GAL80-LL9 FRT80B (BDSC, 5191), yw; FRT82B tub-GAL80-LL3 (BDSC, 5135), hsflp,FRT19A,tubGAL80; tub-GAL4,UASmCD8GFP/cyo-GFP (H. Reichert).
Egg collection were done for 6 hrs and flies were grown at 25°C and for all temperature shift based experiments (involving tub-GAL80ts) 12 hr egg collection was done at 18°C. The aging was calculated as number of hours after egg laying (AEL).
For lacZ quantitation experiments across different knockdowns, females of UAS-dcr2; inscGAL4 UASmCD8-GFP; tub-GAL80ts were crossed to UAS-abdARNAi, UAS-grhRNAi, UAS-NotchRNAi and UAS-p35 (DGRC-108019). Egg were collected for 12hrs at 18°C and shifted to 30°C after 42hrs (approximately late embryonic stage-early L1 stage) for all the three genotypes simultaneously. The larvae were reared at 30°C for approximately 90–91 hrs (by then in late L3 stage) and were dissected in wandering stage (114–120 hr AEL, t-shift as shown in S8B Fig), processed and imaged together. Lac-Z quantification for NBs was done in abdominal region by costaining with anti-β gal, anti-AbdA and anti-Dpn antibodies. The confocal slice which showed maximum lacZ staining (mostly this slice was at the centre of pNB) was selected for the surviving pNBs in all the combinations and compared. LacZ intensity was quantitated by ZEN 2012 software. The background signal intensity observed in region of the image outside larval brain was subtracted from lacZ signal observed in pNB.
For all other RNA knockdowns, females of UAS-dcr2; inscGAL4 UASmCD8-GFP; tub-GAL80ts were crossed to males of respective RNAi lines. The flies were allowed to lay eggs at 18°C for 12 hours and were shifted to 30°C at specific times. The approximate time and temperature shift (t-shift) protocols are detailed in S8 Fig. Result text in each section refers to specific t-shift protocol used.
For Hth and Exd knockdown in Dfd-SEG region, embryos were shifted immediately after egg collection to 30°C after 12 hrs of egg laying (from early embryonic stages) and dissected at late L3 (t-shift as shown in S8E Fig). For Notch knockdown in Dfd-SEG, embryos were allowed to grow at 18°C for 42 hours (from late embryonic stages) and then kept at 30°C until dissection at late L3 (t-shift as shown in S8B Fig).
For AbdA over-expression experiments, males carrying the UAS-AbdA and enhancer-lacZ (F3-lacZ, F4-lacZ, F3B3-lacZ and 717-lacZ) transgenes were crossed to females of the genotype UAS-dcr2; inscGal4,UAS-mCD8-GFP; tubGAL80ts and were allowed to lay eggs at 18°C for 12 hours. Eggs were reared for 6 days at 18°C (till early L3 stage) and then shifted to 30°C for 7hrs for F3-lacZ and F3B3-lacZ (12hrs for F4-lacZ) and dissected immediately thereafter (t-shift as shown in S8F Fig).
MARCM clones were generated as described previously [60]. Embryos were collected over a period of six hours at 25°C and were heat-shocked at 37°C for 60 minutes every 12 hours starting from 24 hrs AEL to 96 hrs AEL and the larvae were dissected at late L3.
VNCs were dissected from larvae of desired stage and fixed in 4% paraformaldehyde in 1X PBS containing 0.1% TritonX-100 for 45 minutes. The following primary antibodies were used: rabbit anti-Dpn, 1/5000; rat anti-Dpn, 1/2000; mouse anti-Dpn, 1/2000 (Bioklone, Chennai); rabbit anti-Grh, 1/2000 (Bioklone, Chennai); mouse anti-Grh, 1/2000; rabbit anti-Dfd 1/500; mouse anti-AbdA, 1/2000; mouse anti- Ubx/AbdA (FP6.87,DSHB), 1/20; rabbit anti-exd and anti-Hth (GP-52) (R. Mann) 1:100; mouse anti-exd (EXD B11M, DSHB, R. White) 1/20; rabbit anti-β-gal (Cappel) 1:1000; chicken anti-β-gal (ab9361, Abcam), 1/2000; mouse anti-NICD (C17.9C6, DSHB, Artavanis-Tsakonas, S). Secondary antibodies conjugated to Alexa fluorophores from Molecular Probes were used: AlexaFluor405 (1/250); AlexaFluor488 (1:500); AlexaFluor555 (1/1000); and AlexaFluor647 (1/500). The samples were mounted in 70% glycerol and images were acquired with Zeiss LSM 510 Meta and LSM 700 confocal microscopes and processed using ImageJ and Adobe Photoshop CS2. In all images yellow arrowheads indicate pNBs and a dotted line shows approximate thoracic (T) and abdominal (A) boundary except for Fig 3 (where it is specified the figure legends). Scale bars are shown in figures. Microsoft Excel and GraphPad Prism was used for all the data analysis (unpaired student t-test was done to check significance of the data).
EMSAs were performed as previously described [61]. All DNA binding experiments were performed with 6X His-tagged forms of Grh (residues 551–1333), AbdA (residues 261–590), full length Exd co-purified with HM domain of Homothorax. The following 5X binding buffer was used: 20% glycerol, 1mg/ml BSA, 2.5mM EDTA, 50mM 1M tris pH 7.5, 50mM KCl, 5mM MgCl2, 2.5mM DTT, 90 μg/ml polyDIDC. All binding reactions were set up in a 20μl volume and incubated at room temperature for 30 minutes.
Bacterial cultures expressing truncated GST tagged AbdA (261 to 590 aa) and His tagged Grh (551–1333 aa) were induced for two hours with 0.5mM IPTG at 18°C. Bead bound GST-AbdA and GST were incubated separately with equal amount of His-Grh lysate for 12 hours at 4°C. Pulled down and bead bound proteins were separated by denaturing SDS–PAGE and then transferred on to Polyvinylidene fluoride membranes (66543, PALL Life Sciences, Bio Trace). The membrane was blocked in 5% skimmed milk in Tris-buffered saline with 0.1% tween 20 (TBST). Primary antibodies- mouse anti-GST (sc-138, Santa Cruz Biotechnology) and mouse anti-His (H1029, Sigma- Aldrich) were diluted 1 in 5000 in 5% milk in TBST and the blot was incubated overnight at 4°C. HRP conjugated secondary antibody (Peroxidase-AffiniPure Rabbit Anti-Mouse IgG + IgM (H+L) (315-035- 048—Jackson Immunological Research Laboratory USA) (1:5000) was used. Visualization was carried out by enhanced chemiluminescence detection (34087- SuperSignal West Pico Chemiluminiscent Substrate, ThermoFisher Scientific).
|
10.1371/journal.pntd.0004133 | Nodular Worm Infections in Wild Non-human Primates and Humans Living in the Sebitoli Area (Kibale National Park, Uganda): Do High Spatial Proximity Favor Zoonotic Transmission? | Nodular Oesophagostomum genus nematodes are a major public health concern in some African regions because they can be lethal to humans. Their relatively high prevalence in people has been described in Uganda recently. While non-human primates also harbor Oesophagostomum spp., the epidemiology of this oesophagostomosis and the role of these animals as reservoirs of the infection in Eastern Africa are not yet well documented.
The present study aimed to investigate Oesophagostomum infection in terms of parasite species diversity, prevalence and load in three non-human primates (Pan troglodytes, Papio anubis, Colobus guereza) and humans living in close proximity in a forested area of Sebitoli, Kibale National Park (KNP), Uganda. The molecular phylogenetic analyses provided the first evidence that humans living in the Sebitoli area harbored O. stephanostomum, a common species in free-ranging chimpanzees. Chimpanzees were also infected by O. bifurcum, a common species described in human populations throughout Africa. The recently described Oesophagostomum sp. found in colobine monkeys and humans and which was absent from baboons in the neighboring site of Kanyawara in KNP (10 km from Sebitoli), was only found in baboons. Microscopic analyses revealed that the infection prevalence and parasite load in chimpanzees were significantly lower in Kanyawara than in Sebitoli, an area more impacted by human activities at its borders.
Three different Oesophagostomum species circulate in humans and non-human primates in the Sebitoli area and our results confirm the presence of a new genotype of Oesophagostomum recently described in Uganda. The high spatiotemporal overlap between humans and chimpanzees in the studied area coupled with the high infection prevalence among chimpanzees represent factors that could increase the risk of transmission for O. stephanostomum between the two primate species. Finally, the importance of local-scale research for zoonosis risk management is important because environmental disturbance and species contact can differ, leading to different parasitological profiles between sites that are close together within the same forest patches.
| Nodular worms frequently infect primates, pigs and ruminants. These intestinal nematodes induce inflammatory masses in the colon wall that cause severe abdominal pain, diarrhea, weight loss, and potential death. Through microscopic and molecular analyses, we studied the presence of nodular worms in three non-human primates (chimpanzees, baboons, black and white colobus) and humans inhabiting the Sebitoli area, at the extreme north of Kibale National Park in Uganda. Three different Oesophagostomum species were identified in the primates studied and we confirmed the existence of a recently described clade in baboons. Because the Sebitoli chimpanzees displayed a high prevalence of infection and because a high spatiotemporal overlap between humans and apes occurred in our study area, the risk of transmission of O. stephanostomum between the two species cannot be neglected. Thus, our results add to our understanding of nodular worm infection in location where non human primates and humans are co-existing, and underline the necessity to conduct further research at a local scale in a public health concern.
| Emerging zoonotic diseases are a serious threat to public health and animal conservation. This is especially true for apes, whose close phylogenetic relationship with humans increases the risk of zoonotic transmission between them. Although humans have always shared habitats with non-human primates, the dynamics of their relationships are rapidly changing nowadays. Indeed, non-human primate populations suffer from forest loss and fragmentation [1–3] and an increasing number of them live in anthropogenically disturbed habitats such as farmlands, human settlements, fragments of forest, and isolated protected areas [4–6]. As a consequence, people and non-human primates live in increasing spatial proximity to each other [7]. So far, several cases of pathogen transmission have been reported to have occurred between non-human primates and humans; these include the transmission of viruses (e.g. [8–10]), bacteria (e.g. [11–13]) as well as blood-borne parasites (e.g. [14–16]), and intestinal parasites (e.g. [17–21]).
Nematodes of the genus Oesophagostomum are intestinal parasites, which frequently infect primates (including monkeys, apes and humans), domestic and wild pigs and ruminants [22, 23]. Uninodular oesophagostomosis (i.e. Dapaong tumor) and multinodular oesophagostomosis [24], which are caused by O. bifurcum in humans [25], have been reported in endemic foci in West Africa (Togo and Ghana) with an estimated 250,000 infected people and a further one million at risk of contracting these parasitic diseases [26]. Third stage larval development in the colon wall induces the aforementioned inflammatory masses that cause severe abdominal pain, diarrhea and weight loss, and occasional death from peritonitis and intestinal occlusion [24]. While a variety of drugs kill Oesophagostomum nematodes, these drugs appear to be less effective on the tissue-dwelling stage because of the difficulty they have passing through the nodule wall [25]).
Only nine cases published in six studies before the 1980s [23, 27] and six cases published in one study in 2014 [18] reported the presence of human oesophagostomosis in Uganda. The few reported cases might be related to a low rate of infection. Nevertheless, oesophagostomosis infections may also be underdiagnosed, notably because obtaining a definitive diagnosis by ultrasound examination [26] is rarely undertaken in Ugandan hospitals and dispensaries [18]. Because transmission occurs through ingestion of the infective third-stage larvae [25] present in water, in food or on the ground, oesophagostomosis is a potential zoonotic risk when humans and non-human primates share the same habitats. In Ghana, identification of genetic differences among Oesophagostomum nematodes found in different primate hosts suggested that infection with these nematodes was rarely zoonotic [28, 29]. Still, the risk of zoonotic infection from the presence of infected chimpanzees in the vicinity of humans was mentioned in a study conducted in Kibale forest in western Uganda [17], while a recent study undertaken in the same area described a novel Oesophagostomum clade that infects humans and five sympatric species of non-human primates [18]. Because of the severity of the clinical consequences of oesophagostomosis, it should not remain a neglected area of public health. Surprisingly, while captive chimpanzees suffer from oesophagostomosis, no evident clinical signs have been observed in wild chimpanzees so far [30], except for one individual from Gombe (Tanzania) who developed weight loss as well as diarrhea prior to death, without other predisposing factors [31]. It has been suggested that ingestion of rough leaves via swallowing might decrease chimpanzee infestation with the parasite. Hairy leaves have a mechanical effect in preventing third stage larvae from penetrating the colon wall and by abrading the intestinal wall thereby leading to the expulsion of the immature and adult worms present in the nodules and in the gut lumen [32]. Wild chimpanzees that harbor these parasites usually do not suffer lethal infections, whereas humans can die from oesophagostomosis.
More specifically, the Sebitoli area, located in the extreme north of Kibale National Park (KNP) in Uganda, is an area with high anthropogenic pressure. Indeed, the human demographic density is high at the forest borders (villages with croplands, tea and eucalyptus plantations, and tea factories) and a tarmac road crosses the forest and the Sebitoli chimpanzee home range [33]. Additionally, the Sebitoli forest was commercially logged in the 1970s [34, 35] and is today mostly degraded and regenerating with 70% of the land cover affected [36]. Studies conducted in Sebitoli but also in KNP and in close-by forest fragments have shown that non-human primates (i.e., chimpanzees, Pan troglodytes; baboons, Papio anubis; black and white colobus, Colobus guereza; redtail monkeys, Cercopithecus ascanius; and vervet monkeys–Chlorocebus pygerythrus) frequently feed on croplands at the forest edge [37, 38]. At these sites, farmers rank baboons as the worst pests among the non-human primates because they regularly forage in large groups on several different crops at varying stages of maturity [37, 39]. Since then, baboons have come to represent a particularly important risk for people living close to the forest area, equally to chimpanzees, which are less frequent crop raiders but our closest relatives. Also, in the Kibale region, almost 9% of the population has reported direct contact with non-human primates [40] via touching carcasses (60.8% of cases) or butchering these animals (16% of cases); this is important because these activities represent a high-risk for zoonotic transmission of pathogens. In addition to poachers, other humans (e.g. researchers, field assistants and rangers) also regularly enter the forest and are in close proximity with the non-human primates living there. These observations underline the necessity to determine which species pose a risk of transmission in an environment where the risk of zoonosis appears to be particularly important.
This study aimed to investigate Oesophagostomum infection in terms of parasite species diversity, rate of infection and parasite loads, in four primate species (humans, chimpanzees–an ape species and closest relative to humans, baboons–a terrestrial monkey species, black and white colobus–an arboreal monkey species) to better understand the zoonotic risk associated with increased spatial proximity locally between humans and wildlife in an area subject to a high rate of environmental disturbance.
KNP is located in southwestern Uganda (0°13′ to 0°41′N and 0°19′ to 30°32′E), covers 795 km2 [41] and declines in elevation from 1590 m in the north to 1110 m in the south [42], while temperatures in the park range from 23.3 to 24.2°C (annual mean daily maximum; [43]). The park, home of 13 non-human primates species [44], is a mosaic of mature forest (58%), colonizing forest formally used for agriculture (19%), grassland (15%), woodland (6%) and lakes and wetlands (2%) [45] (Fig 1).
At Sebitoli, a long-term research project the “Sebitoli Chimpanzee Project” was initiated in 2008. The chimpanzee habituation level in this area did not allow researchers to collect identified feces from these animals at the time of the study (see below) and other non-human primates are not under habituation. Human pressure around Sebitoli is high. In fact, the human population density within 5 km of the boundary is ~260 inhabitants/km2 to the west and 335 inhabitants/km2 to the east of the park [46], and 82% of the Sebitoli chimpanzee home range borders are in contact with anthropogenic features [36]. Additionally, an asphalted road with high traffic intensity linking Kampala to the Democratic Republic of Congo crosses the forest [33].
Human fecal samples (N = 326) were collected during five different periods (July–August 2010, July–August 2011, March–April 2012, February–March–April 2013, December 2013–January 2014) from people in six villages less than 500 m from the border of the park (Fig 1). Some people were sampled several times at a minimum of 1-month intervals. Participants received instructions on how to collect and store the fecal samples and researchers retrieved them within 1 day of collection.
Fecal samples from baboons (Papio anubis) (N = 97) and black and white colobus (Colobus guereza) (N = 96) were collected during the first three study periods and fecal samples from chimpanzees (Pan troglodytes schweinfurthii) (N = 228) were collected during the five study periods. These samples (< 6 hours old) were collected non-invasively in the forest and immediately transferred to plastic bags. It is likely that samples from the same animals were collected several times because we did not know the identity of the individual from whom the stool was collected. Also, the sample sizes were larger than the number of animals within the groups or community studied. Fecal samples were collected from a relatively large area of the forest (25 km2).
Fecal samples from humans and non-human primates were inspected before processing them to check for the presence of macroscopic parasites and to note the consistency (liquid, soft or pasty, solid or normal, and dry or hard, according to a method published previously [47]).
Two grams of a fresh fecal sample from a human or a non-human primate was preserved in 18 mL of 10% formalin saline solution. These samples were analyzed at the Department of Parasitology (Ecole Nationale Vétérinaire d’Alfort, France) and a direct microscopic examination of two 50-μL smears was performed to access the presence of hookworm-like eggs (at 100–400x magnification after homogenization). Because collecting fecal samples from different animal species and from humans living in separated villages was time consuming, as well as the fecal sample storage for both microscopy and molecular analyses, we were unable to conduct systematically other microscopic examination methods such as fecal flotation or sedimentation on fresh samples [48]. Each egg was identified according to its size, color, shape and morula aspect. Eggs were not classified at the genus level because the Oesophagostomum genus cannot be distinguished with certainty from hookworm nematodes (i.e. Ancylostoma sp., Necator sp.) by microscopy alone. To establish an arithmetic parasite load (eggs per gram of feces; epg), the total number of eggs, larvae and adults from a 100-μL aliquot was counted and multiplied by 100. Then, a corrected parasite load (CPL) was obtained according to the stool consistency (i.e. x2 when the feces were soft, x3 when the feces were liquid [49]).
Fecal samples (N = 15 baboon samples; N = 22 black and white colobus samples; N = 39 chimpanzee samples; N = 39 human samples) were randomly selected and then stored differently according to the study periods: (1) at least 4 g was diluted in 18 mL of 95% ethanol (2010); (2) at least 4 g was diluted in 18 mL of 95% ethanol over a 24 h period, after which the supernatant was removed and the sedimented feces dried on a silica gel beads (2011, 2012); and (3) 10 mL of coproculture products were stored in 50 mL of 95% ethanol (2012, 2013). One gram of fresh feces was mixed with charcoal and vermiculite, and cultured for 10 to 17 days in a Petri dish at room temperature (approximately between 18°C and 26°C). During the culture, regular inspection was done to keep the culture moist and to softly stir the mixture to minimize fungal growth. At the end of the culture, all the mixture was transferred on two layers of gauze and then larvae products were collected via the Baermann procedure, described in [48].
Molecular analyses were performed at the Eco-anthropology and Ethnobiology Laboratory (National Museum of Natural History, France). DNA was extracted from coproculture products, feces in ethanol or dried feces. DNA from 10 mL of a larval culture was extracted with a QIAamp DNA Mini Kit Tissue (Qiagen, Chatsworth, CA, USA) according to manufacturer’s protocols but with the following modifications. In step 1, 1 mL of phosphate-buffered saline (PBS) was added to the sample then mixed thoroughly by vortexing, centrifuged and the supernatant removed (3 times), followed by an overnight incubation at -20°C with the ASL buffer included in the kit. DNA was extracted from 100 mg of dried sample or from 2 mL of ethanol sample with a QIAamp DNA Stool Kit (Qiagen). We made a modification to step 1 where 1 mL of PBS was added to the sample, the sample allowed to sit for 10 min at room temperature, and then incubated overnight at 70°C with ASL buffer. An external PCR targeting the ribosomal internal transcribed spacer 2 gene (ITS2) using NC1 and NC2 primers [50], followed by an internal semi-nested PCR using OesophITS2-21 [18] and NC2 primers were performed. PCR reactions were cycled in a BioRad CFX (Bio-Rad Laboratories, Hercules, CA, USA) with a mix of sterile water, Taq polymerase, buffer, primers, dNTPs, fluorochrome and an intercalating agent (Ssofast Evagreen).
The following temperature profile was used for the external PCR: 94°C for 2 min; 45 cycles of 94°C for 10 sec, 60°C for 30 sec, 72°C for 1 min, and a final extension at 72°C for 10 min. The semi-nested PCR followed a slightly different temperature profile: 95°C for 2 min; 45 cycles of 95°C for 10 sec, 55°C for 30 sec, 72°C for 1 min; and a final extension at 72°C for 5 min. The PCR products were sequenced at the Pasteur Institute (Lille, France) by the Genoscreen Laboratory using primer NC2 and OesophITS2-21. DNA sequences were hand-edited and cleaned with 4Peaks software.
Sequence alignments were performed on SeaView software by inputting our sequences with the sequences obtained by Ghai et al. [18] (accession numbers: KF250585 –KF250660) and by Krief et al. [17] (KT592234, KT592235). In addition, we included three Oesophagostomum stephanostomum reference sequences (AF136576, AB821022, AB821031), one O. bifurcum sequence (AF136575) and five outgroups (HQ844232, Y11736, Y11735, Y10790, AJ006149). Phylogenetic trees were established using the maximum likelihood method in MEGA [51] and the Hasegawa-Kishino-Yano substitution model with five discrete gamma categories [52]. To assess the phylogenetic robustness of the tree, 1000 bootstrap replicates were performed.
All the relevant sequences have been deposited in GenBank under the accession numbers: KR149646 –KR149658.
Special attention was paid to avoiding contamination during all the process stages, that is, during the field collections (utilization of gloves and tongue depressors), during sample storage (use of sterile instruments on different days of collection for each primate species), and during DNA extraction and amplification (separation of the samples by species on the plates and repetitions). The percentage values of the fecal samples that were positive for hookworm-like eggs were considered a proxy for the infection prevalence and the mean corrected parasite load (including infected and non-infected samples) as a proxy for the infection intensity. All statistical tests were performed via R software [53] and were two-tailed with the criterion of statistical significance set at P < 0.05. When samples sizes were small or data were not normally distributed, nonparametric procedures were used.
The Uganda National Council for Science and Technology, the Uganda Wildlife Authority and the National Museum of Natural History in France (Memorandum of Understanding SJ 445–12) reviewed and approved the animal care and human research protocols. The free-ranging chimpanzees and monkeys were studied without invasive methods and without interacting with the researchers. Additionally, we obtained the approval of each village chairperson to conduct our research, and human volunteers gave their written informed consent. All volunteers were free to withdraw from the study at any time. The purpose, methods and preliminary findings of the research were explained to all volunteers. Each fresh human sample collected herein was analyzed microscopically within 12h of collection via a fecal flotation to ascertain the parasite species present [49] and to immediately inform the volunteer whether or not he or she had an infection. Following the recommendations of the local dispensary in the area, a single dose anthelmintic treatment (albendazole) was given to any person infected with nematodes (hookworm-like species, Ascaris lumbricoides or Trichuris trichiura). Persons who received anthelmintic drug treatment could be resampled but at 8-monthly intervals as a minimum.
The proportions of samples containing hookworm-like eggs varied significantly among the host species (Chi-square = 395.2; df = 3; P<<0.001). Chimpanzees had the highest percentage of positive samples (77.2%; 176/228), followed by baboons (71.1%; 69/97). Humans (6.4%; 21/326) and black and white colobus (2.1%; 2/96) were the two species with low proportions of samples positive in hookworm-like eggs. The mean hookworm-like egg load was also significantly higher in chimpanzee feces (CPLmoy = 535 epg) and in baboon feces (CPLmoy = 343 epg) compared with human feces (Mann-Whitney tests: W = 64443; P<<0.001 and W = 26443; P<<0.001, respectively) and in black and white colobus feces, whose mean corrected parasite loads were < 100 epg (Mann-Whitney tests: W = 19512; P<<0.001 and W = 8013; P<<0.001, respectively) (Fig 2). During the dry season, the parasite load was higher in baboons than in chimpanzees (Mann-Whitney test: W = 2269; P<0.01) but in the wet season, it was three times higher (statistically significant) in chimpanzees than in baboons (Mann-Whitney test: W = 3071; P<0.001). In comparison with the dry season, during the wet season, the mean hookworm-like egg load was significantly lower in baboon feces (Mann-Whitney test: W = 551; P<<0.001) while it was higher in chimpanzee feces (Mann-Whitney test: W = 7014; P<<0.001) (Fig 2).
PCR products from 61 out of the 115 fecal samples tested (92.3% (36/39) of the chimpanzee samples, 93.3% (14/15) of the baboon samples, 36.4% (8/22) of the black and white colobus samples and 28.2% (11/39) of the human samples) produced interpretable DNA sequences. Approximately 50% of the samples stored in ethanol or in ethanol followed by silica gel and more than 75% of the samples that generated coproculture products gave interpretable DNA sequences (Table 1).
Fifty-seven DNA sequences (34 from chimpanzees, 14 from baboons, 5 from black and white colobus and 4 from humans) matched the Oesophagostomum ITS2 sequences already published. Sequences corresponding to O. stephanostomum were found in 82.1% (32/39) of the chimpanzee samples, 18.2% (4/22) of the black and white colobus samples and 10.3% (4/39) of the human feces samples. Sequences from one man (60 years of age) and three women (16, 32 and 41 years of age) clustered with the published O. stephanostomum sequence. Sequences matching with O. bifurcum were obtained from 80% (12/15) of the baboon samples, from 5.1% (2/39) of the chimpanzee samples and from 4.5% (1/22) of the black and white colobus sample (Fig 3). Only 13.3% (2/15) of the baboon fecal samples were positive for the new sequence type of Oesophagostomum sp., recently described in Ghai et al. [18] (Fig 3).
Microscopic examinations showed that despite the high infection prevalence in olive baboons and the low infection prevalence in humans being common between sites, Sebitoli black and white colobus monkeys had a low prevalence of infection, as has been described previously in Kanyawara by Gillespie et al. [54] but not by Ghai et al. [18]. The infection prevalence in chimpanzees and the arithmetic mean corrected for parasite load were significantly higher in Sebitoli than in Kanyawara (Mann-Whitney test: W = 6020, P<0.001) (Table 2).
At the molecular level, the Oesophagostomum clades obtained from chimpanzees in Sebitoli and Kanyawara were similar in terms of the predominance of O. stephanostomum and O. bifurcum (less frequent) (Table 3). All of the Kanyawara baboon samples were positive for the O. bifurcum clade. In addition to O. bifurcum, the newly described Oesophagostomum clade was identified in 14.3% of the Sebitoli baboon fecal samples (Table 3). The rate of Oesophagostomum infestation of human feces and colobus feces was twice as low in Sebitoli as in Kanyawara. Specifically, 60% of the Oesophagostomum species in Kanyawara black and white colobus monkeys comprised the newly described Oesophagostomum sp. clade [18], whereas 80% were O. stephanostomum in Sebitoli colobine monkeys but Oesophagostomum sp. was not detected. Humans living in the Sebitoli area harbored only O. stephanostomum, while villagers from the Kanyawara area were only infected by Oesophagostomum sp. (Table 3).
In the present study, microscopic and molecular approaches were used to reveal the prevalence and parasitological load of Oesophagostomum sp. in three non-human primate species and humans living in close proximity in a forested area. Our results provide the first evidence that some humans living in the Sebitoli area are infected by O. stephanostomum, a common species in free-ranging chimpanzees. Moreover, the chimpanzees also harboured O. bifurcum, a species commonly described in humans. Finally, the existence of the new clade Oesophagostomum sp. described in black and white colobus monkeys and humans in Kanyawara (a neighboring site to Sebitoli) by Ghai et al. [18], was confirmed in the Sebitoli region as two baboon fecal samples were infected with it. Microscopy revealed that the infection prevalence and the parasite load were significantly higher in the Sebitoli chimpanzees than in the Kanyawara ones.
Sebitoli chimpanzees had a high infection prevalence compared with colobine monkeys and humans. While the Oesophagostomum species isolated from Sebitoli and Kanyawara chimpanzees are similar, both the prevalence of infection and the corrected arithmetic mean parasite load were higher in the Sebitoli apes. Without additional data, it appears to be difficult to attribute these differences to specific causes. Indeed, individuals may differ in terms of infection rates and parasitic loads according to demography (1.5 individuals/km2 at Kanyawara vs. 3.2 individuals/km2 at Sebitoli, leading to increased transmission among individuals; [58, 59]) and behaviour (e.g. higher association strength between members of the same community with increased grooming sessions; [60]). These observations might also result from a difference in the chimpanzees’ physiology, immunity, and environment (e.g. difference in proximity to humans and their livestock). Because of their proximity to humans, Sebitoli chimpanzees are more affected by stress, which can be evidence through increased signs of anxiety when chimpanzees leave the forest to crop raid at the borders, and when they cross a tarmac road with high traffic cutting their home range [33, 38]). Stress could decrease their immunity level [61, 62], and thereby make them more susceptible to parasitic infections. Additionally, about 30% of the chimpanzees have limb deformities caused by poaching, 10% of the individuals suffer from facial dysplasia [63] and one of them has a cleft lip [64]. Such mutilations and congenital diseases—suspected to be caused by prenatal exposure to teratogen chemicals—may be associated with other health disorders in the affected individuals and decrease individual’s immunity to pathogens such as parasites [65–67]. Other individuals—without abnormal phenotypes—may also experience effects of such exposure.
A similar prevalence of Oesophagostomum spp. and other hookworms was observed in villagers living in Sebitoli and Kanyawara (28.2% of the Sebitoli villagers and 25% of the Kanyawara villagers), but two different Oesophagostomum clades were distinguished: O. stephanostomum in Sebitoli and Oesophagostomum sp. in Kanyawara. O. stephanostomum is common in non-human primates, particularly great apes [17, 18]. In the present study, humans infected with O. stephanostomum came from Sebitoli village (one man and two women) and from Kyansimbi village (one woman). They were used to seeing chimpanzees and baboons in their gardens, at locations less than 500 m from the forest edge. Interestingly, one of them was used to sleeping in a small hut to prevent animal crop raiding during the night. However, Sebitoli chimpanzees often feed in maize gardens in large parties and they can stay a long time in croplands [38], likely contaminating the fields when defecating. In addition, maize begins to ripen and be consumed by both humans and chimpanzees at the end of the wet season (Cibot et al., submitted) when the rate of Oesophagostomum infection in chimpanzees was the highest. This period poses the highest risk of pathogen transmission through the fecal-oral route. Taken together, these results and observations showing a high degree of spatiotemporal overlap between humans and chimpanzees, which are phylogenetically close species, represent factors that could enhance the risk of transmission for O. stephanostomum between chimpanzees and people.
In this study, black and white colobus had a low prevalence of infection, which could be related to their arboreality (making them less vulnerable to nematode parasites with a life cycle including soil [18, 68]) and to their diet (important ingestion of secondary compounds in leaves with potential anthelminthic properties [30]), and they are rarely in contact with humans (Cibot et al., submitted). While oesophagostomosis lesions in baboons are identical to the ones described in chimpanzees and humans (nodules in the intestinal wall; [31, 69]), O. bifurcum, which appears to be the most common species in Sebitoli baboons, was not found in any of the villagers. Nevertheless, because Sebitoli baboons are reported to be the most frequent crop raiders (Cibot et al., submitted) and two baboon fecal samples were infected with the new Oesophagostomum sp. clade described in humans from Kanyawara, we still cannot exclude the potential disease transmission between the two species. Surprisingly, in baboons, the mean corrected parasitic load of hookworm-like eggs was higher during the dry season compared than it was in the wet season (the opposite of what was observed in humans and chimpanzees in this study). Indeed, lower ambient temperatures and higher humidity rates likely favor survival of eggs and larvae in feces, increasing the risk of infection during the wet season [70]. A long-term survey with an increased sample set should be initiated to confirm this result and investigate why an opposite seasonality pattern exists between chimpanzees and baboons.
Our findings raised the need for better public health awareness of oesophagostomosis in the Kibale region. Further studies should be conducted to better understand the epidemiology of Oesophagostomum infections in Uganda, and the role played by domestic animals (cows, sheep, goats, pigs), and other wild animals (antelopes, buffalos, wild pigs) in its transmission. Indeed, a recent study undertaken in Tanzania revealed that crop fields regularly used by both chimpanzees and domesticated animals represented potential hotspots for Cryptosporidium transmission [71]. However, when comparing studies, we should interpret carefully data because different methods may have been used. Indeed, in the present study, we detected N. americanus using the OesophITS2-21 primer, which was supposed to be specific and only allow amplification of the Oesophagostomum genus [18]. This result could be caused in part by a difference in the storage procedure between the two studies, with coproculture storage likely favoring the amplification of Necator sp. Similarly, we need to be cautious with the prevalence we obtained after PCR and sequencing since we amplified materials issued from different methods and we sampled unidentified individuals for wild primates compared to other studies. Indeed, the relatively low prevalence in Oesophagostomum spp. in humans and in colobine monkeys could result from an underestimation of all the nematode species or of certain species of Oesophagostomum, due to the culture of fecal samples, which could allow differential development of larva-stage nematodes and which is less sensitive than other methods (e.g. agar plate culture, which requires sterilization systems that are not available in field settings). Then, we likely underestimate the public health problem. Finally, we should also remain prudent when we employ the term “species” in our study, since DNA amplification was only based on a short region of a single gene. While we demonstrated a relatively high genetic diversity within the Oesophagostomum genus, the sequencing of additional genes or the morphological identification of the third stage larvae and adults worms should be established to confirm the species level differentiation.
As Bortolamiol et al. [36], comparing three sites within the same park including Sebitoli and Kanyawara, revealed that small-scale analysis was needed to obtain a better understanding of chimpanzee diet, repartition, density and land use, the parasitological profiles were different between sites and showed that research is required at a more local scale for zoonosis management. For example, the higher prevalence of Oesophagostomum spp. in colobine monkeys observed in Kanyawara compared to our present study may be explained by a difference in the Oesophagostomum species harbored in the black and white colobus monkeys, which could have consequences for zoonosis management. Moreover, working at a small scale is also essential for the human-wildlife zoonotic management as local knowledge and traditional beliefs can differ significantly between people living in relatively close locations and even within the same villages. In fact, near the KNP area, different ethnicities live within the same villages [37] and an increasing number of migrants, notably Congolese people fleeing conflicts in the Democratic Republic of Congo, join western Uganda [72]. Migrants may have a different perception of the risks associated with living in close proximity to wild animals, or the risks associated with hunting animals and eating bush meat [40, 73]. In any case, health-risk education programs should be better integrated into conservation programs and measures against crop-raiding and other practices such as throwing food from passing vehicles on the Sebitoli road to feed baboons should be implemented. It should also be important to better follow Ugandan patients in hospitals and dispensaries with clinical examination and ultrasonography to better evaluate impacts on human health. Today, in Northern Togo and Ghana, the Oesophagostomum infections seem to have been eradicated after a large scale and intense mass treatment on humans [23]. Finally, our findings reinforce the fact that zoonotic parasites, in the context of increased proximity between non-human primates and humans, should be considered a priority concern for researchers, wildlife managers and health care systems.
|
10.1371/journal.pcbi.1004058 | Heterogeneous CD8+ T Cell Migration in the Lymph Node in the Absence of Inflammation Revealed by Quantitative Migration Analysis | The three-dimensional positions of immune cells can be tracked in live tissues precisely as a function of time using two-photon microscopy. However, standard methods of analysis used in the field and experimental artifacts can bias interpretations and obscure important aspects of cell migration such as directional migration and non-Brownian walk statistics. Therefore, methods were developed for minimizing drift artifacts, identifying directional and anisotropic (asymmetric) migration, and classifying cell migration statistics. These methods were applied to describe the migration statistics of CD8+ T cells in uninflamed lymph nodes. Contrary to current models, CD8+ T cell statistics are not well described by a straightforward persistent random walk model. Instead, a model in which one population of cells moves via Brownian-like motion and another population follows variable persistent random walks with noise reproduces multiple statistical measures of CD8+ T cell migration in the lymph node in the absence of inflammation.
| Migration is fundamental to immune cell function, and accurate quantitative methods are crucial for analyzing and interpreting migration statistics. However, existing methods of analysis cannot uniquely describe cell behavior and suffer from various limitations. This complicates efforts to address questions such as to what extent chemotactic signals direct cellular behaviors and how random migration of many cells leads to coordinated immune response. We therefore develop methods that provide a complete description of migration with a minimum of assumptions and describe specific quantities for characterizing directional motion. Using numerical simulations and experimental data, we evaluate these measures and discuss methods to minimize the effects of experimental artifacts. These methodologies may be applied to various migrating cells or organisms. We apply our approach to an important model system, T cells migrating in lymph node. Surprisingly, we find that the canonical Brownian-walker-like model does not accurately describe migration. Instead, we find that T cells move heterogeneously and are described by a two-population model of persistent and diffusive random walkers. This model is completely different from the generalized Lévy walk model that describes activated T cells in brains infected with Toxoplasma gondii, indicating that T cells exhibit distinct migration statistics in different tissues.
| A primary challenge of immunological imaging experimentation is to understand the nature of cell migration statistics, and the role that these statistics play in immune function. Over the last decade, two-photon microscopy has transformed the understanding of the role of cell migration in the immune response [1–3]. However, although improved statistical approaches are still being developed [4–6], many existing methods for analyzing migration statistics are susceptible to experimental artifacts that can lead to inaccurate conclusions about leukocyte behavior. Similar questions arise in analyzing the migration of organisms ranging from bacteria [7] to vultures [8] and human hunter-gatherers [9]. Migration tracks can be directional or random and can be characterized by a bewildering array of models. This poses the question of how best to analyze migration tracks in an unbiased fashion, given experimental data that is often gathered in a limited field of view over a short period of time.
Many immune functions are thought to be directed by chemotactic signals, and directional migration has been observed in numerous cases, such as neutrophil response to sterile inflammation [10], migration of positively selected T cells in the thymus [11], and T cell priming by dendritic cells in lymph nodes [12, 13]. While the directional bias in these studies is clear, they use measures of directionality that can be susceptible to experimental artifacts. These issues range from technical constraints such as the finite imaging field and global drift to the intrinsic limitations of widely-used quantities such as the meandering index and motility coefficient [2, 4, 5, 14, 15]. Such artifacts can affect quantitative analyses and can even lead to inaccurate qualitative conclusions in cases where directional motility is subtle.
Immune cell migration also has a stochastic component [1, 3, 6, 16–18]. Commonly used random walk analyses [14, 15] assume that cells obey Brownian statistics. In several cases, however, it has been argued that cells migrate via persistent random walks [5, 16, 19–22] or even exhibit Lévy behavior [6, 23]. Despite this knowledge, many analyses of random motion implicitly assume that the statistics are described by Brownian walks even at short time scales, by assuming that the mean-squared displacement increases linearly in time and extracting a motility coefficient [14, 15]. The accurate identification of the persistence time for persistent random walks [5, 16, 19–22], or of more exotic forms of migration statistics such as Lévy behavior of migrating microglia [23] or of CD8+ T cells in Toxoplasma gondii-infected mouse brains [6], requires a description that goes beyond use of the mean-squared displacement as a distinctive identifier of migration statistics.
A more complete and accurate description of migration statistics requires methods capable of detecting subtle directional bias that can also handle more general forms of random walks, given experimental data gathered over rather short time periods. Furthermore, in order to investigate the correlation between cell migratory behavior and immune function, it is necessary to develop a description that rigorously characterizes both stochastic and directional migration without any initial assumptions regarding the location of possible targets. Current methodologies are of limited use in achieving these goals. Here, we describe a set of analytical and computational methods that can be used to identify various types of directional, anisotropic (asymmetric), and stochastic migration. These methods can be applied to any type of motile organism or cell. In order to demonstrate the practical implementation of the methods, we apply them to the migration of CD8+ T cells in uninflamed mouse lymph nodes. Unexpectedly, this system is well-described by a model containing two populations of T cells, in which one population obeys Brownian statistics and the other population migrates via heterogeneous, persistent random walks. While this model shares similarities with previous persistent random walk and run/pause models [5, 16, 19–22], we show that it reproduces several key statistical measures beyond the mean-squared displacement alone. Our results show that CD8+ T cells in uninflamed mouse lymph nodes migrate differently from those in the brains of mice chronically infected with Toxoplasma gondii.
In the analysis that follows, the migration statistics of CD8+ T cells in uninflamed lymph nodes are described using green fluorescent protein (GFP)-expressing OT-I T cells that were transferred to C57BL/6 mice as described in [24] and the Methods section. During the period of 16–24 hours after transfer, T cells were imaged in excised lymph nodes in a 500 μm × 500 μm × 68 μm volume. In the following subsections, there are frequent references to numerical simulations of various walk models that were conducted to either illustrate our points or to describe our experimental data. We distinguish between results from simulations and from experiments by always identifying the walk model used in the simulation.
A pivotal question for many studies of immune cells is whether they migrate and respond to signals with specific directional motion or biases [2, 3, 14, 15]. More generally, anisotropic motion — motion that is not statistically identical in all directions — can indicate a directional bias, such as chemotaxis or chemokinesis due to a chemical gradient, or to an asymmetric feature of a particular direction, such as confinement. However, since even isotropic trajectories appear directed on short enough time scales, and, conversely, directed tracks typically contain a stochastic component, discerning directionality and anisotropy is not a simple task [15, 16].
The most commonly used method for identifying directional motion currently consists of plotting cell tracks with their starting points translated to the origin, measuring the MSD, calculating the meandering index, and measuring the mean displacement vector [5, 10–13, 15, 18]. However, as discussed in the following sections, these methods are only sensitive to obvious directionality, suffer from several experimental artifacts, and depend quantitatively on details of the experimental set up. Of these methods, even the quantitative tests are only capable of identifying global drift. Thus, they cannot detect other types of anisotropic motion, such as directed motion towards a single target (or scattered collection of targets) or Brownian-walk-like motion with different motility coefficients for different spatial directions. Other methods for identifying motion in a specific direction, such as measuring the component of velocity in that direction or angle of motion with respect to a target, require prior knowledge of the existence of a target or special direction [15].
Because of these issues, we developed and tested several techniques to detect and determine the amount and type of anisotropy. These methods are sensitive to small anisotropies and do not rely on having prior knowledge about the directional motion.
The mean-squared displacement, while straightforward to calculate, is highly susceptible to artifacts that can lead to misinterpretations of data. For example, typical imaging experiments can only visualize cells within a relatively small part of the space that cells can actually explore. Thus, as noted previously, cells may exit the field of view before the time series has ended, which can bias the analysis [5, 6, 15, 21, 30]. The magnitude of this effect can be estimated by calculating the mean time, 〈texit〉, for a Brownian-random-walking cell to reach the boundary of the imaging field, which has a shortest dimension (typically, depth) of length L [27]:
〈 t exit 〉 = L 2 12 D ,(1)
where D is the diffusion (motility) coefficient. For typical values derived from multiple studies of T cell movement, L ≈ 40 μm and D ≈ 30 μm2/min (e.g., refs. [1, 5, 6, 15, 28–30, 34]), 〈texit〉 is just 4.4 minutes, with some cells exiting the field of view even more quickly.
This limitation of imaging has especially significant consequences for the mean-squared displacement (MSD). Since fast-moving cells tend to leave the imaging field more quickly than others, data at late times becomes biased toward slow-moving cells [5, 15, 21, 30]. This can distort the shape and magnitude of the MSD as a function of time. Furthermore, this issue plagues alternatives to the standard practice [5, 14, 15] of measuring the motility coefficient from the slope of the best-fit line to the MSD versus time curve. Since the standard motility coefficient method is inaccurate due to short-time directional persistence [5], one may try either fitting only the MSD at late times to a line or fitting the MSD with a function of both the motility coefficient, D, and persistence time, tp [5, 21]. However, the first option exacerbates the finite imaging field problem, underestimating the diffusion coefficient by as much as 20% under simulated typical conditions. The second option is also inadequate because the fit parameters (motility coefficient and persistence time) are sensitive to the duration of the “early time” segment used. Alternatively, if long time segments are used, the fits converge on common parameters [5], but we have found in simulations that they underestimate the motility coefficient and persistence time by as much as 20% and 40%, respectively, due to cells exiting the imaging volume.
When considered together, these limitations have several practical implications. Since the mean time 〈texit〉, for a cell to leave the imaging volume is typically a few minutes, and many cells exit earlier than 〈texit〉, the MSD and quantities derived from the MSD are unreliable for rigorously assessing migration data. Counterintuitively, in a finite imaging field, measuring the MSD over longer time intervals can lead to erroneous conclusions rather than deeper insights. It should also be noted that the MSD does not uniquely specify an underlying model for random (or directional) motion [15, 35, 36]. Therefore, this measurement should only be used for qualitative comparisons between experiments with identical imaging dimensions or as a complementary consistency check for any proposed migration model.
Instead of focusing on the MSD, we rely on two main quantities to characterize the displacements: (1) the probability distribution, PΔt(r), of cell displacements, r, as a function of the time interval Δt, over which the displacement occurs and (2) the correlations, C(t1, t2;τ), between the displacements during one time interval τ starting at time t1, with displacements during a later time interval τ, starting at t2. Utilizing these quantities mitigates problems due to the finite field of view. Probability distributions, PΔt(r), reveal displacements of various sizes and at all times, instead of focusing on large displacements at late times, which are the most profoundly impacted by the limited field of view. The correlation function typically reveals features such as persistence in the early-time data, while minimizing significant artifacts.
The last decade has seen tremendous advances in the ability to image the behavior of lymphocyte populations [3], but there are several important limitations to imaging experiments and data analyses of cell tracks that have hindered efforts to interpret cell migration quantitatively. For example, the finite imaging volume and z-depth of experiments can skew the interpretation of migration data, and the effects of the finite image volume negate any advantage gained by imaging cell populations for longer times. Due to the effects of the shallow depth of the imaging volume, there are severe truncation effects due to cells prematurely leaving the field of view. It is important to recognize that these effects cannot be remedied by simply imaging over a longer time interval; instead, improvements require experiments with greater imaging volume and thus increased z-depth. Less obvious effects such as global drift can further obscure the true nature of cell migration; this may be mitigated by adjusting tracks according to the motion of auto-fluorescent particles. These experimental issues necessitate a comprehensive analysis that goes beyond standard measures such as the mean-squared displacement and meandering index.
In addition, several tests for anisotropy and directionality are required because different measurements capture different types of anisotropic behavior. To directly analyze migration statistics, one should construct probability distributions of displacements over various time intervals and measure correlations in cell trajectories. Even without experimental artifacts such as the limited field of view, common measures such as the mean-squared displacement are not sufficient to distinguish details of cell migratory behavior without further information. For instance, a major drawback of the MSD analysis is that different models may produce identical MSD curves, yet differ in a variety of key aspects including mean velocity or persistence time. We have presented quantitative methods to minimize and account for these limitations, and quantitatively describe cell migration. Together, these techniques minimize errors due to drift (Fig. 1), reliably detect anisotropic cell migration (Fig. 2), and provide a strong connection between migration models and experimental data (Fig. 3).
With these methods, we have found that the migration of CD8+ T cells in lymph nodes in the absence of inflammation is reasonably well-described by a model with two distinct populations of stochastic walkers (Fig. 3). The first population migrates by a pure Brownian walk. The other population, comprising most of the walkers, migrates by a persistent random walk and is subject to a small amount of Brownian noise. This model can also be interpreted as the aggregate of a paused population and an active population with a small amount of overall noise.
While this model has some similarities with existing models, it differs from previous models for CD8+ T cell migration in the lymph node, which describe cells as a single, homogeneous population of persistent random walkers [5, 19–22]. Our model explicitly incorporates heterogeneity in the CD8+ T cell population. Furthermore, in contrast to previous run/pause models for T cells in the lymph node [21], runs and pauses are not well-mixed. Instead, in our refined model, cells that are essentially paused except for slow Brownian-like motion, remain in the paused state for many minutes at a time (the entire duration), and similarly, cells migrating by persistent random walks move continuously for at least ten minutes. This model, in contrast to existing models and common practice [1, 41–43], does not choose an arbitrary speed cutoff below which cells are assumed to be paused and data is discarded. Most importantly, this model successfully describes multiple cell migratory statistical measures (Fig. 3), rather than just the MSD.
There are a variety of possible explanations for the observed heterogeneity in migratory behavior. For instance, in these experiments, CD8+ T cells were imaged throughout the lymph node, and thus, likely in multiple zones within the lymph node. Thus, it is possible that the two populations in the model represent migration in distinct regions of the lymph node. Alternatively, previous studies using this experimental system have shown that this population of OT-I CD8+ T cells express variable levels of the chemokine receptors CCR5 and CCR7 [24], which could impact migration. Finally, it is unlikely that the paused population of cells arises from an experimental artifact such as phototoxicity. Indeed, paused cells are alive since they exhibit shape fluctuations and even, after very long time scales, begin to migrate.
Interestingly, T cells in uninflamed lymph nodes do not migrate via generalized Lévy walks, as activated T cells do in the brain during chronic toxoplasmosis [6]. While generalized Lévy walks may enable T cells to efficiently find rare target parasites [6], the long runs in the Lévy walk may be less beneficial for T cells that must frequently interact with dendritic cells. The observed differences in migration statistics may be indicative of the cell extrinsic or intrinsic differences between the two cell populations. For instance, it is likely that the structural features within these tissues, which may act as a scaffold for cell crawling [1, 21, 44], are different in the two tissues. In addition, the T cell population in lymph nodes in the steady state is not as activated as the cells in the brain, which could also affect migratory behavior. Together, these observations suggest that CD8+ T cell populations with distinct functions migrate differently.
While the model is successful in characterizing many aspects of cell migration (Fig. 3), there are deviations from measured cell statistics. These differences reflect the difficulty in systematically identifying a simple model that accurately and comprehensively describes walk statistics. Both fluctuations within individual cells and variations within cell populations complicate the overall behavior and ensuing analysis. This problem may be mitigated by accumulating better statistics; for example, if individual cells could be followed for hours throughout the entire lymph node, instead of minutes in a small volume, the whole-population analysis could instead be carried out for individual cell tracks. Additionally, further improvements to analytical and computational methods will lead to more accurate cell migration modeling. However, while various measures and techniques to understand and describe migratory behavior have been developed [4–6, 15, 36, 45–49], less has been done to systematically build robust migration models. The methodology described in this paper, combined with generalizations of new techniques for analyzing heterogeneous migration statistics [36, 46, 48, 49], could achieve this goal.
In general, use of these statistical approaches require relatively large amounts of data (more than 100 cell tracks) and numerical simulations of random walk models. Despite these difficulties, these methods can provide powerful insights into cell migratory behavior, and they will be useful for characterizing migration in future studies. In turn, developing these more accurate models will help connect cell migration to immune function and lead to a deeper understanding of immune response.
All procedures involving mice were reviewed and approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania (Animal Welfare Assurance Reference Number #A3079–01) and were in accordance with the guidelines set forth in the Guide for the Care and Use of Laboratory Animals of the National Institute of Health.
CD8+ T cells were isolated as previously described [24]. Briefly, cells were isolated from the spleen and peripheral lymph nodes of DPE-GFP OT-I transgenic mice (OT-IGFP). Single cell suspensions were obtained by mechanical homogenization. Red blood cells were removed by hypotonic lysis. T cells were purified using the mouse T cell enrichment columns (R&D systems, Minneapolis, MN). 2–5 × 106 purified OT-IGFP cells were injected into recipient mice intravenously.
Mice were euthanized by CO2 asphyxiation 16–24 hours following T cell transfer. The mesenteric lymph nodes were removed immediately, with minimal mechanical disruption. The lymph nodes were immobilized in 1% agarose in a heated chamber where specimens were constantly perfused with warmed (37°C), oxygenated media (phenol-red free RPMI 1640 supplemented with 10% FBS, Gibco). The temperature in the imaging chamber was maintained at 37°C using heating elements and monitored using a temperature control probe. Imaging was performed using a Leica SP5 multi-photon microscope system (Leica Microsystems, Mannheim, Germany) equipped with a resonant scanner, picosecond laser (Coherent Inc., Santa Clara, CA), and external detectors that allow simultaneous detection of emissions of different wavelengths. Enhanced GFP was excited using laser light of 920 nm. Images were obtained using a 20X water-dipping lens. Four-dimensional imaging data was collected by obtaining images from the x-, y-, and z-planes, with a z-thickness of 68 μm and step size of 4 μm to allow for the capture of a complete z-series every 20 seconds for period of 15 minutes. 8 separate image series were taken. Individual cells were tracked using Volocity software (PerkinElmer, Waltham, MA), giving the x-, y-, and z-coordinates of each cell at every time point.
We implement standard Brownian dynamics algorithms for numerical walker models [50]. For each walker, we draw a run time, t, and speed, | v ⃗ |, from distributions using pseudorandom number generators. A direction for the velocity vector v ⃗ is also uniformly randomly chosen. At each time step, we add v ⃗ δ t to the walker position. We use a time step of δt = 0.001 s. When the run time, t, has passed, we randomly draw new run times, speeds, and directions. Diffusive noise is added by adding ( 6 D δ t ) r ^ to the walker position at each time step, where r ^ is a unit vector that points in a random direction. We simulate 5,000 walkers in a 600 μm × 600 μm × 170 μm volume, but only collect data on walkers if they are within a particular 500 μm × 500 μm × 68 μm “imaging” volume. Prior to data collection, we simulate walkers for a short period of time (typically a minute) in order to avoid artifacts at early times.
The 2D moment of inertia tensor, I, is given by [31, 32]:
I = ( I x x I x y I y x I y y ) ,I x x = ∑ i N y i 2 ,I x y = I y x = − ∑ i N x i y i ,I y y = ∑ i N x i 2 .(4) Ixx, Ixy, Iyx, and Iyy are summations of products of displacements, xi and yi. For the average moment of inertia tensor, I ¯, the summation runs over all cellular motions, so that i indexes individual steps and N is the total number of steps. For individual track tensors, In, the summation is over all displacements for the individual cell, so while i still indexes individual displacements, N is now the number of frames for the track. An inertia tensor, In, can be calculated for each track, but only tracks of the same length (e.g., complete tracks) should be averaged together. In order to find the eigenvalues of I, follow standard linear algebra methods [31, 32]. The eigenvalues are found by the solving det ( I − λ 1 ) = 0 so that λ ± = I x x + I y y ± I x x 2 − 2 I x x I y y + I y y 2 + 4 I x y 2, where λ1 = λ+ and λ2 = λ−. Finally, note that for the asphericity calculation, one typically takes the moment of inertia tensor about the center of the track [33]; to do this replace xi and yi in Eq. 4 with x i − x ¯ and y i − y ¯, respectively, where x ¯ = 1 N ∑ i N x i and y ¯ = 1 N ∑ i N y i.
|
10.1371/journal.pntd.0006583 | Identifying residual transmission of lymphatic filariasis after mass drug administration: Comparing school-based versus community-based surveillance - American Samoa, 2016 | Under the Global Programme to Eliminate Lymphatic Filariasis (LF), American Samoa conducted seven rounds of mass drug administration (MDA) from 2000–2006. The World Health Organization recommends systematic post-MDA surveillance using Transmission Assessment Surveys (TAS) for epidemiological assessment of recent LF transmission. We compared the effectiveness of two survey designs for post-MDA surveillance: a school-based survey of children aged 6–7 years, and a community-based survey targeting people aged ≥8 years.
In 2016, we conducted a systematic school-based TAS in all elementary schools (N = 29) and a cluster survey in 28 villages on the two main islands of American Samoa. We collected information on demographics and risk factors for infection using electronic questionnaires, and recorded geo-locations of schools and households. Blood samples were collected to test for circulating filarial antigen (CFA) using the Alere Filariasis Test Strip. For those who tested positive, we prepared slides for microscopic examination of microfilaria and provided treatment. Descriptive statistics were performed for questionnaire variables. Data were weighted and adjusted to account for sampling design and sex for both surveys, and for age in the community survey.
The school-based TAS (n = 1143) identified nine antigen-positive children and found an overall adjusted CFA prevalence of 0.7% (95% CI: 0.3–1.8). Of the nine positive children, we identified one microfilariaemic 7-year-old child. The community-based survey (n = 2507, 711 households) identified 102 antigen-positive people, and estimated an overall adjusted CFA prevalence of 6.2% (95% CI: 4.5–8.6). Adjusted village-level prevalence ranged from 0–47.1%. CFA prevalence increased with age and was higher in males. Of 86 antigen-positive community members from whom slides were prepared, 22 (25.6%) were microfilaraemic. School-based TAS had limited sensitivity (range 0–23.8%) and negative predictive value (range 25–83.3%) but had high specificity (range 83.3–100%) and positive predictive value (range 0–100%) for identifying villages with ongoing transmission.
American Samoa failed the school-based TAS in 2016, and the community-based survey identified higher than expected numbers of antigen-positive people. School-based TAS was logistically simpler and enabled sampling of a larger proportion of the target population, but the results did not provide a good indication of the overall CFA prevalence in older age groups and was not sensitive at identifying foci of ongoing transmission. The community-based survey, although operationally more challenging, identified antigen-positive individuals of all ages, and foci of high antigen prevalence. Both surveys confirmed recrudescence of LF transmission.
| Lymphatic filariasis (LF) is caused by infection with filarial worms that are transmitted by mosquito bites. Globally, 68 million are infected, with ~36 million people disfigured and disabled by complications such as severe swelling of the legs (elephantiasis) or scrotum (hydrocele). The Global Programme to Eliminate LF (GPELF) aims to interrupt disease transmission through mass drug administration (MDA), and to control illness and suffering in affected persons by 2020. The World Health Organization recommends conducting Transmission Assessment Surveys (TAS) in school children aged 6–7 years, to determine if infection rates have dropped to levels where disease transmission is no longer sustainable. American Samoa made significant progress towards eliminating LF. Following seven rounds of MDA, American Samoa passed TAS in 2011–2012 and 2015, with antigen prevalence of <1%. Despite passing TAS, recent studies have provided evidence of ongoing disease transmission in American Samoa, questioning the suitability of TAS for conducting surveillance after MDA has stopped. We compared a school-based survey of children aged 6–7 years and a community-based survey targeting people aged ≥8 years as tools for conducting post-MDA surveillance of LF. Our study provides recommendations for strengthening of post-MDA surveillance as countries approach the GPELF elimination targets.
| Lymphatic filariasis (LF) is a neglected tropical disease caused by Wuchereria bancrofti and Brugia species of helminth worms. The disease is transmitted by mosquito vectors including Aedes, Anopheles, Culex and Mansonia species. Globally, an estimated 68 million people are infected; with 36 million microfilaemic people and 36 million people who are disabled or disfigured because of complications including lymphoedema, elephantiasis and scrotal hydrocoeles [1]. In 2000, the World Health Organization (WHO) launched the Global Programme to Eliminate Lymphatic Filariasis (GPELF), which aims to eliminate LF as a public health problem by 2020. The GPELF includes two strategies, (i) to interrupt transmission of LF by conducting mass drug administration (MDA) in all disease endemic regions, and (ii) morbidity management and disability prevention for infected people [2]. The GPELF is estimated to have delivered 6.2 billion treatments to over 820 million people since its inception [3]. Prior to the formation of the GPELF, the Pacific Programme to Eliminate LF (PacELF) was formed in 1999 to support 22 Pacific Island Countries and Territories in the Western Pacific Region [4]. As of 2017, Cook Islands, Niue, the Marshall Islands, Tonga and Vanuatu have successfully achieved elimination targets established by WHO [5].
WHO recommends conducting Transmission Assessment Surveys (TAS) in children aged 6–7 years for epidemiological assessment of transmission after the completion of MDA [2]. A minimum of two TAS are recommended at 2–3 year intervals, until the absence of transmission can be validated. The first TAS is designed to be conducted at least six months after the final round of MDA to decide if MDA can be stopped, while subsequent TAS are conducted to establish the absence of ongoing transmission. The rationale for choosing children aged 6–7 years as the target population for TAS is because they were born during or after MDA, and any infection in this population would most likely indicate recent and/or ongoing transmission. Transmission is considered not sustainable when the mean antigen (Ag) prevalence in an evaluation unit drops below the TAS threshold. Critical cut-off values are thresholds below which transmission is considered not sustainable and depend on the filarial parasite and vector. Critical cut-off values are calculated so that the likelihood of an evaluation unit passing is at least 75% if true Ag prevalence is 0.5%, and no more than 5% if the true Ag prevalence is ≥1% [2]. In regions with endemic W. bancrofti and where transmission is dominated by Aedes spp. mosquitoes, the TAS threshold is based on an Ag prevalence of 1%. Recent studies have highlighted the limitations of relying solely on Ag-based TAS of young children as a post-MDA surveillance tool, especially as prevalence reaches low levels, and detection of any residual transmission becomes increasingly challenging. For example, in Sri Lanka, TAS of children aged 6–7 years were less sensitive at detecting low-level transmission compared to community-based surveys of people aged ≥10 years, antibody detection in school children aged 6–7 years, or xenomonitoring [6].
In American Samoa, where LF is endemic, W. bancrofti is the only known species of filarial worm, transmitted by both day and night biting Aedes spp. mosquitos. Ae. polynesiensis, the dominant vector, is highly efficient and day-biting [7]. In 1999, the Ag prevalence using rapid immunochromatographic test (ICT) was estimated to be 16.5% [8, 9]. Under PacELF, the American Samoa Department of Health delivered seven rounds of MDA during 2000–2006. In 2007, Ag prevalence by ICT in a community-based survey had declined to 2.3% with microfilaria prevalence of 0.5% [8, 9]. Another round of MDA was recommended by the WHO Western Pacific Region Technical Advisory group Meeting held in 2008 [10], but was not conducted at large-scale due to logistical reasons [8, 11].
School-based TAS are recommended in regions (e.g. American Samoa) where net school enrolment is ≥75% [2]. The sample size and threshold for TAS are designed to estimate Ag prevalence for the entire evaluation unit. Thus, TAS may not be able to detect small and highly focal residual clusters of transmission, particularly if there is significant spatial variation in prevalence within an evaluation unit. In addition, the age group (6–7 years) tested in TAS is likely to have lower prevalence than older ages, making it even more difficult to detect residual foci.
In American Samoa, TAS-1 (conducted in 2011–2012) identified two Ag-positive children, and TAS-2 (conducted in 2015) identified one Ag-positive child. The Ag-positive children identified during both TAS all attended the same school. As the number of Ag-positive children identified was below the critical cut-off of six, American Samoa passed both TAS-1 [12] and TAS-2 [13].
Despite passing two TAS, recent community-based seroprevalence studies and molecular xenomonitoring studies of mosquitoes provided evidence of low-level but widespread ongoing transmission [14, 15]. In a retrospective study of serum samples collected from adults in 2010, Ag (Og4C3) positive samples were identified from participants living across the main island of Tutuila, with higher Ag prevalence in two localised areas. One of these areas included the school where the Ag-positive children were identified during TAS-1 and TAS-2 [14]. A subsequent study in 2014 confirmed ongoing transmission within the two localised areas by identifying high Ag prevalence and microfilaraemic individuals. In addition, Ag prevalence (ICT) of 1.1% (95% CI 0.2–3.1) was found in children aged 7–13 years who attended the school where Ag-positive children were identified in TAS-1 and TAS-2 [14, 16].
The above findings raise concerns about the suitability of school-based TAS for post-MDA surveillance, not only in American Samoa but globally. In 2016, we conducted a study to compare the effectiveness of two survey designs for post-MDA surveillance: a school-based TAS of children aged 6–7 years and a community-based survey of individuals aged ≥8 years. We also evaluated the use of school-based TAS results as indicators of community-level Ag prevalence and/or foci of ongoing transmission. American Samoa was an optimal study site for answering these operational research questions for two reasons: (i) there had been no MDA since 2007 and any infection in children aged ≤9 years would have been acquired after the last round of MDA, and (ii) evidence from recent studies were highly suggestive of ongoing LF transmission [14–16]. In this paper, we report our key findings and discuss their implications for strengthening of TAS for post-MDA surveillance.
American Samoa is a US Territory in the South Pacific (14.2710° South, 170.1322° West), consisting of small inhabited islands with a total population of ~55,519 persons living in ~70 villages [17]. Over 90% of the population resides on the main island of Tutuila and the adjacent island of Aunu'u. The remote Manu'a islands were not included in this study as recent seroprevalence studies did not provide any evidence of local LF transmission [14].
This study consisted of two components: A) a school-based survey and B) a community-based survey. Each of the survey designs and sampling methods are described below.
For the school-based survey, information sheets and consent forms were distributed to parents/guardians of all Grade 1 and 2 children approximately one week prior to scheduled school visits. All children with valid consent forms were included, and assent was sought from all participants. For the community-based survey, signed informed consent was obtained from adult participants or from parents/ guardians of those aged <18 years, along with verbal assent from minors.
Ethics approvals for the study were granted by American Samoa Institutional Review Board and the Human Research Ethics Committee at the Australian National University (protocol number 2016/482). The study was conducted in collaboration with the American Samoa Department of Health and the American Samoa Community College. Official permissions for school and village visits were granted by the Department of Education and the Department of Samoa Affairs, respectively. All field activities were carried out in a culturally appropriate and sensitive manner with bilingual local field teams, and with verbal approval sought from village chiefs/ mayors prior to conducting the community surveys. Surveys were conducted in English or Samoan depending on the participants’ preference. The Institutional Review Board of the U.S. Centers for Disease Control and Prevention (CDC) determined CDC to be a non-engaged research partner.
We recruited a total of 3650 participants, including 1143 and 2507 persons from the school-based and community-based surveys respectively (Table 1).
We included all elementary schools (N = 29) on the two main islands of Tutuila and Aunu’u. Of 2180 Grade 1 and 2 students who were eligible to participate, 1143 (52.4%) students returned signed consent forms and were included in the study. The average participation rate by school was 57% (range 18.2–91.7%).
Table 2 summarises characteristics of participants included in the school-based TAS. Of the 1143 students, we identified nine Ag-positive children, equivalent to a crude CFA prevalence of 0.8%. As the critical cut-off for passing TAS was fewer than six Ag-positive children, American Samoa failed the school-based TAS.
The estimated overall CFA prevalence after adjusting for participation rates by school and sex was 0.7% (95% CI: 0.3–1.8). The design effect for the school-based survey was 1.9. Adjusted CFA prevalence in males was 0.5% (95% CI: 0.1–1.9) and in females was 0.9% (95% CI: 0.4–2.4).
Valid FTS results were available for all 1143 (100%) children. Of the nine Ag-positive children, four (44.4%) attended the same school in Pago Pago, and two (22.2%) attended the same school in Nua. The other three Ag-positive children attended different schools located in the villages of Ili'ili, Nu'uuli and Faga'alu. Both Ag-positive children from the school in Nua lived in Fagali’i, one of the suspected hotspots identified in previous studies [14, 16]. Seven (77.8%) Ag-positive children were born in American Samoa and reported to have lived there for their entire lives, and two (22.2%) were born in Samoa.
Of the nine FTS-positive children, one (11.1%) was microfilaraemic with Mf density of 1075 Mf/ml. The child was a 7 year old male who lived in Vaitogi, another potential hotspot identified in previous studies and attended the school where Ag-positive children were identified in TAS-1 and TAS-2.
We visited all 30 selected PSUs and sampled 2507 persons from 711 households. The villages of Ili’ili and Pava’ia’i were split into two segments during the selection process, and both segments of these villages were selected for the survey. For the purposes of analyses, data from the different segments of each village were pooled, and results were presented for 28 villages (comprising 30 PSUs).
The average household size was 6 (range 1–25) persons aged ≥8 years. We recruited participants from 77.6% of the selected households, and 83.2% (range 14.3–100%) of eligible household members (aged ≥8 years). On average, 16.8% of eligible household members were not recruited, and non-response was mostly associated with household members being absent at the time of visit, rather than refusal to participate. We recruited 1,140 (45.5%) males and 1,367 (54.5%) females (Table 3). Of the 2507 participants tested, 11 (0.4%) had invalid tests and were excluded from analyses (Tables 1 & 3). Of the 2496 participants with valid tests, 102 were Ag-positive, equivalent to an overall crude CFA prevalence of 4.1%. Of the 102 FTS-positive persons, 79 were male (crude CFA prevalence 7.0%) and 23 were female (crude CFA prevalence 1.7%), and this difference was statistically significant (p<0.001).
The original target sample size for the community-based survey, calculated based on an expected CFA prevalence of 1%, was 4620. After the first two weeks of recruitment, the observed CFA prevalence (~4%) was significantly higher than anticipated, and it was determined that a smaller target sample size of 2981 would provide adequate statistical power (Table 4).
The age and sex distribution of the community-based survey participants and the general population of American Samoa are presented in Fig 3, showing that the survey included proportionately more females but provided a good representation of all age groups. After adjusting for the survey design, and age and sex distribution of American Samoa, the overall adjusted CFA prevalence was 6.2% (95% CI: 4.5–8.6). The design effect for the community-based survey was 4.2. The adjusted CFA prevalence by age and sex in the selected villages are presented in Fig 4A. Notably, in children aged 8–9 years, who were born after MDA had stopped, the adjusted CFA prevalence was 2.2% (95% CI: 0.8–6.1).
Of the 102 FTS-positive individuals, we were able to prepare slides for 86 (84.3%) participants. Of these, 22 (25.6%) were microfilaraemic, of whom 19 (86.4%) were male (Fig 4B). The geometric mean Mf density was 60.7 Mf/ml (range 5.6–916.7 Mf/ml).
Adjusted village-level CFA prevalence varied from 0% to 47.1% (Table 4 and Fig 5). Of the 28 villages, Ag-positive individuals were identified in the majority of villages, and only 6 (21.5%) villages had no Ag-positive individuals. Microfilaraemic people identified from the community survey were dispersed throughout the island and lived in 10 of the 28 selected villages. Of the 22 microfilaraemic people, six (27.3%) lived in Vaitogi (focal area with ongoing transmission identified in previous studies), the same village as the Mf-positive school child.
Ag-positive school children from the school-based survey lived in four (14.3%) of the 28 villages included in the community-based survey and attended school in three villages (10.7%). A total of six villages (21.4%) were identified as communities where Ag-positive children lived and/or attended school. Three (33.3%) and two (22.2%) Ag-positive children identified in the school-based survey, respectively, lived in and attended school in villages that were not selected for this survey; these data were therefore not included in the analysis of sensitivity, specificity, PPV and NPV. Table 5 shows the sensitivity, specificity, PPV and NPV for using Ag-positive children as indicators of villages with adjusted CFA prevalence of greater than 1%, 2%, 5%, 10%, and 20%. These findings suggest that follow-up of Ag-positive school children would have identified areas of transmission with high specificity (>73% for all scenarios), but low sensitivity (<24% for all scenarios), even if we considered villages where Ag-positive lived and/or attended school. Ag-positive children were a poor indicator of villages with higher prevalence (10 or 20%), with PPV of 25% or less for any of the scenarios tested.
Our study confirmed recrudescence of LF transmission in American Samoa 10 years after the last round of MDA. Through the school survey, we identified nine Ag-positive children (adjusted CFA prevalence 0.7%), including a seven-year-old microfilaraemic child. The community survey identified 102 Ag-positive persons (adjusted CFA prevalence 6.2%) and 22 microfilaraemic individuals. American Samoa failed school-based TAS in 2016, and the community-based survey identified higher numbers of Ag-positive people compared to Ag prevalence of 2.3% (by ICT) in 2007 and 3.2% (by Og4C3 Ag) in 2010 [8, 14]. The adjusted CFA prevalence in the school-based survey of children aged 6–7 years (0.7%) was significantly lower than the community-based survey of people aged ≥8 years (6.2%), consistent with our knowledge that CFA prevalence is generally higher in older age groups.
Our study identified advantages and limitations of using school-based versus community-based surveys. The school-based survey was logistically simpler, faster and cheaper, while the community-based survey was practically more challenging and time-consuming but provided more detailed information on estimates of community-level CFA prevalence, highlighted foci of high prevalence, and identified Ag- and Mf-positive people who are capable of perpetuating transmission. With the school-based survey, the identification of Ag-positive young children who have lived in American Samoa for their whole lives provided clear evidence of transmission within the past 6–7 years. The community survey also provided evidence of recent transmission by identifying Ag-positive children aged 8–9 years (adjusted CFA prevalence of 2.2%, 95% CI 0.8–6.1), who were either born after the last round of MDA in 2008 or were too young to participate. In addition, the community-based survey provided detailed epidemiological information for a wider age range, including identification of older Ag-positive people who may serve an important reservoir of parasites and maintain transmission in the post-MDA setting [28]. Our results indicate that TAS conducted among young children (who have lower seroprevalence) may not be sufficiently sensitive to identify all areas of residual transmission (i.e where Ag-positive people aged ≥8 years were identified). This finding supports the conclusions of a recent study which modelled the efficiency of surveillance protocols based on the combination of ability to identify transmission foci, sample size required and high PPV, and found that testing of adults would be more efficient at detecting transmission in low prevalence settings compared to testing children aged 6–7 years [28].
The school-based TAS did not identify any difference in CFA prevalence between male and female young children. The community-based survey indicated that males (particularly those aged ≥20 years) had higher CFA prevalence and greater proportion of those who were Ag-positive had detectable microfilaria. Higher prevalence in adult males could be related to more time spent outdoors for work and recreation, and/or lower rates of participation in MDA [14, 29]. Hormonal and pregnancy-mediated regulation of the immune system may also contribute to lower infection rates in females, particularly during the reproductive years [30].
The school-based survey was a systematic survey where all elementary schools on the two main islands were included. Overall, 52.4% of our target population (6–7 year olds) participated in the survey, and we do not have reasons to suspect differences in Ag prevalence amongst those children who did not participate. The school-based survey identified two FTS-positive children living in Fagali’i, an area of known high LF transmission [16], and a cluster of four FTS-positive children who attended one school and lived in Fagatogo and Pago Pago. Considering that both villages of residence had low estimated CFA prevalence of 2.7% and 2.3%, respectively (below the overall estimated CFA prevalence of 6.2%), the school-based clustering raises the possibility that transmission might be occurring in and around the school, particularly in the presence of a highly efficient day-biting vector [7]. The school-based survey had limited utility for detecting focal areas of ongoing transmission, (Table 5); while follow-up of villages where Ag-positive children lived or went to school might help identify areas of transmission with high specificity (range 83.3–100%), the low sensitivity (range 0–23.8%), PPV (0% in villages with >20% Ag prevalence) and NPV (range 25–83.3%) suggests that even if further surveillance was conducted in all the villages where Ag-positive children lived and/or attended school, many high prevalence villages would still have been missed.
The community-based survey was a modified WHO cluster survey, which is recommended for surveying large populations in resource-limited settings. By using a population representative sampling design and correcting for clustering during analyses [27, 31], we believe our results are an accurate estimate of the country-wide CFA prevalence. The community-based survey demonstrated significant heterogeneity in CFA prevalence between villages, even in the very small and isolated island. Similar observations were made in the 2007 survey in American Samoa, and during post-MDA surveillance studies in other small island countries such as Sri Lanka and Samoa [6, 8, 32]. A limitation of the survey design was that the school and community-based surveys were not completely geographically aligned because all schools were included, but only 30 of the 70 PSUs were sampled, i.e. some children tested in the school-based survey lived in villages that were not selected for the community-based survey. However, as we surveyed a large proportion of the selected villages, and many villages are contiguous along the limited number of roads in American Samoa (Fig 5), geographical concordance between the two surveys is unlikely to be a major issue in this study.
The reasons for recrudescence of LF in American Samoa are unclear but could be associated with a combination of factors including some areas of poor-coverage or systematic non-compliance during MDAs [33, 34], travel and migration of people from other countries in the Pacific where LF transmission is still ongoing [14, 29, 34]. An outdoor lifestyle in the presence of highly efficient day and night biting mosquitoes could also have contributed to recrudescence, and lower target thresholds may need to be considered in such settings [7, 15].
LF has a long prepatent period [35], leading to low prevalence in young children even when prevalence is high in adults [8]. Thus, it was unlikely that all the Ag-positive people of all ages identified in our study acquired infection between 2015 (when American Samoa passed TAS-2) and 2016 (when TAS-3 was failed). In hindsight, early signals of ongoing transmission were evident from the population-representative serological survey of adults conducted in 2010, and further studies in 2014 which confirmed high Ag prevalence and identified Mf positive individuals within two foci of residual transmission in American Samoa [14, 16]. Ag-positive young children detected in TAS-1 and TAS-2 attended school in one of the foci identified by research studies, which could have provided early signals of focal transmission. A molecular xenomonitoring study conducted in 2011 also identified evidence of widespread low-grade infection [7], and findings were strongly geographically correlated with village-level human seroprevalence [15]. Taken together, these observations and the results of the current study indicate that, compared to current protocols and thresholds of TAS of 6–7 year old children, other surveillance strategies have the potential to detect ongoing transmission earlier and with greater geographic precision. Our findings therefore support the need for more sensitive and innovative post-MDA surveillance strategies to achieve elimination goals of the global programme.
Surveillance strategies that warrant further consideration include community-based surveys of both adults and children, school-based surveys that include a wider age range, lowering the threshold for the current TAS protocol, spatially explicit surveillance strategies, adaptive or snowball sampling (e.g. testing household and/or community members of Ag-positive children), testing for antifilarial antibodies, molecular xenomonitoring, or a combination of these strategies. We do not currently have sufficient evidence to specifically recommend any of these strategies over another, and operational research will be required to provide robust guidance for future surveillance. However, inclusion of older individuals in surveillance strategies might enable earlier detection of Ag-positive people in American Samoa and in other Polynesian countries, with highly efficient day-biting vectors and similar age-specific prevalence curves.
As countries approach the GPELF elimination targets, WHO recommends developing sustainable post-MDA surveillance strategies that are cost-efficient and can be integrated into routine surveillance activities [2, 16]. Although community-based surveys can be operationally more challenging, surveillance activities could take advantage of opportunistic and cost-effective methods of targeting community members [16, 28], such as testing high-risk occupation groups, screening at workplace clinics and antenatal clinics, or during routine health check-ups for chronic illnesses and school-based vaccination campaigns.
|
10.1371/journal.ppat.1006936 | Nubbin isoform antagonism governs Drosophila intestinal immune homeostasis | Gut immunity is regulated by intricate and dynamic mechanisms to ensure homeostasis despite a constantly changing microbial environment. Several regulatory factors have been described to participate in feedback responses to prevent aberrant immune activity. Little is, however, known about how transcriptional programs are directly tuned to efficiently adapt host gut tissues to the current microbiome. Here we show that the POU/Oct gene nubbin (nub) encodes two transcription factor isoforms, Nub-PB and Nub-PD, which antagonistically regulate immune gene expression in Drosophila. Global transcriptional profiling of adult flies overexpressing Nub-PB in immunocompetent tissues revealed that this form is a strong transcriptional activator of a large set of immune genes. Further genetic analyses showed that Nub-PB is sufficient to drive expression both independently and in conjunction with nuclear factor kappa B (NF-κB), JNK and JAK/STAT pathways. Similar overexpression of Nub-PD did, conversely, repress expression of the same targets. Strikingly, isoform co-overexpression normalized immune gene transcription, suggesting antagonistic activities. RNAi-mediated knockdown of individual nub transcripts in enterocytes confirmed antagonistic regulation by the two isoforms and that both are necessary for normal immune gene transcription in the midgut. Furthermore, enterocyte-specific Nub-PB expression levels had a strong impact on gut bacterial load as well as host lifespan. Overexpression of Nub-PB enhanced bacterial clearance of ingested Erwinia carotovora carotovora 15. Nevertheless, flies quickly succumbed to the infection, suggesting a deleterious immune response. In line with this, prolonged overexpression promoted a proinflammatory signature in the gut with induction of JNK and JAK/STAT pathways, increased apoptosis and stem cell proliferation. These findings highlight a novel regulatory mechanism of host-microbe interactions mediated by antagonistic transcription factor isoforms.
| The numerous human diseases caused by aberrations in intestinal immunity and integrity urge a better understanding of the regulatory interactions that balance the output of host-microbe interactions. In this study, we discovered a novel phenomenon of transcriptional antagonism exerted via two isoforms encoded from the same gene. Balanced expression of the two forms was necessary for ensuring normal immune gene expression and maintaining immunological homeostasis in the Drosophila gut. We performed genetic manipulations to skew the balance of these isoforms. This resulted in a dysregulated immune system, changed levels of gut bacteria and altered host lifespan. Moreover, when we overexpressed the activating form flies quickly succumbed to oral bacterial infection despite an enhanced immune response. We suggest that antagonistically acting transcription factor isoforms may constitute a general mechanism for adjusting gene expression in various biological processes.
| The innate immune system of mammals and insects is regulated by intracellular signaling pathways and transcriptional programs that show remarkable signs of evolutionary conservation. Well-known examples are the Toll/Toll-like receptor (TLR), immune deficiency (IMD)/tumor necrosis factor-α (TNF-α), JAK/STAT and JNK signaling pathways and their respective downstream transcriptional activators, nuclear factor kappa B (NF-κB), STAT and AP-1, which regulate innate immune responses in both Drosophila and mammals [1–4]. Pathway activation triggers a vast set of genes that encode effector molecules such as antimicrobial peptides (AMPs) and cytokines, which in mammals also support the induction of adaptive immune responses [5]. The underlying regulation is complex, especially in the intestine and other barrier epithelia that are in constant contact with the commensal microbiota. Improper control of the innate immune system and loss of tissue homeostasis can cause inflammation and other autoimmune diseases, and may lead to system failure and early death [1, 6]. In Drosophila, the IMD and Toll pathways and downstream NF-κB homologs, Relish (Rel) [7] and Dorsal-related immunity factor (Dif) [8], are crucial activators of immune genes in response to infection. About 25% of Drosophila immune-regulated genes (DIRGs) are, however, expressed independently of these pathways [9]. Additional transcriptional activators [10–13] as well as repressors [14–17] have been implicated in the immune response during specific conditions. However, little is known concerning how such factors compete for the same targets and interact to balance responses and maintain homeostasis.
The POU/Oct transcription factor family is a subclass of the homeodomain proteins present in all metazoans [18] and encompasses crucial regulators of developmental decisions, metabolism, immunity and cancer [19]. Human Oct-1 (POU2F1) was originally discovered as a regulator of adaptive immune responses [20] and has more recently been shown to regulate numerous target genes involved in both innate and adaptive immunity, stress resistance, metabolism and cellular proliferation [21]. In a yeast screen for transcriptional regulators of innate immune response genes, we have previously isolated three candidates from the Drosophila POU/Oct family [22]. The Oct-1 homolog, nub, was subsequently found to encode a negative regulator (Nub-PD) of NF-κB/Rel-driven immune gene expression in flies [15]. Nub-PD is necessary to control aberrant gene activation as nub1 mutant flies are short-lived and have a severely distorted gut microbiota [23]. Both Oct-1 and Nub regulate target genes by binding to the canonical octamer DNA sequence motif (ATGCAAAT) or variants thereof, via their C-terminal POUS and POUH domains [24–26]. Thus, the DNA-protein interaction surfaces appear conserved, further emphasizing the evolutionary relationship of these ancient transcriptional regulators.
It has earlier been reported that alternative transcript forms of nub exist [27]. However, functional studies have, to our knowledge, been focused solely on the Nub-PD isoform. In this study, we demonstrate that nub encodes a novel isoform, Nub-PB, which is a strong activator of immune gene expression. Furthermore, both isoforms are expressed in midgut enterocytes and regulate the same immune target genes antagonistically. We show that such tuning of the transcriptional output of Nub target genes is crucial for host immunity, fitness and survival.
Recent public annotations indicate two promoters of the nub gene, which in combination with promoter-specific splicing produce two independent proteins, Nub-PB (104 kDa) and Nub-PD (65 kDa; Fig 1A). Both proteins contain identical C-terminal domains, in which the DNA-binding POUS and homeodomain (POUH) are located. Thus, Nub-PB and Nub-PD are expected to bind the same DNA sequences, while abilities for protein-protein interactions are likely to differ due to the distinct N-termini of the proteins. Initial experiments were aimed at detecting and elucidating any immune-regulatory features of Nub-PB. We applied the Gal4-UAS system and constructed a UAS-nub-RB line to drive its overexpression in immune competent tissues (midgut and fat body) using the c564-Gal4 driver. To simultaneously detect the ability of Nub-PB to regulate AMP gene expression, we combined c564>nub-RB with a lacZ reporter construct for Cecropin A1 (CecA1). This resulted in very prominent β-galactosidase activity in the abdominal fat body (Fig 1B). Nub-PB hence appears to function as a transcriptional activator of immune effector genes, in stark contrast to Nub-PD.
We carried on by analyzing the global transcriptional profiles of c564>nub-RB in comparison to driver controls (c564>+, from hereon referred to as wild type [wt]). To enable expression analysis of fat body and gut independently, mRNA from the fly body without head and gut (“rest”) and the digestive system (“gut”) were extracted and analyzed separately. The raw data were normalized, preprocessed and filtered to remove genes that were not expressed above the detection level (S1 Table; Methods). Unbiased principal component analysis (PCA), correlation analyses and hierarchical clustering showed that the tissue variable (“rest” versus “gut”) accounted for the largest distinction, as expected (PC1, explains 47% of the variance; Figs 1C and S1A). The PCA furthermore indicated that genotype accounted for the second largest separation in the dataset (PC2 explains 9.5%; most apparent in “gut”). After adjusting p-values for a false discovery rate of 1%, 1177 (rest) and 545 (gut) probes indicated significant differential expression (S1 Table), of which 132 were in common for both tissues (S1B Fig, S2 Table). Gene set enrichment analyses (GSEA) according to “Biological process” of upregulated transcripts in respective tissue resulted in a single major GO cluster encompassing various aspects of immunity (Fig 1D and 1E, S3 Table). Of transcripts that were found downregulated, the only significant subnode in “rest” was “proteolysis” (also significant for upregulated probes in “rest”), whereas those from gut samples formed GSEA clusters of several overlapping cellular and developmental processes such as cell fate commitment, organ development, Notch signaling pathway and sensory responses along with GO categories of antibacterial defense and transcriptional regulation, suggesting a multifaceted role of Nub-PB in this organ (S2 Fig, S3 and S4 Tables). To detect additional DIRGs, the dataset was manually curated and run against a collection of putative immune genes (https://lemaitrelab.epfl.ch/resources, accessed February 2015), which yielded 152 DIRGs in total (Fig 1F, S5 Table). Overall, a striking coherence with well-characterized DIRGs typical for activated Toll, IMD and JAK/STAT pathways was observed. In summary, the global expression analysis demonstrated a broad immune activation following nub-RB overexpression, which further indicates that Nub-PB is a transcriptional activator of immune genes.
A comparative analysis of the differentially expressed genes in response to Nub-PB overexpression (this study) and nub1 mutant flies (disrupts Nub-PD, but not Nub-PB expression) [15, 27] revealed an extensive overlap as 65 immune-regulated genes were upregulated in both transcriptional profiles (S6 Table). Out of the most highly expressed immune-process genes in “rest” (FC>10, 26 genes), 25 were upregulated at least 2-fold in nub1 mutants. To compare the capacity of Nub-PB and Nub-PD to regulate AMP genes, c564-driven overexpression of either transcriptional regulator was performed in parallel experiments (Fig 2A). In validation of the transcriptional profiling, overexpression of nub-RB promoted strong upregulation of all eight assayed AMP genes (AttA, AttB, CecA1, CecC, DptA, Drs, Drsl2 and Drsl3) from whole fly extracts. Conversely, overexpression of nub-RD decreased CecA1, DptA, Drsl2 and Drsl3 significantly compared to controls. We next evaluated the combined effect of the two isoforms in co-overexpression assays. To circumvent developmental and secondary effects, a temperature sensitive (ts) c564-Gal4; Tub-Gal80 driver line was applied. This resulted in expression levels of AttA, CecA1, CecC and DptA similar to those in the control cohort (Fig 2B), indicating that the two Nub isoforms overall neutralize each other’s activity. Co-expression did not dampen the increased expression of Drsl3, whereas Drsl2 was partially restored to control levels, suggesting that Nub-PB induces these targets with greater affinity than the corresponding repression mediated by Nub-PD. We conclude that the Nub isoforms are able to regulate the expression of several AMP genes in an antagonistic manner.
It has previously been shown that Nub-PD is a repressor of Rel-target genes in the absence of infection [15]. The observation that Nub-PB overexpression activates the same gene set (S6 Table) prompted us to explore its putative cooperation with Rel. Co-transfection of mbn-2 cells with nub-RB and Rel induced CecA1-luciferase expression significantly stronger (42.8-fold) than single transfections (nub-RB, 18.6-fold; Rel, 2.2-fold), which also indicated a synergistic effect (Fig 2C). We therefore hypothesized that Nub-PB acts as a co-activator of Rel. To investigate this, AMP expression was assayed in whole flies following overexpression of nub-RB in a Rel mutant (RelE20) or wild type background (Fig 2D–2F). To induce a robust IMD pathway response, separate cohorts were subjected to systemic infection with Enterobacter cloacae β12. Three AMP genes were assayed based on their known dependency of Rel and proximal Nub/Oct-sites: DptA (Rel-dependent, one Nub/Oct-site), CecA1 (Rel-dependent, several Nub/Oct-sites) [15, 28] and Drsl2 (Rel-independent, one Nub/Oct-site) [29]. Of note, binding of Nub to the proximal promoter region of DptA and CecA1 has been demonstrated biochemically [15]. Overexpression of Nub-PB was sufficient to drive expression of CecA1, but not DptA, in absence of Rel and independent of infection status (Fig 2D and 2E). Drsl2, specifically expressed in the gut via the JAK/STAT pathway [29] was, as expected, not affected by the RelE20 mutant background (Fig 2F). Taken together, the data suggest that Nub-PB can influence AMP gene transcription both independently of, and together with Rel.
The proximal promoter region of CecA1 contains a cluster of six Oct/Oct-like sites required for Nub-PD mediated repression of a CecA1-luciferase reporter construct in vitro [15]. To investigate the requirement of this cluster for the regulatory capacity of Nub-PB and Nub-PD in vivo, the expression of different reporter constructs was analyzed in female (Fig 3) and male flies (S3 Fig). As expected, flies carrying the CecA1-lacZ construct with an Oct cluster deletion (pA10ΔOct; Fig 3A) displayed a derepressed and hence stronger reporter gene expression in fat body than flies with the complete upstream region (Figs 3B, 3C, 3H and S3A and S3B). Since deletion of the cluster promoted reporter gene expression per se, the incubation time was decreased to circumvent saturation of the response (Figs 3D–3G and S3C–S3F) and combined with c564-driven overexpression of Nub-PB (Figs 3F, 3G and S3E and S3F). The full-length pA10 CecA1-lacZ reporter gene responded strongly to Nub-PB overexpression (Fig 3F and 3H; also shown in Fig 1B). Overexpression combined with the Oct cluster deletion did, however, not promote full transcriptional activation (Fig 3G and 3H). Still, an intermediate level of β-gal reporter gene expression was observed, significantly stronger than in the Oct deletion control (Fig 3E and 3H), suggesting that the CecA1 locus may contain additional Nub target sequences outside the pA10ΔOct region. We conclude that both Nub isoforms require the Oct cluster for accurate regulation of the CecA1 gene (Fig 3I).
Transcription factor antagonism requires a co-localized expression. We explored the spatial distribution in vivo of nub-RB and RD transcripts from dissected tissues of adult male flies (Fig 4A). Notably, the RNA levels of nub-RD exceeded that of nub-RB in the gut, whereas expression of the two appeared roughly similar in other tissues. To further investigate the spatial expression, we applied transgenic reporter lines specific for either the nub-RB or nub-RD transcript. A dual isoform marker line (nub-RBGFP, nub-RDAC-62>mCherry) was applied to visualize the spatial expression of both isoforms in parallel (S4 Fig). Prominent dual fluorescence was observed in larval wing discs, leg discs and foregut, with varying degree of overlap, whereas only nub-RD>mCherry was expressed in the larval brain (S4A–S4C Fig). For adult expression, tissues from nub-RBGFP and nub-RDAC-62>mCherry flies were studied separately. Fluorescence was observed in wing veins and leg joints with either line, whereas mainly nub-RD appeared expressed in the abdominal fat body or adjacent tissues (S4D–S4I Fig). A strong GFP signal was observed throughout the midgut of nub-RBGFP flies (Fig 4B) while nub-RDAC-62-driven mCherry expression was present only in the anterior region and faded within days in newly eclosed adult flies (S4A’ Fig). Therefore, additional nub-RD-Gal4 driver lines were screened and especially nub-RDVT6452 drove prominent mCherry expression in the midgut (Figs 1A and 4B). For the purpose of the present work, we conclude that the isoforms are expressed to varying degrees in immune competent tissues and thereby should be able to act as transcriptional antagonists in a competitive manner.
In contrast to the observed isoform transcript levels in Fig 4A, Western blot experiments on midgut and carcass extracts using a non-discriminatory Nub antibody yielded an overall stronger band for Nub-PB (S5A Fig) [15]. This suggests that additional regulation occurs post-transcriptionally. Since nub-RD mutant flies exhibit chronic immune activation and microbial dysplasia in the gut [15, 23], we hypothesized that this isoform is required to suppress aberrant activity of Nub-PB in these tissues. In fact, Nub immunostaining is commonly used to mark the nucleus of gut enterocytes [30], which suggests that at least one of the isoforms is strongly expressed in this cell type. The enterocyte-specific driver NP1-Gal4ts was therefore applied to drive RNA interference (RNAi) of nub-RB (nub-RB-IR) in adult flies. This resulted in a roughly two-fold decrease of the targeted transcript accompanied by a significant reduction of Drsl2, Drsl3, DptA, CecA1 and upd3 in midgut tissues (Fig 4C). The latter encodes an infection-inducible cytokine and ligand of the JAK/STAT pathway [31, 32] and was found upregulated 4.8-fold in the gut, albeit below the signal threshold in the global transcriptional profile (S1 Table). The RNAi was further confirmed at the protein level, as Nub-PB decreased 2-fold in extracts from whole flies, strongly indicating that enterocytes constitute a major source of this isoform (S5B Fig). Similar effects were observed with c564-Gal4ts in the midgut, but not in the abdominal carcass (encompassing the fat body), suggesting that the endogenous Nub-PB acts as a positive regulator of AMPs specifically in the midgut (S5C and S5D Fig). Conversely, RNAi-mediated downregulation of nub-RD, using a similar assay as in Fig 4C, resulted in an overall increased expression of the same set of genes (Fig 4D). Comparative overexpression of the two isoforms in gut enterocytes confirmed their opposite regulatory effects on immune genes, previously observed in whole flies (Fig 2A and 2B), as Drsl2 and upd3 were strongly up- and downregulated in midgut extracts following nub-RB and nub-RD overexpression, respectively (Fig 4E). CecA1 and Drsl3 levels were increased by nub-RB but unaltered by nub-RD whereas neither overexpression affected DptA significantly. Interestingly, NP1ts>nub-RD decreased the expression of nub-RB 2.4-fold, but not vice versa. To investigate whether Nub-PB expression affected the gut bacterial load, we performed a qPCR analysis of the relative 16S rDNA levels from midgut extracts. NP1ts-driven overexpression of nub-RB resulted in a ~56% reduction of bacteria after 24 h (Fig 4F). Conversely, RNAi using the same driver increased bacterial loads by 67% (Fig 4G). The same set of flies displayed striking lifespan phenotypes as nub-RB overexpression significantly shortened median longevity (females, 19.3; males, 20.3 days; average from three individual experiments) relative to controls (females, 46.2; males 45.8 days), whereas its downregulation increased longevity (females, 52.3; males, 48 days) (Figs 4H, 4I and S6A–S6C). Overexpression of nub-RD resulted in longer lifespan (females, 29; males 33 days) compared to that of nub-RB, but shorter than controls (S6B and S6C Fig). Surprisingly, antibiotic supplementation of the diet enhanced longevity of nub-RB-overexpressing flies (females 15.8%, males 28.9%) but not those of nub-RB RNAi (females -3.1%; no change in males), suggesting that the microbial composition, rather than loads influences host lifespan (Figs 4F, 4G and S6B and S6D). The germ-free conditions enhanced female (6.8%), but not male controls. Conversely, germ free nub-RD overexpressing males displayed enhanced longevity at early time points, but were overall not significantly benefited, whereas females were equally long-lived in comparison to conditionally reared flies. Together, these findings demonstrate that Nub isoforms regulate the expression of midgut AMPs in opposite manners and that Nub-PB expression correlates with gut microbiota and host lifespan.
To investigate the role of Nub-PB after bacterial challenge, we performed oral infections using Erwinia carotovora carotovora 15 (Ecc15), a well-characterized bacterium with generally low oral pathogenicity in adult flies. Strikingly, overexpression of nub-RB caused a hypersensitive phenotype as all males and roughly 70% of the females succumbed within a day post infection (Fig 5A and 5B). Interestingly, the median lifespan of the remaining overexpressing female survivors after bacterial exposure was 10 days, suggesting that death occurs primarily due to acute effects. RNAi directed against nub-RB caused a significant female (~25%), but not male, mortality during the acute stage of infection compared to infected driver control flies. We next explored the induction of the immune response mediated by the combined overexpression of nub-RB and Ecc15 infection (Fig 5C–5F). At three hours post infection (hpi), Drsl2 expression was upregulated by three orders of magnitude relative to the similarly infected driver control cohort, while at 24 hpi, levels were more comparable to those of the uninfected overexpression cohort (Fig 5C). This indicates a rapid and transient hyperinduction of this gene by the combined effect of overexpression and infection. The induction of upd3 and CecA1 was also strongly enhanced by the combined effect of overexpression and infection, with peak expressions occurring later than for Drsl2 (Fig 5D and 5E). Conversely, and in line with Fig 4E, DptA was not affected by nub-RB overexpression with NP1 (Fig 5F). We hypothesized that such a strong effect on the immune response would likely be detrimental for the host, while at the same time enhance clearance of the infection. In line with this, enterocyte expression levels of nub-RB correlated with the relative clearance of Ecc15 compared to controls (Fig 5G–5J). In comparison, overexpression of nub-RD caused a moderate, non-significant decrease in clearance (S7A and S7B Fig). To exclude a confounding effect of feeding rate, a capillary feeding assay was performed where the different genotypes were found to consume similar amounts to controls (Figs 5K, 5L and S7C). Proinflammatory responses in the gut could potentially cause epithelial damage and ultimately cause gut leakage. To test this, we applied the Smurf assay [54] but did not observe any flies turning blue, neither from genotype, infection, nor the combination of both, suggesting that death occurs due to other causes (Fig 5M). We also observed upregulated levels of both isoforms in the midgut of orally infected flies, albeit stronger for nub-RB than nub-RD (Fig 5N). This suggests that the balance is temporarily skewed towards the activating function of Nub-PB during the acute stage of infection. Following recovery on regular fly food, the expression of both isoforms returned to pre-infection levels around 48 hpi, indicating a pattern typical for transiently induced DIRGs. Taken together, these data indicate that Nub-PB is involved in the midgut immune response to ingested Ecc15 and that the activity of this transcription factor requires tight control to avoid detrimental effects on the host.
Among the identified IMD/Toll-independent DIRGs in the transcriptional profiles, several targets of the JAK/STAT pathway were induced, such as the gut-specific and infection-inducible Drsl2, stress-regulated Turandots (Tots) and the immunomodulatory cytokine Diedel (in the fat body), suggesting that Nub-PB either acts above, or at the level of, the JAK/STAT pathway (Fig 1F, S1 and S5 Tables) [10, 29]. In Drosophila gut enterocytes, the JNK pathway has been implicated in the regulation of Upd3, which in turn acts as a ligand for the JAK/STAT pathway in response to bacterial infection and stress [31]. Pathway activation triggers intestinal stem cell differentiation and proliferation to replenish extruded enterocytes [31]. Our observations that the expression of upd3 and Drsl2/3 is regulated by Nub-PB led us to investigate the role of the above-mentioned pathways in this context (Fig 6). Prolonged nub-RB overexpression for five days resulted in prominent induction of reporters for JNK (Fig 6A–6D), upd3 (Fig 6E–6F”) and JAK/STAT (Fig 6G and 6H). This was accompanied by a general disorganization of enterocytes (Fig 6E’ and 6F’), increased number of mitotic cells (Fig 6I, 6J and 6M) and apoptosis (Fig 6K and 6L). Combined overexpression of nub-RB and targeted RNAi against the JNK-homolog basket (bsk-IR) in gut enterocytes attenuated the induction of Drsl2 but not upd3, suggesting a dependency on JNK pathway activity for the former, but not the latter target (Fig 6N). To investigate the role of JAK/STAT, nub-RB was co-expressed together with a dominant negative form of the receptor Domeless (DomeDN; Fig 7O–7Q). Similarly to the findings in Fig 6N, and independent on infection status, this diminished the induced expression of Drsl2, but not upd3. Nub-PB might hence act together with the transcription factors of the JNK and JAK/STAT pathways to induce midgut-specific immune genes not typically regulated by NF-κBs. In support of this, Drsl2 was found to contain putative DNA-binding motifs for Nub [15], AP-1 and Stat92E [33] in the proximal promoter region (Fig 2F). As expected, the expression of nub-RB was similar between the overexpression genotypes and was, in line with observations in Fig 6H, also induced by Ecc15 infection in the driver control line (Fig 6N and 6Q). Together these results indicate that Nub-PB is sufficient to drive most, if not all, of the documented aspects of intestinal immunity and inflammation.
We have shown that the large isoform encoded by nub, Nub-PB, is a novel and exceptionally strong transcriptional activator of immune genes in Drosophila. Compared to the major immune regulatory factors, Rel and Dif, Nub-PB can potentially target an even broader set of DIRGs (Fig 1F). This could be explained by the notoriously promiscuous nature of Oct factors in terms of their conformations, dimer formations, protein-protein interactions and post translational modifications [21]. In humans, the Nub homolog, Oct-1, has been proposed to act as a switchable stabilizer of either repressed, induced or poised states of genes depending on its protein-protein interactions [34]. Furthermore, interactions between human Oct-1 and NF-κB have been demonstrated biochemically [35]. In agreement, we found that Nub-PB and Rel synergistically induce CecA1 transcription (Fig 2C and 2E). Moreover, the Nub binding sites (Oct sites), located in immediate proximity to the κB-sites in the proximal promoter region of CecA1, were required to both repress [15] and fully induce expression (Figs 3 and S3), suggesting that the isoforms bind the same motifs to antagonistically regulate the transcription of Rel-target genes (Fig 3I).
Several Oct family members encode alternative isoforms [36]. Knowledge of their unique functions is, however, sparse. Drosophila appears to be no exception as both nub and its paralogs pdm2/miti and pdm3 have similar gene organizations and encode promoter-specific isoforms with unknown functions. The underlying mechanism of the antagonism demonstrated in this study remains to be deciphered and is likely under multilayered regulation through isoform expression, mRNA/protein stability and the potential to form homo- and heterodimers. The very short unique N-terminus of Nub-PD implies that the domains required for transcriptional activation, e.g. via protein-protein interaction, are located in the larger N-terminal part of Nub-PB, although prediction analyses of protein functions did not reveal any distinct domains. Nub-PD might act as a passive repressor in a competitive manner, by binding to target sequences and prevent recruitment of additional factors, or alternatively form inactive heterodimers with Nub-PB. In accordance, isoform dimerization of the human POU protein Brn results in its inactivation [37]. Opposing effects of rat Oct-2 isoforms in vitro have also been reported and were dependent on the sequence and position of the octamer motif [38]. Finally, the human Oct-1 locus encodes at least three N-terminally distinct isoforms, which have been reported to act in partly distinct, albeit not opposite, manners [39]. It is hence plausible that distinctly acting, or even antagonistic isoforms of POU/Oct factors are an evolutionarily conserved phenomenon.
Intestinal immune and stress responses need to be tightly controlled to avoid excessive damage to host tissues. The gut microbial level and composition in adult flies correlate with host immunity and lifespan [6, 40, 41]. Age-related gut dysfunctions have been linked to dysbiosis, chronic inflammation and ultimately host death [42]. In line with these studies, we have previously observed that flies with the nub1 mutation (disrupts Nub-PD, but not Nub-PB expression) display chronic immune activation, microbial dysplasia and shortened lifespan [15, 23]. Importantly, we found that both isoforms are expressed in midgut enterocytes (Figs 4A–4D and S5B) and that overexpression of Nub-PB in these cells resulted in overall similar phenotypes as those previously found in nub1 mutants (Fig 4E, 4F, 4H and 4I), suggesting that the function of Nub-PD in this context is to counteract Nub-PB and repress aberrant responses. In fact, prolonged enterocyte-specific overexpression of Nub-PB was sufficient to drive most aspects of immune and inflammatory responses previously recorded during oral infection [31] including JNK and JAK/STAT pathway activation, Upd3 induction and increased gut mitosis and apoptosis (Fig 6). It is hence plausible that Nub-PB represents a missing node in the Upd3 mediated signaling from enterocytes to intestinal stem cells resulting in subsequent activation of JAK/STAT-driven stem cell proliferation and epithelial renewal [31], although a recent study demonstrated that a large number of transcription factors could potentially be involved [43]. The shortened lifespan of these flies correlated with a decreased microbiota (Fig 4F, 4H and 4I) and enhanced clearance of orally administered Ecc15, indicating that death is not a direct consequence of bacterial overgrowth but rather occurs due to a hyperactive immune response. In agreement, genetic manipulations to inhibit feedback regulation of the IMD pathway impair host lifespan and survival to infection [44–45]. Also, similar to nub1 mutants [23], microbial depletion extended longevity in nub-RB overexpressing flies, which furthermore suggests that bacterial exposure could aggravate the inflammatory response triggered by imbalance between Nub isoforms (S6B Fig). Enterocyte-specific and adult-restricted RNAi of nub-RB yielded overall the opposite phenotypes: reduced DIRG transcription, increased level of gut bacteria and enhanced lifespan, suggesting that Nub-PB is both necessary and sufficient to drive these phenotypes (Fig 4C and 4G–4I).
Importantly, genes involved in mounting an immune response at all levels from recognition to effectors and cytokines were activated by Nub-PB (Fig 1F). In contrast to Rel-driven immune responses [2], very few components providing negative feedback regulation were induced by Nub-PB. We hence speculate that endogenous Nub-specific transcription might be regulated via feedback on the Nub isoforms per se, possibly through protein degradation, a typical feature of many transcription factors. In addition, the combined effect of Nub-PB overexpression and infection resulted in immune gene expression levels up to three orders of magnitude above those of infected control flies, which implies that Nub-PB can enhance transcription during infection. Such vast expression levels are likely to impact host tolerance to microbial exposure [46] through loss of homeostasis, generation of self-inflicted damage, stress or even a metabolic collapse, which may explain the shortened life span and hypersensitivity to infection. Interestingly, Oct-1-deficient mouse fibroblasts are hypersensitive to stress [47], of which some parallels can be drawn to Nub-PB overexpressing, as well as nub1 flies [15, 23]. We propose that Nub-PD, in analogy with to Oct-1, acts as a stress sensor to neutralize the activity of Nub-PB, and that a balanced ratio between the isoforms is required to maintain a healthy gut environment.
Our finding that two N-terminal isoforms of the Oct-1/Oct-2 homolog, Nub, play antagonistic roles in immune/stress gene transcription provides genetic evidence of a novel switch-like regulation mediated via the same gene. We further suggest that Nub-PB and Nub-PD together form a molecular rheostat that dynamically tunes the transcriptional output to balance responses and efficiently eradicate pathogens while avoiding excessive activation and autoimmune-like reactions. This raises the possibility that Nub protein isoforms also regulate other physiological and developmental processes in opposite directions. Furthermore, the findings highlight a potential need to scrutinize the present view of POU/homeodomain networks by considering the presence of antagonistically operating isoforms, which may radically alter the transcriptional output. It remains to be explored whether similar modes of molecular rheostasis constitute a general and evolutionarily conserved mechanism to ensure flexible adjustment to environmental and developmental cues.
The following transgenic fly lines were used in this study. (A) w1118;; (RRID:BDSC_5905) was applied as wild type. (B) y1w*; nubMI05126 (RRID:BDSC_37920). In this stock, the MiMIC cassette [48] is inserted into the 5’ UTR of nub-RB and the GFP expression derived from the MiMIC cassette is under control of the endogenous nub-RB promoter. (C) w*; nubAC-62; UAS-mCherry/TM3 (RRID:BDSC_38418). The nubAC-62 allele carries a Gal4 enhancer trap inserted into the upstream region of nub-RD promoter, and has been recombined with nuclear UAS-mCherry (Müller and Affolter, personal communication to Flybase). (D) w*; nubGAL4.K (RRID:BDSC_42699). The nubGal4.K line carries a Gal4 reporter driven by a 5.3 kb fragment from the nub-RD promoter and approximately 5 kb of upstream sequence[49]. CecA1-lacZ reporter lines: (E) pA10 [50] and (F) pA10ΔOct, (this work). Lines for overexpression of transcription factors: (G) w;; UAS-nub-RD (15), (H) w; UAS-nub-RB (this work), (I) w;; UAS-DomeDN/TM3, sb1 (a gift from Nicolas Buchon). Lines for RNAi (J) w; UAS-nub-RB-IRKK113120 (RRID:FlyBase_FBst0476872, no predicted off-targets), (K) y1, v1;; nub-RD-IRP{TRiP.JF02973}attP2/TM3, sb1 (RRID:BDSC_28338; moved to a w1118 background prior to experiments), (L) w*;; UAS-bsk-IR. Gal4 driver lines for expression in specific tissues: (M) w1118; c564-Gal4 (RRID:BDSC_6982) and (N) w1118; c564-Gal4; tub-gal80ts for fat body and midgut, and (O) w; NP1-Gal4 (a gift from Bruno Lemaitre) combined with Tub-gal80ts or (P) w; NP1-Gal4; Tub-gal80ts, UAS-GFP (a gift from Nicolas Buchon) for midgut enterocytes and (Q) w;; nub-RD-Gal4VT006452 (RRID:FlyBase_FBst0483181). Mutant fly strains: (R) w;; RelE20, es (RRID:BDSC_9457). (S) w; UAS-nub-RB; UAS-nub-RD was constructed from (G) and (H); (T) w; UAS-nub-RB; UAS-DomeDN from (H) and (I); (U) w[1118]; UAS-nub-RB; UAS-bsk-IR from (H) and (L); (V) w1118; c564-Gal4; RelE20, es from (M) and (R); [23] w1118; UAS-nub-RB; RelE20, es from (H) and (R) using the double balancer line (W) w1118; if/CyO; MKRS/TM6B, tb1. (X) w;; TRE-eGFP, (Y) w;; puc-lacZ/TM3 (gifts from Ulrich Theopold /Dirk Bohmann), (Z) w; upd3-lacZ and (AA) w; 10xStat92E-GFP (gifts from Nicolas Buchon) were recombined to (I) and crossed to (O) or (P) for immunostainings.
Flies were maintained on instant potato medium [23] in mixed female/male populations at 25 °C, 60% RH, with a 12 h light/12 h dark cycle. For experimental crosses, flies were reared at 18 °C until at least two days post eclosure, and then switched to 29 °C. Overexpression experiments were carried out following two days of incubation at 29 °C. For RNAi, flies were maintained at 29 °C for 5–7 days prior to experiments. All experiments were performed using 5–10 day old females with the exception of lifespan (newly eclosed males and females, which were recorded daily) and survival assays post infection (5–10 day old males and females maintained separately). For lifespan analysis in germ free conditions, food was supplemented with an antibiotic cocktail [23].
Drosophila mbn-2 cells (DGRC, cat. no. 147) were cultured in Schneider’s medium (Gibco) supplemented with 10% fetal bovine serum (Gibco) in 5 ml plates at 25 °C, to a cell density of approximately 6–7 × 106 cells/plate.
Total RNA was isolated from OrR males with TRIsure (Bioline) and used as template for reverse transcriptase (RT). Due to the gene locus size, two RT-PCR reactions were performed in parallel, using coupled AMV RT and Tfl DNA polymerase (Access RT-PCR system, Promega Biotech AB), to amplify the 5’ and 3’ fragments of the nub-RB cDNA separately. The 5’ forward primer was constructed with a NotI site-containing overhang and an internal reverse primer; the reverse with an internal forward primer and a reverse primer with a BamHI site-containing overhang at the 3’-end of the nub-RB cDNA. The two cDNAs, 1603 bp and 1803 bp respectively, which partly overlap and contain a common EcoRV site in exon 5, were cloned into the pGEM-T easy vector. After DNA sequence verification, the two nub-RB cDNA halves were excised with NotI/EcoRV and EcoRV/BamHI respectively and then ligated into the pcDNA3.1(-) vector to create the complete 2883 bp nub-RB coding sequence with short 5’ and 3’ UTRs. Drosophila expression plasmids were created using Gateway Technology (Invitrogen, Carlsbad, CA, USA). Briefly, nub-RB coding cDNA was amplified from pcDNA3.1(-)nub-RB using Pfu DNA polymerase (Thermo Fisher Scientific, Waltham, MA, USA) according to standard procedures. The purified PCR product was cloned into the pENTR/D-TOPO vector using pENTR Directional TOPO Cloning (Invitrogen) followed by recombination of the nub-RB cDNA into the pTW and pAW destination vectors (obtained from TD Murphy) using LR Recombination and the LR Clonase enzyme mix (Invitrogen). P element transformation of w1118 flies with pTW-nub-RB was performed according to standard procedures[51]. Transfection of cells with pAW-nub-RB is described below.
Plasmids and transgenic fly lines with CecA1-luciferase and CecA1-lacZ (pA10) reporter constructs with the complete CecA1 upstream region and a short 5’ UTR have been described previously [13, 47]. To create a CecA1-lacZ reporter with the Oct cluster deleted, pA10 was used as template for inverse PCR amplification with phosphorylated primers and cloned as described previously for CecA1ΔOct-luc [15]. The whole CecA1ΔOct-lacZ fragment was thereafter excised by XbaI-XhoI digestion and ligated into the P element vector pW8 plasmid, opened with the corresponding enzymes. P element-transformation of w1118 was carried out according to standard procedures [51].
Transfections were performed with 1 μg of pA10-luc construct and mixed with 1 μg of pAW-nub-RB, 500 ng of pAct-Rel, and 100 ng of Pol III-Renilla luciferase (Addgene plasmid 37380) as internal reference. Carrier DNA was added to reach 10 μg in each sample, and transfections were performed using a calcium phosphate transfection kit (Invitrogen) as described previously [15]. Luciferase values were measured by the Dual-luciferase Reporter Assay System (Promega). To stimulate immune activation, peptidoglycan was added in the form of a crude lipopolysaccharide (LPS) preparation (25 μl of 2 mg/ml), 4 h prior to harvest.
For analysis of CecA1-lacZ reporter gene expression in transgenic flies, adults were dissected to remove heads and separate the digestive system from the rest of the fly, then fixed in 1% glutaraldehyde in phosphate-buffered saline and stained for β-gal activity using 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal) as substrate, as described previously [52]. Incubation with substrate was done for 2 h at 37 °C, and in some cases continued for 16 h at 25 °C. To estimate X-gal intensity, 8-bit images were first processed in Adobe Photoshop (version 2015 CC) using the grayscale tool to convert all colors to white except blue (converted to grayscale). Mean gray values per fly were subsequently measured in ImageJ (in the range 0 (white, no staining) to 255 (black)) using the freehand tool (N = 6).
Total RNA extractions were carried out using TRIsure (Bioline) from adult females (three flies for whole fly extracts and six dissected tissues, respectively, per replicate), followed by DNAse treatment and cDNA synthesis as previously described with a few modifications[15]. Taqman probes/primers were used to measure gene expression according to the manufacturer´s instructions (Applied Biosystems). Primer/Probes: nub-RD (CG34395-PD): Dm01841366_m1 (Applied Biosystems); nub-RB (CG34395-PB): Dm01812808_s1 (Applied Biosystems). Primer/probes for AMP gene expression were as previously published [15]. Samples were analyzed in biological triplicates or quadruplicates and relative amounts of each target were quantified relative to a set standard curve pooled from all samples in the analysis and finally normalized relative to those of RpL32.
Overexpression of nub-RB, was carried out in flies of the genotype w; UAS-nub-RB/c564-Gal4 and compared with flies carrying the c564-Gal4 driver but no UAS target gene (w; c564-Gal4/+). Flies were raised at 18 °C and adults reared in mixed sex populations at 25 °C. Female flies were used at 10 days of age, dissected to separate guts and the rest (flies without guts and heads). Total RNA extractions were carried out using TRIsure (Bioline), followed by DNAse treatment with Turbo-DNase (Ambion) and purification using RNeasy (Qiagen). Tissues (25 per replicate) from three or four independent pools of flies were analyzed as biological replicates on Drosophila Agilent microarrays. Total RNA (150 ng) was amplified and labeled using Low Input Quick Amp Labeling Kit according to the manufacturer’s instructions. Cyanine 3-CTP-labelled c-RNA (1.65 ug) was used for 17 hour of hybridization at 65 °C to the Drosophila (V2) Gene Expression Microarray, 4x44K. The hybridized arrays were washed and scanned with the Agilent DNA microarray scanner G2505C. The fluorescent intensities of the scanned images were extracted and preprocessed using the Agilent Feature Extraction Software (version 10.7.3.1). Preprocessing of the raw data was done according to the standard analysis pipeline at the Bioinformatics and Expression Analysis Core Facility at Karolinska Institutet, Huddinge, Sweden. In short, Agilent processed signals (i.e. feature gProcessedSignal) were imported to Partek Genomics Suite and subjected to quantile normalization. After preprocessing and normalization, the data was filtered to remove genes that were not expressed at detectable levels (estimated background signal). A factorial map of principal component analysis was executed on the whole expressed data using Bioconductor 3.1 and R 3.1. Multiple T-tests of the entire dataset were performed in Graphpad Prism 6 and p-value thresholds adjusted using the FDR approach (Q set to 1%). Filtered data (above detection limit in at least one of the groups compared and below the adjusted p-value threshold) were further used throughout the extended explorative downstream analysis with a few exception were a subsequent fold change cut-off were additionally applied (where denoted). The GSEA to reveal enriched GO biological processes was performed using Cytoscape (version 3.6.0) and the plugin Bingo (version 3.0.3). The analysis was executed using the hyper-geometric test with Benjamini-Hochberg FDR correction. Venn diagrams were constructed using the web-based software Venny (version 2.1).
Female guts were dissected in PBS, pH 7.4 and fixed with 4% paraformaldehyde (PFA) at room temperature for 1 h. Following two 5 min washes in PBS-T (1x PBS +0.1% TritonX-100), tissues were incubated in blocking solution (PBST+ 0.5% normal goat serum) at room temperature for 30 min and probed with primary antibodies (mouse α-β-gal [1:20; DSHB, RRID:AB_528101]; rabbit α-PH3 [1:300; Millipore, RRID:AB_310177]; rabbit α-cleaved caspase-3 [1:300; Cell Signaling, RRID:AB_2341188]) at 4 °C over night. The next day, tissues were washed in PBST 4×15 min, incubated with secondary antibodies (goat α-mouse Alexa594 [1:1000; Invitrogen, RRID:AB_141372]; goat α-rabbit Alexa594 [1:1000; Invitrogen, RRID:AB_141359]) at room temperature for 2 hours, washed again in PBST 4×15 min and then stained with DAPI (Sigma) at room temperature for 10 min. Stained samples were mounted on a glass slide with DABCO (Sigma) and confocal images were acquired using a LSM 780 microscope (Zeiss).
Carried out as previously described with a few modifications [15]. Briefly, whole fly extracts were prepared by grinding three flies per replicate in 2x standard Laemmli buffer in a 1.5 ml microcentrifuge tube using a plastic pestle followed by 10 min heating at 70 °C and 15 min centrifugation at 4 °C to remove debris. Samples were run on 10% polyacrylamide gels (125V, 90 min), wet transferred to PVDF membranes (Millipore; 10V O/N followed by 40V for 1 h) and blocked with 5% non-fat dry milk in Tris-buffered saline, 0.1 Tween-20. Membranes were incubated with rabbit-α-Nub [15] in TBS-T at 4 °C overnight, washed and incubated for 1 h at RT with HRP-conjugated donkey-α-rabbit (1:10,000; GE Healthcare, cat. no. LNA934V/AG). Bands were obtained using SuperSignal West Pico (Thermo Fisher Scientific). After quick stripping (5 min protocol using 0.3 M NaOH), PVDF membranes were blocked and reprobed with mouse α-β-Actin (1:10,000; Abcam) followed by HRP-conjugated sheep-α-mouse (1:10,000; GE Healthcare, cat. no. LNA931V/AG) as loading control.
Erwinia carotovora carotovora 15 (Ecc15) and Ecc15-GFP, kindly donated by Bruno Lemaitre, were cultured in LB medium at 30 °C with shaking [53]. For oral infections, overnight cultures of bacteria were pelleted and resuspended to OD100 in a 1:1 ratio of bacterial medium and Milli-Q H2O with 5% sucrose and 1% isotonic phosphate-buffered saline (PBS), pH 7.4. Infection vials were prepared by depositing 100 μl of bacteria or control solution on a filter paper placed on top of a 1% PBS, 2% agar gel to maintain humidity. Prior to challenge, flies were starved and desiccated in empty vials for 2 h, then briefly anesthetized to allow transfer to infection vials. Flies were maintained on this diet for 1 h (bacterial count) or 24 h (survival analysis) and subsequently transferred to fresh food vials. For systemic infections, E. cloacae β12 was grown over night at 37 °C with shaking, pelleted and resuspended in PBS at OD 1. Flies were infected by abdominal injections with approximately 50 nl of bacterial suspension under brief anesthesia. Following all infections, flies were maintained at 29 °C. Smurf assay [54] was initiated at 16 hpi by transferring flies onto regular fly food supplemented with Coomassie Brilliant Blue FCF (commercial, food grade). Flies were maintained up to two weeks post infection on this diet and observed for smurfs daily.
For relative 16S rDNA quantification of gut bacteria, midguts from six flies per replicate were dissected and bacterial DNA extracted using the DNA Blood and Tissue (Invitrogen) kit according to instructions, including the Gram+ lysis step. For bacterial counts of Ecc-15 post infection, Ecc15-GFP was applied and cultured in LB with carbenicillin (50 μg/ml). Oral infections were performed as described above. At indicated time points, individual flies were anaesthetized and ground in 100 μl PBS on ice. Ten-fold serial dilutions were added to LA-carbenicillin plates and subsequently incubated over night at 30 °C. GFP-positive bacterial colonies were quantified from seven individual flies per replicate. The experiment was performed three times for conditional rearing, and once under germ free conditions.
Carried out as described by Ja et. al with a few modifications [55]. Two females per replicate were placed in a vertically standing microcentrifuge tube with the bottom part excised and sealed with cotton to allow air exchange. Microcapillaries (3.2 mm, 5 μl; Drummond) were filled with a 5% sucrose, 5% yeast extract solution placed through a small hole in the cap of the tube. To prevent evaporation, tubes were maintained in a high humidity climate chamber, with quick replacement of microcapillaries every 24 h. Following 48 h entrainment at 18 °C, tubes were placed at 29 °C to induce Gal4-activity. For overexpression, measurements were performed between day 2–3; for RNAi between day 5–6 after the temperature shift.
Analyses of two sample means were performed using a two-tailed Student’s unpaired t-test (Figs 4C, 4D, 5I–5L, 6M and S5B–S5D, S7B and S7C). Equal variances between the groups were ensured using an F-test (p>0.05). Multiple comparisons were carried out using a one-way (Figs 2A–2C, 3H, 4E–4G, 5N and 6N), or two-way ANOVA (Figs 2D–2F, 5C–5G and 6O–6Q) combined with Tukey’s post hoc test. In Fig 5C–5F, denoted significant differences between cohorts were derived from the interaction between the two factors: 1) time post infection and 2) nub-RB overexpression as determined by a two-way ANOVA without further post-hoc analysis. In Fig 5N, Dunnett’s test was applied to analyze significant fold changes post infection relative to at 0 hpi (set as control). Lifespan and survival assays were analyzed using Mantel-Cox log-rank test with Bonferroni-corrected thresholds for significance (p<0.05/number of comparisons). Normalized qRT-PCR data were log2-transformed in order to model proportional changes prior to statistical analysis. Statistical analyses and graph constructions were carried out in Graphpad Prism 6. For experiments involving dissections, sample sizes were set to N = 3–4 to allow collections within the designated time point and to minimize sample degradation due to handling. Parametric tests were chosen based on previous experience of the normality of gene expression of the selected targets. Fold changes in expression were quantified relative to the mean value of the control cohorts.
|
10.1371/journal.ppat.1005987 | Swimming Motility Mediates the Formation of Neutrophil Extracellular Traps Induced by Flagellated Pseudomonas aeruginosa | Pseudomonas aeruginosa is an opportunistic pathogen causing severe infections often characterized by robust neutrophilic infiltration. Neutrophils provide the first line of defense against P. aeruginosa. Aside from their defense conferred by phagocytic activity, neutrophils also release neutrophil extracellular traps (NETs) to immobilize bacteria. Although NET formation is an important antimicrobial process, the details of its mechanism are largely unknown. The identity of the main components of P. aeruginosa responsible for triggering NET formation is unclear. In this study, our focus was to identify the main bacterial factors mediating NET formation and to gain insight into the underlying mechanism. We found that P. aeruginosa in its exponential growth phase promoted strong NET formation in human neutrophils while its NET-inducing ability dramatically decreased at later stages of bacterial growth. We identified the flagellum as the primary component of P. aeruginosa responsible for inducing NET extrusion as flagellum-deficient bacteria remained seriously impaired in triggering NET formation. Purified P. aeruginosa flagellin, the monomeric component of the flagellum, does not stimulate NET formation in human neutrophils. P. aeruginosa-induced NET formation is independent of the flagellum-sensing receptors TLR5 and NLRC4 in both human and mouse neutrophils. Interestingly, we found that flagellar motility, not flagellum binding to neutrophils per se, mediates NET release induced by flagellated bacteria. Immotile, flagellar motor-deficient bacterial strains producing paralyzed flagella did not induce NET formation. Forced contact between immotile P. aeruginosa and neutrophils restored their NET-inducing ability. Both the motAB and motCD genetic loci encoding flagellar motor genes contribute to maximal NET release; however the motCD genes play a more important role. Phagocytosis of P. aeruginosa and superoxide production by neutrophils were also largely dependent upon a functional flagellum. Taken together, the flagellum is herein presented for the first time as the main organelle of planktonic bacteria responsible for mediating NET release. Furthermore, flagellar motility, rather than binding of the flagellum to flagellum-sensing receptors on host cells, is required for P. aeruginosa to induce NET release.
| Pseudomonas aeruginosa leaves a large footprint in human disease because it causes infections in immunocompromised patients. Its ability to quickly adapt to diverse environments and to form biofilms poses a significant challenge to the medical community. Neutrophil granulocytes, professional phagocytes found cruising through the body’s circulatory system and tissues, provide the most efficient immune response against P. aeruginosa. Neutrophils utilize multiple strategies to eliminate bacteria. Formation of neutrophil extracellular traps (NETs), a DNA-based scaffold with attached antimicrobial proteins, provides an efficient mechanism to trap P. aeruginosa. The detailed mechanism of NET release induced by bacteria remains unclear. Our data show that the flagellum, the organelle that provides swimming motility to P. aeruginosa, is the main factor required to induce NET release. Our novel findings indicate that the flagellum, and in particular swimming motility, mediates P. aeruginosa-induced NET extrusion independently of the well-characterized flagellin receptors. The novel data presented here also suggest that down-regulation of flagellar motility characteristically seen in P. aeruginosa lung infections in cystic fibrosis is relevant for P. aeruginosa to avoid neutrophil attacks.
| Pseudomonas aeruginosa is a ubiquitous opportunistic Gram-negative pathogen found in the environment. P. aeruginosa rarely infects healthy individuals and mainly causes lung infections in patients with compromised immune defenses [cystic fibrosis (CF), chronic obstructive pulmonary disease (COPD), HIV, non-CF bronchiectasis and hospital-acquired pneumonia] [1–6]. P. aeruginosa colonizes up to 80% of CF patients, 4–15% of COPD patients, 8–25% of HIV patients with pneumonia, 28% of non-CF bronchiectasis patients and 18–20% of patients with hospital-acquired pneumonia [4, 7–9]. The high incidence of P. aeruginosa infections among these patients demonstrates that this bacterium represents a serious clinical problem.
Polymorphonuclear neutrophilic granulocytes (PMN) play a critical role in fighting P. aeruginosa. Mammalian species lacking phagocytic cells or innate immune defense molecules are highly susceptible to infection with P. aeruginosa [10–12]. Humans deficient in key neutrophil-mediated antimicrobial mechanisms, such as specific granule deficiency or leukocyte adhesion deficiency (LAD), are prone to P. aeruginosa infection [10]. Neutropenia, caused by chemotherapy, HIV infection or autoimmune disorders, predisposes patients to P. aeruginosa pneumonia [13–15]. Only patients with the full defensive arsenal of PMNs are able to defeat P. aeruginosa infections. An adequate immune response to P. aeruginosa requires the full spectrum of neutrophilic defenses.
PMNs are the first to arrive at the site of infection where they fight pathogens via various mechanisms. In addition to phagocytic killing [16], PMNs also trap and kill microbes via an alternative mechanism known as Neutrophil Extracellular Trap (NET) formation [17]. NETs are composed of a DNA scaffold associated with histones and neutrophil granule components, such as myeloperoxidase (MPO) and neutrophil elastase (NE) [17–19]. Only NET-forming PMNs and not apoptotic or necrotic PMNs release protein-DNA complexes (MPO-DNA, NE-DNA or histone-DNA) [17, 20–22]. Signaling pathways leading to NET formation are largely unknown. The few known players are: NADPH oxidase, MPO, HNE (human neutrophil elastase) and histone citrullination mediated by peptidylarginine deiminase 4 (PAD4) [23, 24]. Both MPO and HNE are required for NET release [23]. The neutrophil respiratory burst produced by the NADPH oxidase is also essential for induction of NET formation by most bacterial stimuli studied [25, 26]. PAD4-mediated citrullination of histones is required for NET formation [27, 28]. These citrullinated histones are only present in NETs, not in resting PMNs [29] and PAD4-deficient murine PMNs do not form NETs [28, 30]. PAD4-deficient mice have impaired NET-mediated antibacterial defenses [27].
Robust neutrophil infiltration and NETs have been detected in most of the diseases associated with P. aeruginosa lung infection [31–39]. This suggests that P. aeruginosa-triggered NET formation takes place in vivo under those disease conditions. Several independent studies performed by us and other groups confirmed that P. aeruginosa induces robust NET release in human PMNs [19, 33, 40–45]. We found that P. aeruginosa-induced NET formation requires the NADPH oxidase that leads to the release of citrullinated histones [19, 41]. However, the mechanism by which P. aeruginosa initiates NET extrusion from PMNs remains unknown.
Here, we aimed to identify components of planktonic bacteria and their associated mechanism(s) responsible for inducing NET release in PMNs. We identified the flagellum as the main bacterial component required to trigger maximal NET release. Interestingly, flagellum-mediated swimming motility, and not flagellum production itself, proved to be the main inciting mechanism. Our studies provide novel insight into P. aeruginosa-induced NET formation, a host-microbe interaction clinically relevant in several airway diseases.
Although bacteria have been shown to trigger NET release, it is unknown which microbial components mediate this process. To gain insight into this question, we monitored P. aeruginosa’s ability to trigger NET release at various phases of growth as P. aeruginosa expresses different phenotypic features depending on its growth phase [46]. Early exponential phase cultures are characterized by motility and expression of virulence factors, while in later growth phases these features are lost and quorum-sensing molecules and extracellular polysaccharides become expressed to a greater extent [47]. We used early exponential phase (OD = 0.4 at 600nm, ~6hrs incubation) (Fig 1A), early stationary phase (~24 hrs) and late stationary phase (~48 hrs) cultures of two laboratory strains of P. aeruginosa, type A flagellin-producing PAK and type B flagellin-producing PAO1 [48]. PMA (phorbol 12-myristate 13-acetate, a potent activator of PKC) is capable of inducing robust NET release by PMNs [19] and was utilized as a positive control. At the indicated times, bacteria were washed and exposed to human PMNs to measure NET release. As Fig 1B and 1C show, early exponential phase cultures of both strains induced the greatest NET release with decreasing induction of NET release by bacteria at each subsequent time. The PAK strain reproducibly induced larger amounts of NETs than PAO1 (Fig 1B and 1C). These data show that P. aeruginosa in its early exponential growth phase triggers the most robust NET release.
This finding suggested that bacterial components expressed at this early growth stage but lost at later stages are the main inducers of NET release. Flagellum-promoted swimming motility is often a hallmark characteristic of planktonic bacteria in their exponential growth phase [49, 50]. To study the role of the flagellum in NET formation we used flagellum-deficient PAO1 and PAK strains (PAO1 fliC and PAK flgC). The fliC gene encodes the flagellin monomer that polymerizes to form the flagellar filament, and the flgC gene encodes the flagellar hook to which the flagellar filament attaches [51]. As expected, both flagellum-deficient strains were immotile whereas their parental, flagellated counterparts displayed strong swimming motility (S1 Fig).
There are currently no commercially available methods to quantitate flagellin production in P. aeruginosa. Immunoblotting performed on bacterial lysates using an anti-P. aeruginosa flagellin antibody is described in the literature, but this method provides only semi-quantitative results [52]. Therefore, we developed an ELISA assay using a commercially available antibody capable of accurate quantitation of P. aeruginosa flagellin levels in bacterial lysates. Briefly, bacterial lysates are immobilized to the bottom of high-binding ELISA plates, blocked and exposed to anti-P. aeruginosa flagellin antibody, followed by repeated washes and addition of a secondary, peroxidase-labeled, anti-murine IgG antibody (S2A Fig). Reliable and highly reproducible standard curves can be established using commercially available, purified P. aeruginosa flagellin resulting in a tight correlation between flagellin levels and optical density (S2B Fig). To show the specificity of the assay for flagellin obtained from P. aeruginosa, we tested the ELISA assay with identical concentrations of P. aeruginosa and Shigella flexneri flagellin. The assay detected flagellin derived only from P. aeruginosa, not from S. flexneri (S2C Fig). With this new tool, we observed no flagellin expression by the PAK flgC strain, as opposed to confirmed flagellin expression by its parental strain (PAK WT) (S2D Fig).
Previously, we demonstrated that human PMNs release active MPO and HNE in the presence of P. aeruginosa PA14 [19]. Fig 2A shows that PAO1 and PAK strains also induce MPO and HNE release in human PMNs. Flagellum-deficiency significantly reduced P. aeruginosa-triggered MPO release [PAO1: 53.5+/-12.3% reduction (p = 0.0495), PAK: 44.3+/-13.1% reduction (p = 0.0296)] (Fig 2A). HNE release was also reduced in the case of both strains [PAO1: 56.4+/-11.5% reduction (p = 0.0482), PAK: 32.0+/-6.0% (p = 0.0467)] (Fig 2A). Our previous data also show that MPO remains enzymatically active after being released from PA14-exposed PMNs [19]. This was also true using both PAO1 and PAK strains (Fig 2B). Flagellum-deficient strains induced significantly less release of active MPO than their corresponding wild-type strains [PAO1: 68.0+/-11.8% reduction (p = 0.020), PAK: 80.9+/-6.2% reduction (p = 0.004)] (Fig 2B). Thus, the bacterial flagellum is required to maximal release of HNE and active MPO from PMNs upon P. aeruginosa exposure, allowing for a more impactful immune response.
Our previously published data suggest that NET formation provides the primary mechanism of MPO and HNE release from PMNs in the presence of P. aeruginosa [19]. We next tested how flagellum deficiency affects P. aeruginosa-initiated NET release. Non-flagellated P. aeruginosa strains induced only minimal extracellular DNA release (ecDNA) while their flagellated counterparts triggered a signal closer to that induced by PMA in human PMNs (Fig 2C). Lack of flagellum resulted in a 74.1+/-6.3% (PAO1, n = 12) or 81.8+/-3.6% (PAK, mean+/-S.E.M., n = 5) reduction in ecDNA release (Fig 2C). To specifically quantitate NETs, we used established ELISA assays detecting NET-specific MPO-DNA and HNE-DNA complexes developed in our laboratory [41, 53]. These assays do not detect NET components alone (DNA, MPO, HNE or nucleosomes) (S3 Fig). We observed robust NET release triggered by the wild-type flagellated P. aeruginosa (PAO1 and PAK) but not by isogenic non-flagellated bacteria (Fig 2D). Lack of flagellum resulted in a reduction in P. aeruginosa-induced MPO-DNA release of 77.2+/-20.7% (PAK, n = 3) and 61.4+/-17.1% (PAO1, n = 3), as well as, a HNE-DNA release reduction of 88.0+/-12.5% (PAO1, n = 3) and 109.9+/-11.5% (PAK, mean+/-S.E.M., n = 3) (Fig 2D).
Since NETs have a distinctive morphology [17] and we had previously shown that MPO and citrullinated histone H4 co-localize with DNA in P. aeruginosa-induced NETs [19, 41], we compared immunofluorescence staining of human neutrophils exposed to flagellum-deficient bacteria to those exposed to wild-type strains. The absence of flagellum greatly reduced NET release, as assessed by the amount of characteristic DNA structures expelled from PMNs (Fig 3A and 3B). MPO and citrullinated histones co-localized with DNA in NETs triggered by flagellated P. aeruginosa (Fig 3A and 3B), further confirming that NET formation is the main neutrophil mechanism responding to P. aeruginosa [19, 41].
P. aeruginosa flagellum is required for phagocytosis by macrophages [50]. Previously, we reported the requirement of a functional cytoskeleton for human PMNs to release NETs triggered by P. aeruginosa [19]. Based on this, we tested whether the bacterial flagellum is also essential for P. aeruginosa engulfment by PMNs. Our results in Fig 4A demonstrate that phagocytosis of flagellum-deficient PAO1 is greatly diminished in comparison to its flagellated counterpart.
In addition to phagocytosis, the NADPH oxidase has also been described as a mediator of NET formation induced by different stimuli [26]. Although recently emerging data indicate the existence of NADPH oxidase-dependent and NADPH oxidase-independent mechanisms of NET release, we have previously shown that NET formation stimulated by P. aeruginosa requires the NADPH oxidase [19, 41]. Therefore, to determine if the flagellum is required for NET formation upstream or downstream of the NADPH oxidase, we measured the neutrophil respiratory burst upon exposure to P. aeruginosa. Absence of the flagellum results in markedly reduced superoxide production triggered by PAO1 and PAK (Fig 4B and 4C).
Taken together, these results identified that the flagellum plays a key role in P. aeruginosa-induced NET release via both phagocytosis and NADPH oxidase-mediated superoxide production.
Our data herein established that the flagellum is the main bacterial component of P. aeruginosa mediating induction of NET release in human PMNs (Figs 2 and 3). To determine the mechanism of this finding, we asked whether purified flagellin, the monomer constituent of flagella, is capable of triggering NET release. As flagellin of other bacterial species including Listeria monocytogenes has been shown to stimulate superoxide production in PMNs [54], we first assessed the PMN respiratory burst in the presence of commercially available, recombinant P. aeruginosa flagellin. Flagellin stimulated PMN superoxide production in the micromolar range in a dose-dependent manner (Fig 5A and 5B). However, same concentrations of flagellin failed to trigger NET release (Fig 5C and 5D) indicating that flagellin alone is not sufficient to induce NET extrusion in human PMNs. We also tested P. aeruginosa type a and type b flagellins purified from P. aeruginosa as described [55] and observed no NET release induced by them (Fig 5E). Similarly, purified flagellin of Shigella flexneri also failed to induce NET formation in human PMNs (S4 Fig).
The main surface receptor for extracellular bacterial flagellin is Toll-like receptor 5 (TLR5) that is expressed in PMNs [56, 57]. It is unknown whether TLR5 has any role in NET formation. To assess the potential contribution of the TLR5-flagellin interaction to P. aeruginosa-induced NET release, we stimulated human PMNs with P. aeruginosa in the absence or presence of a neutralizing antibody against human TLR5. This antibody inhibits P. aeruginosa flagellin-stimulated superoxide production in human PMNs in a dose-dependent manner (Fig 6A). The same concentrations of an isotype control antibody had no effect (Fig 6B). Blockade of TLR5 on PMNs with the neutralizing antibody had no effect on NETs expelled in response to P. aeruginosa PAO1 (Fig 6C). These data suggest that the flagellum mediates P. aeruginosa-induced NET formation in human PMNs in a TLR5-independent manner.
Inhibitors and blocking antibodies are currently the only options to experimentally manipulate human PMNs since these cells cannot be genetically modified in vitro. Due to the limited experimental repertoire of human PMNs, we also used primary murine PMNs allowing us to test cells obtained from genetically engineered animals. Murine PMNs isolated from bone marrow are capable of releasing NETs both in vitro [58] and in vivo [59]. We isolated viable, highly pure PMNs from murine bone marrow (S5 Fig). Using fluorescence microscopy, we observed NETs expelled by murine PMNs after exposure to P. aeruginosa with MPO and DNA co-localization (Fig 6D). Flagellum-deficient P. aeruginosa did not trigger significant NET release in murine PMNs (Fig 6D–6F). These results confirm similar human PMN data and demonstrate that murine PMNs serve as an excellent model to study the role of flagellum in P. aeruginosa-induced NET extrusion.
In addition to TLR5 sensing extracellular flagellum, flagellin in the cytosol is sensed by NOD-like receptor CARD domain containing 4 (NLRC4) [60] that is expressed in PMNs [61]. To further assess whether TLR5/NLRC4-mediated flagellin recognition has any role in NET formation, we subjected murine PMNs expressing (wild-type, WT) or deficient in both TLR5 and NLRC4 (TLR5-/- NLRC4-/- DKO) to flagellated P. aeruginosa. Flagellated PAK and PAO1 induced NET release in murine PMNs (Fig 6F and 6G). Interestingly, lack of ability to sense flagellin by murine PMNs (TLR5-/- NLRC4-/- DKO) did not affect P. aeruginosa-induced NET formation (Fig 6F and 6G). This confirms our previous data with human PMNs showing that flagellin recognition pathways are dispensable for neutrophilic deployment of NETs against P. aeruginosa.
P. aeruginosa flagellar motility has been shown to activate the PI3K/Akt pathway to induce phagocytic engulfment [62]. To study whether this pathway is required for NET formation mediated by P. aeruginosa swimming motility, we used wortmannin to inhibit the PI3K pathway. Inhibiting PI3K had no significant effect on P. aeruginosa-induced NET formation in neither human, nor murine PMNs (Fig 6H). In our previous study, the same dose of wortmannin significantly inhibited NET formation stimulated by pseudogout-causing calcium pyrophosphate microcrystals suggesting that PI3K involvement in NET release is stimulus-dependent [29].
The finding of P. aeruginosa-initiated NET formation being flagellum-dependent but TRL5- and NLRC4-independent (Figs 2–6) could be explained by the fact that the bacterial flagellum not only binds to its host receptors but also confers the ability to swim. Flagellar motility is characteristic during the early exponential phase and is lost at later stages of bacterial growth. Motility is an underappreciated feature of bacterial interactions with the host immune system that is recently gaining recognition [50, 62]. Flagellar motility, not simply flagellum production, has been shown to be a key player in initiating immune responses in macrophages [63, 64]. No study, however, has investigated the role of flagellar motility in PMN activation and NET release. To separate these two functions of the flagellum from each other we took advantage of P. aeruginosa mutants deficient in genes responsible for propulsion of the flagellum [51]. The P. aeruginosa flagellum is powered by a complex bacterial motor consisting of multiple proteins encoded by two sets of homologous motor genes: 1) motA, motB and 2) motC, motD [51, 65]. Disruption of both loci is required to completely abolish swimming motility; deletion of either set of operons is not sufficient to eliminate swimming [51, 65]. We characterized motility and flagellin production in 2 mutant PAK strains deficient in the following mot genes: strain LMP16 (ΔmotCD motB) and strain LMP50 (ΔmotAB motD) [51]. As expected, swimming motilities of the flagellar motor-deficient strains (LMP16 and LMP50) and the flagellum-deficient (flgC) mutant were abolished (Fig 7A). On the other hand, flagellin production (measured by western blotting and ELISA in bacterial lysates) was only missing in the flgC strain (Fig 7B). Flagellar motor-deficient mutants produced flagellin to an extent similar to that of the wild-type (WT) strain (Fig 7B). Together, these data demonstrate that the flagellar motor-deficient strains produce a flagellum that is paralyzed for rotation; they have impairment in swimming motility but express normal levels of flagellin.
To assess the potential role of flagellar motility in NET release induced by bacteria, we stimulated human PMNs with wild-type, flagellum-deficient and flagellar motor-deficient P. aeruginosa PAK strains to measure NADPH oxidase activity and NET release. Superoxide production was abolished in the absence of both flagellar motility and flagellum production (Fig 7C): nonstimulated (6.08+/-4.24), PMA (57.25+/-7.08), PAK WT (21.44+/-3.10), LMP16 (8.21+/-2.89), LMP50 (8.03+/-2.81) and flgC (7.68+/-3.43) (*109 RLU, mean+/-S.E.M., n = 4). Similarly, P. aeruginosa-induced ecDNA release was significantly reduced when motility was abolished (Fig 7D): 81.8+/-3.6% reduction with LMP16, 85.8+/-4.6% with LMP50 and 88.6+/-4.1% with flgC (mean+/-S.E.M., n = 5). PAK-induced MPO release showed a similar pattern (Fig 7E). Wild-type PAK triggered 767.6+/-101.0 ng/ml MPO release, whereas flagellum-deficient P. aeruginosa induced considerably less MPO release of 405.8+/-37.7 ng/ml (mean+/-S.E.M., n = 10) (Fig 7E). Immotile, flagellum-expressing PAK mutants induced PMN behavior similar to that of the flagellum-deficient strain: 387.9+/-47.9 ng/ml MPO release by LMP16 and 360.0+/-47.9 ng/ml by LMP50 strains (Fig 7E, mean+/-S.E.M., n = 10). Of note, a significant portion of P. aeruginosa-stimulated MPO release, unlike other measures, is independent of the flagellum (Fig 7E). These data confirm our previous observations that NET formation is the main, but not the only, mechanism to mediate MPO release from human PMNs stimulated by P. aeruginosa [19].
The fact that a motile flagellum provides the main mechanism of P. aeruginosa-triggered NET extrusion was further confirmed by our MPO-DNA ELISA data, showing greater amounts of MPO-DNA complexes released in response to motile P. aeruginosa (Fig 7F). NET inductions by P. aeruginosa were reduced: by 84.5+/- 7.4% (LMP16), by 73.3+/-3.5% (LMP50) and by 73.2+/-3.7% (flgC) (mean+/-S.E.M., n = 3) (Fig 7F). Immunofluorescence staining of NETs following stimulation with wild-type and PAK mutants confirmed these data qualitatively (Fig 7G and 7H). As reported previously, MPO co-localized with extracellular DNA in PAK-induced NETs (Fig 7G). Human PMNs exposed to flagellum-deficient or flagellar motor-deficient P. aeruginosa strains released only minimal amount of NETs (Fig 7G). Murine PMNs exposed to the same PAK mutants exhibited an identical pattern of NET release (Fig 7I).
Taken together, the results presented in Fig 7 indicate that flagellar motility, and not flagellum production alone, is the main factor of NET release, in the presence of flagellated P. aeruginosa.
To support our previous findings we centrifuged wild-type, flagellum- and flagellar motor-deficient strains of P. aeruginosa on human PMNs and measured NET release. Centrifugation of bacteria on PMNs bypasses the need for motility to establish cell-cell contact. All four immotile bacterial strains were capable of inducing close to maximal DNA release in human PMNs upon centrifugation on PMNs (Fig 8A and 8B). NET release by the wild-type bacterium was not affected by centrifugation (Fig 8A and 8B). To confirm that live bacteria are required to induce NET formation, human PMNs were stimulated with heat-killed P. aeruginosa and superoxide production and DNA release were measured. Both readouts were inhibited by the heat-treatment indicating that live bacteria are needed to induce maximal NET formation (S6 Fig). Thus, bypassing the requirement for motility to enable live P. aeruginosa-PMN contact restores the ability of immotile bacterial strains to trigger NETs.
To assess which motility genes are required for P. aeruginosa to induce NET formation in human PMNs, we tested PAK strains deficient in either one of the motAB or motCD loci [51]. As shown in Fig 9, motCD-deficient P. aeruginosa (LMP9, LMP84 and PAO1014) exhibited significant impairment in inducing NET formation while motAB-deficient bacteria (LMP13, LMP09 and PAO1020) remained largely unaffected. These data suggest that the motCD genes are more crucial for inducing NETs. However, NET release triggered by ΔmotCD PAK was still higher than that induced by the completely immotile bacteria that have lesions in both loci, suggesting that a fully functional motor complex is needed to trigger maximal NET formation (Fig 9).
Next we aimed at restoring the impaired NET-inducing ability of motAB-deficient bacteria by cloning and reintroducing functional motAB genes in both PAK and PAO1 backgrounds. The following PAK strains were created: wild-type PAK transformed with empty vector (LMP 80), motCD-deficient PAK with the vector (LMP84) and motCD-deficient PAK complemented with functional motCD (LMP85) (Table 1).
Similar PAO1 strains were generated: PAO1 wild-type containing empty vector (PAO1010), motCD-deficient PAO1 containing vector (PAO1014) and motCD-deficient PAO1 complemented with motCD genes (PAO1015) (Table 1). MotCD-deficiency lead to significantly impaired induction of NET release; whereas introduction of functional motCD genes resulted in increased NET release in both PAK and PAO1 backgrounds (Fig 9A and 9B). Similar pattern was observed with superoxide production (S7 Fig).
Complemented strains were also created for the motAB mutants: PAK wild-type transformed with empty vector (LMP82), motAB-deficient PAK (LMP90) and motAB-deficient PAK complemented with functional motAB genes (LMP91); PAO1 wild-type transformed with empty vector (PAO1012), motAB-deficient mutant (PAO1020) and motAB-deficient strain complemented with functional motAB genes (PAO1021) (Table 1). Consistent with previous data, motAB mutants did not result in impairment in NET-induction and there were no significant differences between the motAB-deficient strain and its complemented derivatives in each background (Fig 9A and 9B). Swimming motility of the strains was characterized in semisolid motility agar (Fig 9C). Both, motAB- and motCD-deficiencies led to significant losses of swimming motility in semisolid agar while complemented derivatives regained swimming motility (Fig 9C).
Thus, our experiments suggest that the motCD genes play a primary role in determining the extent of NET induction by planktonic P. aeruginosa but both sets of motor genes are required for maximal NET-induction.
The purpose of this study was to identify the main bacterial component of P. aeruginosa triggering NET formation and to gain insight into its mechanism. P. aeruginosa is an opportunistic pathogen representing a serious medical problem. PMNs, the primary immune cells fighting this bacterium, release large amounts of NETs when challenged with P. aeruginosa [19, 33, 41, 45]. These NETs then kill and trap bacteria [17]. NETs, however, also cause tissue damage in airway diseases characterized by P. aeruginosa infections (e.g., CF and COPD) [31, 32, 70]. Therefore, it is important to illuminate the cellular-molecular details of P. aeruginosa-induced NET formation to better understand its clinical relevance in various disorders. Our data show that early growth-phase bacteria are the strongest NET-inducers. Our data identify flagellum as the main component of bacteria triggering NETs, thereby filling in a major gap in our understanding of the molecular details of bacterium-triggered NET formation. Until now, only two bacterial components (LPS and pyocyanin) were described as weak NET-inducers [17, 40]. Our detailed characterization documents the major contribution of the flagellum to P. aeruginosa-induced NET extrusion in both human and murine PMNs, thus adding a new mechanism to the proinflammatory repertoire of flagellum [71].
Detection of P. aeruginosa flagellin was traditionally performed by immunoblotting or electron microscopy [72]. Neither of these methods is, however, suitable for accurate absolute quantitation. The new ELISA assay established here enables easier detection of varying P. aeruginosa flagellin levels in a large number of samples. The highly reproducible standard ensures accurate absolute quantitation of flagellin levels that is vital to study P. aeruginosa flagellum interactions within host cells including PMNs.
TLR5 is expressed on airway epithelial cells and several innate immune cell types including PMNs [56, 73]. TLR5 is the main receptor mediating activation of airway epithelial cells of P. aeruginosa via flagellin recognition [73] and the TLR5-flagellin interaction is a major mediator of airway inflammation in CF [74]. TLR5 also acts as a modifier gene in CF [74]. Therefore, it was very surprising to observe that P. aeruginosa-induced NET release is independent of TLR5 in both human and murine PMNs. In further support of this finding, we also show that recombinant P. aeruginosa flagellin monomers are capable of stimulating NADPH oxidase activity without inducing NET release. Flagellin monomers bind to and activate TLR5 [57]. Thus, our data suggest that the flagellum mediates P. aeruginosa-induced NET release by a novel mechanism independent of the flagellum-sensing machinery of PMNs (TLR5/NLRC4). Moreover, we found that flagellum-mediated swimming motility is the key mediator of NET release. These data add to currently published reports emphasizing the importance of swimming motility in P. aeruginosa virulence in macrophages [50, 62] and imply that flagellum also contributes to bacterial virulence by mechanisms other than activation of the TLR5 signaling pathway. Motility-mediated bacterial virulence mechanisms represent a large gap in the scientific literature and require more detailed studies. It cannot be automatically assumed that motile bacterial pathogenesis is entirely the result of direct flagellum-receptor interactions. Instead, we must consider that other motility-based mechanisms can also take place. The presence of a rotating flagellum enables P. aeruginosa to swim which significantly increases the chances for bacterium-neutrophil encounters. It is very likely that pattern recognition receptors other than TLR5 and NLRC4 are responsible for direct binding of P. aeruginosa to PMNs during initiation of NET extrusion. The identities of these receptors remain to be elucidated.
The polar flagellum of P. aeruginosa is powered by a complex motor containing dual stator units, MotAB and MotCD [reviewed in [75]]. The stators generate the torque used to turn the flagellar rotor. In liquid medium, either stator is sufficient to power the flagellum as deletion of either the motAB or motCD locus has little effect on swimming speed and only the deletion of both loci renders the bacterium immotile [51, 75]. However, under conditions requiring higher torque, for example in semi-solid motility agar as (shown in Fig 9C) or on a swarming motility plate, the MotCD stator plays the dominant role [51, 69]. Results presented here reveal that the motCD mutants strongly impair P. aeruginosa-induced NET formation, motAB mutants have little effect, and mutants with defects in both motAB and motCD, are most severely impaired for NET release induced in human PMNs. These findings suggest that motility is key to trigger NET formation. Moreover, they suggest that the stator that can provide the highest torque to the rotor is more important in NET formation. In line with previous studies, our results also suggest that flagellar motility genes could provide novel targets of pharmaceutical intervention to intervene with P. aeruginosa motility as a virulence mechanism currently gaining recognition.
A flagellum is typically expressed in environmental isolates of P. aeruginosa and early clinical isolates of CF patients [76, 77]. Loss of the flagellum is one of the characteristic changes accompanying the adaptation of P. aeruginosa in CF airways [76–80]. In chronic CF patients, P. aeruginosa mainly exists in biofilms [81]. However, biofilms are dynamic structures, and motile, flagellated bacteria likely break free from biofilms, possibly interacting with PMNs and shedding flagella. This is supported by recent data showing that P. aeruginosa flagellin is detected in sputa of chronic CF patients [82]. Thus, we speculate that the mechanism described here not only can have clinical relevance in early but also in chronic CF airway disease.
Our studies provide a potential, novel explanation as to why it is advantageous for P. aeruginosa to lose its flagellar motility early on in colonization of the airways in CF. PMNs and NETs could provide a significant external pressure for P. aeruginosa’s down-regulation of flagellin expression in CF airways. This is supported by published data showing that NE cleaves flagellin and down-regulates flagellum expression in P. aeruginosa [83, 84]. Loss of flagella and swimming motility could be the primary mechanism by which late-phase and mucoid CF isolates of P. aeruginosa acquire resistance against NET-mediated killing [33, 41, 45, 85].
It is important to note that while P. aeruginosa-induced NET release is almost entirely flagellum- and flagellar motility-dependent, total MPO and HNE release is only partially dependent on these factors. This is in line with our previous observations stating that NET formation is the main, but not the only, mechanism of MPO and HNE release from PMNs challenged with P. aeruginosa [19, 41]. Excessive degranulation could potentially be responsible for the NET-independent release of these primary granule components [86, 87].
Our results demonstrating that murine PMNs also expel NETs in response to P. aeruginosa in a flagellum-dependent but TLR5-independent manner confirm the usefulness of murine PMNs as a model to study the mechanism of NET release stimulated by planktonic forms of bacteria. Having established murine NET measurements enables us to test ex vivo NET release in genetically modified murine PMNs. This genetic approach complements results obtained with human PMNs that are not suitable for genetic modifications.
Our data show that early growth-phase planktonic bacteria are the strongest NET-inducers in human PMNs. Although the idea that planktonic bacteria can induce NET release has previously been challenged [88], our observation is in line with numerous articles published by several independent groups all reporting robust NET release induced by (planktonic or individual) bacteria of a wide variety of different species [19, 33, 41, 45, 89–91]. That said, it is very likely that different mechanisms are responsible for NET release induced by bacteria or large microbes (fungal hyphae) [88]. Although it does not form the focus of the current study, but our data showing that flagellum is required for both phagocytosis and NET release and bacterium-PMN contact is essential to induce NET formation, indicate that phagocytosis of P. aeruginosa is necessary for NET release. This idea suggests that the same PMN can engulf bacteria and undergo subsequent NET release, as well. Previously, it has been proposed that a single PMN either performs phagocytosis or undergoes NET formation but not two functions in one cell [88]. Likely, the primary response of PMNs to planktonic bacteria is phagocytosis that is followed by NET release once uptake of more microbes is not feasible. This mechanism has already been proposed earlier [26] and is also supported by our data. Future focused studies need to be performed to understand the very exciting question what factors determine PMN effector mechanisms and cell fate in response to bacteria. It is highly important to address this problem to learn about the unanswered questions of NET formation [92], to understand what leads to unnecessary PMN activation in several diseases [93] and to be able to develop novel PMN-based therapies [94].
Overall, the results presented here reveal a novel proinflammatory mechanism of the bacterial flagellum and identify it as the main factor of flagellated bacteria triggering NET formation. We also identified flagellar motility as its primary mechanism to mediate P. aeruginosa-induced activation of PMNs that likely occurs in CF airways, contributes to disease pathogenesis and possibly points to a new, future therapeutic target.
The Institutional Review Board of the University of Georgia approved the human subject study to collect peripheral blood from volunteers anonymously (UGA# 2012-10769-06). Enrolled healthy volunteers were non-pregnant and heavier than 110 pounds without any infectious disease complication. All adult subjects provided informed consent, and no child participants were enrolled into the study. The studies were performed following the guidelines of the World Medical Association's Declaration of Helsinki.
The Institutional Animal Care and Use Committees (IACUC) of the University of Georgia and the Georgia State University reviewed and approved the mouse protocols used in this study: UGA IACUC protocols: A2012 11-003-Y3-A3, A2014 08-019-Y2-A0 and GSU IACUC protocol: A14033. All animal experiments were performed in accordance with NIH guidelines, the Animal Welfare Act and US federal law. Animals were housed in centralized research facilities accredited by the Association of Assessment and Accreditation of Laboratory Animal Care International.
Human PMNs were isolated as described previously [19, 41]. Briefly, whole blood was drawn at the Health Center of the University of Georgia from volunteers. Coagulation was prevented with heparin. Red blood cells were removed by Dextran sedimentation (GE Healthcare), and PMNs were separated using Percoll gradient centrifugation. Cell viability was determined by Trypan blue dye extrusion (>98%). Neutrophil purity was assessed by cytospin preparations and flow cytometry. Autologous serum was prepared from coagulated blood by centrifugation and sterile filtration. Calcium- and magnesium-containing HBSS (Mediatech, Manassas, VA, USA) supplemented with 1% autologous serum, 5 mmol/l glucose and 10 mmol/l HEPES was used as the assay buffer.
Wild-type (WT) C57BL/6 mice were purchased from Jackson Laboratories and maintained in the animal facility of the College of Veterinary Medicine, University of Georgia, Athens. 10-15-week-old mice were used throughout the study. TLR5/NLRC4 double gene-deficient mice on a C57BL/6 background were kept in the Georgia State University animal facility. TLR5KO mice used here were originally generated by Dr. Shizuo Akira (Osaka University, Osaka, Japan) [95] and backcrossed/maintained as previously described [96]. NLRC4 KO mice generated on a pure C57BL/6j background were kindly provided by Genentech (Genentech, Inc. South San Francisco, CA) [97]. Age- and sex-matched healthy C57BL/6 mice were used as controls. Mice were euthanized on the day of the experiment by CO2 asphyxiation and cervical dislocation according to University of Georgia and Georgia State University IACUC guidelines.
Murine bone marrow-derived PMNs were collected from femur and tibia. Bones were flushed with RPMI-1640 medium (Corning, Manassas, VA) and washed in sterile PBS. 1 ml AKC buffer (Lonza, Walkersville, MD) was used to lyse red blood cells. Cells were passed through a 40 μm pore size Nylon Mesh strainer (Fisherbrand, Fisher Scientific, Pittsburgh, PA, USA) and subsequently washed twice with and re-suspended in sterile PBS before being layered on top of a two-step Percoll (Sigma-Aldrich, St. Louis, MO, USA) gradient (62% and 81%), as described previously [98]. After centrifugation (1600 g, 30 min), PMNs accumulated at the interface of 81% and 62% Percoll layers were collected and washed twice in sterile PBS. Cell numbers were determined with a hemocytometer. Cell viability determined by Trypan Blue exclusion was always higher than 98%. Neutrophil purity was also periodically confirmed with hematoxylin and eosin (H/E) staining (Sigma, St. Louis, MO, USA) of cytospins prepared using CytopsinTM Cytocentrifuge (ThermoScientific, Waltham, MA, USA) (S5C Fig).
The purity of the murine neutrophil preparations was routinely confirmed via flow cytometry (LSRII, BD Technologies). Anti-Gr-1 antibody (Miltenyi Biotec, San Diego, CA, USA) was used against Gr-1-expressing granulocytes following the manufacturer’s recommendations. Cells were analyzed in BD LSRII flow cytometer (BD Biosciences, San Jose, CA, USA) using BD FACSDiva 6.0 software (BD Biosciences, San Jose, CA USA) at the Imaging Core Facility of the Department of Infectious Diseases at UGA. The described protocol resulted in more than 95% PMNs (S5 Fig).
The following Pseudomonas aeruginosa strains were used in this study. The PAO1 parental strain (wild-type, WT) was MPAO1 and its flagellum-deficient PAO1 mutant (strain ID: 245, genotype: PA1092-G03::lacZbp01q1, referred to as “fliC”) were obtained from the Pseudomonas aeruginosa PAO1 transposon mutant two-allele library (University of Washington, Seattle, WA; Manoil laboratory) established using NIH funds (grant#: P30 DK089507) [66]. The PAK wild-type (WT) and flagellum-deficient mutant flgC1::Tn5 were obtained from Pathogenesis Corporation (referred to as “flgC”). The flagellar motor mutant strains were described elsewhere (51). Strains are listed in Table 1 that also includes the complementation plasmids [67, 68]. The gene designations were revised to be consistent within the Pseudomonas aeruginosa field and refer to PA1460-61 (motCD) and PA4954-53 (motAB) [51]. P. aeruginosa strains were cultured in Luria-Bertani broth for the indicated periods of time. Bacteria were washed twice in PBS and resuspended in calcium- and magnesium-containing HBSS. Bacterial cultures were set to an optical density (OD) = 0.6 at 600 nm in 96-well microplates measured using a Varioskan Flash combined microplate reader (ThermoScientific, Waltham, MA USA). This corresponds to a bacterial density of 109/ml, as determined by serial dilutions and colony forming unit (CFU) assays [40, 99]. In some experiments, optical density of bacterial cultures was followed over time in an Eon Microplate Spectrophotometer (BioTek, Winooski, VT) to record kinetic growth curves (Fig 1A).
The genotype of the PAO1 fliC mutant was confirmed by PCR using the primers and conditions suggested on the web site of the PAO1 two-allele library. PAO1 WT and fliC bacteria were cultured overnight, washed twice in HBSS and set to an OD = 0.6 as described above. 100,000 bacteria in 2 μL distilled water were added to the PCR reaction mix and served as a DNA template. The PCR mix (20 μl reaction volume) contained 10 mM dNTP (Life Technologies, Carlsbad, CA), 50 mM MgCl2, 10 μM forward and reverse primers and Taq DNA Polymerase (Life Technologies, Carlsbad, CA). The following gene-specific primers were used: fliC (F: 5’- TGCAGCAGTCCACCAATATC-3’; R: 5’- GTTGGTAGCGTTTTCCGAGA -3’, product size: 1081 bp), pilA (F: 5’- GGAATCAACGAGGGCACC -3’; R: 5’- ACCCAGTTTCCTTGATCGTG -3’, product size: 865 bp). PCR reaction parameters were: 94°C for 0.5 min, followed by 35 cycles of 94°C for 30 sec, 60°C for 1 min and 68°C for 90 sec. The PCR reaction was carried out in a Biometra PCR thermocycler (Biometra, Göttingen, Germany). PilA was used as loading control. The PCR products were resolved on 2% agarose gel and stained with Gelstar DNA stain (Lonza, Walkersville, MD, USA). The genotype of the flagellum-deficient PAO1 fliC strain was confirmed by PCR (S1C Fig). Lack of contaminating DNA was confirmed by PCR without template (no bacteria) (S1C Fig).
For the swimming motility assay, bacteria were grown overnight, washed twice in HBSS and set to an optical density (OD) = 0.6 as described previously. 10 μL of bacterial cultures were spotted on the center of freshly prepared LB+0.3% agar plates and incubated at room temperature. After 48 hours, diameters of colonies were measured and expressed in millimeters (mm) [100]. For complementation experiments, the strains were streaked for single colonies on LB medium with 300ug/ml carbenicillin. Three single colonies were toothpicked into tryptone motility agar containing 10 g/L tryptone 5 g/L NaCl and 3 g/L agar. Plates were incubated overnight and photographed.
Flagellin of P. aeruginosa was obtained from two independent sources. First, purified P. aeruginosa flagellin was purchased from Invivogen (San Diego,CA, USA). Flagellin obtained from this commercial source is extracted by acid hydrolysis and is purified by ultrafiltration and chromatography. The identity of the P. aeruginosa strain and the type of the flagellin were not revealed by the company. Second, P. aeruginosa flagellin was also obtained as a kind gift from Dr. Gerald Pier (Massachusetts General Hospital, Boston, MA). Recombinant P. aeruginosa flagellins were purified from E. coli expressing His-tagged type a or b fliC genes as described previously [55, 101].
DNA release from human PMNs was quantitated as described [19, 29]. Briefly, 250,000 PMNs were seeded on 96-well black transparent bottom plates in the presence of 0.2% Sytox Orange (Life Technologies, Grand Island, NY, USA) membrane-impermeable DNA-binding dye. PMNs were infected with 10:1 multiplicity of infection (MOI) P. aeruginosa as indicated. Fluorescence (excitation: 530 nm, emission: 590 nm) was recorded for up to 8 hrs in a fluorescence microplate reader (Varioskan Flash, ThermoScientific, Waltham, MA, USA) at 37°C. DNA release is expressed as % of the maximum obtained by saponin-mediated (1 mg/ml; Sigma-Aldrich, St. Louis, MO, USA) neutrophil lysis and DNA exposure.
Immunofluorescence staining of human MPO and citrullinated H4 was performed as previously described [19, 29, 40, 41]. Briefly, adherent murine or human PMNs were exposed to different strains of P. aeruginosa (10 MOI, 3 hrs, 37°C). After incubation, fixed and permeabilized samples [4% paraformaldehyde (Affymetrix, Celeveland, OH)] were blocked with 5% Normal Donkey serum (Sigma-Aldrich, St. Louis, MO, USA, in PBS) for 30 min at room temperature. Fixed human NETs were incubated with monoclonal mouse anti-human myeloperoxidase/FITC antibody (1:500, Dako, Clone MPO-7) and polyclonal rabbit anti-histone H4 (citrulline 3) (1:1000, Millipore, Billerica, MA) overnight at 4°C. The citrullinated histone staining requires the use of a secondary antibody after three washes: Alexa Fluor 594-labelled donkey anti-rabbit secondary antibody for 1 hr (1:2000, Molecular Probes, Grand Island, NY). Murine PMNs were first stained with goat anti-mouse MPO antibody (R&D Systems, Minneapolis, MN, 1:1,000) overnight at 4°C, followed by staining with FITC-labelled, donkey anti-goat IgG (Jackson ImmunoResearch, West Grove, PA, USA, 1:800, 1 hr, dark). DNA was stained with DAPI (2 min, room temperature, 1:20,000, Molecular Probes, Grand Island, NY). Specimens were washed three times with PBS containing 0.1% Tween-20 (Sigma-Aldrich, St. Louis, MO, USA) between each step. Mounted specimens were analyzed with Zeiss AxioCam HRM fluorescence microscope Axioplan2 imaging software. NET formation was quantitated by counting at least 200 PMNs per sample and by determining the proportion of NET-forming cells compared to the total population.
Concentration of human MPO in PMN supernatants was quantitated by commercial ELISA kit (R&D Systems, Minneapolis, MN, USA) as previously described [19, 41]. Human neutrophil elastase release was assessed by sandwich ELISA: diluted supernatants were applied to 96-well high binding microloan ELISA plates (Greiner bio-one, Germany) pre-coated overnight with anti-human neutrophil elastase rabbit polyclonal antibody (1:500 in PBS, Calbiochem, 481001, EMD Millipore, MA, USA). After blocking with 1% BSA for 1 hr, a secondary anti-human neutrophil elastase antibody was applied (1:2000 in PBS, IgG1, cat #: MA1-10608, ThermoScientific, Hudson, NH) followed by the addition of a horse radish peroxidase-linked (donkey) anti-mouse IgG antibody (1:2000 in PBS, NA934V, GE Healthcare, UK) for 1 hr at room temperature. Blue coloration developed in the presence of the Pierce TMB Substrate Kit (ThermoFisher Scientific, Waltham, MA, USA), and results were quantitated using a human neutrophil elastase standard.
Flagellin concentrations in P. aeruginosa lysates were quantitated by ELISA established in this manuscript. Bacterial cultures were sonicated after repeated washes and their density was adjusted as described above. Bacterial lysates were centrifuged twice (10,000 g, 15 min), and supernatants were used in the ELISA. Supernatants of bacterial lysates and P. aeruginosa flagellin standards (Invivogen, San Diego, CA) were immobilized to 96-well high-binding capacity ELISA plates (Greiner Bio-one, Frickenhausen, Germany) by mixing them with an equal volume of 100 mM carbonate/bicarbonate buffer and incubating the samples overnight at 4°C. Plates were washed three times with PBS containing 0.1% Tween-20 (Sigma, St. Louis, MO, USA) and blocked by 5% bovine serum albumin (Hyclone, Logan, Utah) for 3 hrs at room temperature. After three washes with Tween-20/PBS, anti-P.aeruginosa flagellin antibody was added (1:250 dilution in PBS, 250 ng/ml, mouse IgG1, hybridoma clone: 18D7, Invivogen, San Diego, CA) and incubated overnight at 4°C. Samples were washed again three times, followed by addition of a secondary horse radish peroxidase-labelled, sheep anti-mouse IgG antibody (1:1000 dilution in PBS, GE Healthcare Bio-Sciences, Pittsburgh, PA, USA) for 30 min (room temperature, dark). After four repeated washes with PBS/Tween-20, color reaction was developed with 3,3’,5,5’-tetremthylbenzidine (TMB, 0.16 mg/mL, Sigma, St. Louis, MO) peroxidase solution and the reaction was stopped by adding 1M HCl. Absorbance was read at 450 nm with either Eon (BioTek, Winooski, VT) or Varioskan Flash (ThermoScientific, Hudson, NH) microplate photometers. Absolute quantitation of P. aeruginosa flagellin concentrations were calculated using the standard curve and expressed as micrograms per milliliter (μg/ml) or “ng/6x108 bacteria.”
NETs (MPO-DNA and HNE-DNA complexes) in human PMN supernatants were quantitated by specific ELISA assays as described [41, 53]. Briefly, supernatants of attached human PMNs were treated with 1 μg/ml DNAseI to achieve limited DNA digestion [41, 53]. Diluted samples were added to and incubated overnight on ELISA plates pre-coated with anti-MPO or anti-HNE capture antibodies, followed by addition of horseradish peroxidase-labelled anti-DNA detection antibody [41, 53]. Coloration of added TMB substrate solution (Thermo Scientific, Hudson, NH) was stopped by 1N HCl and absorbance (450 nm) was read either with Eon (BioTek, Winooski, VT) or Varioskan Flash (ThermoScientific, Hudson, NH) microplate photometers. “NET concentrations” were expressed as percentage of the “NET-standard,” consisting of pooled supernatants (5 donors) of PMA-stimulated human PMNs after limited DNAseI-digestion, and were handled parallel with unknown samples [53].
Phagocytosis of PAO1 strains by human PMNs was assessed by measuring the decrease in the number of extracellular (non-phagocytosed) bacteria over time. Human PMNs were mixed with 10 MOI of PAO1 WT or fliC and incubated for 60 min. At different time points (0, 2, 20, 40 and 60 min), aliquots were taken, added to ice-cold PBS and centrifuged (300 g, 3 min, 4°C) to pellet PMNs but leave extracellular bacteria in the supernatant. The centrifugation step was repeated once. 100 μL volume of the supernatant was added to 900 LB growth medium, and bacterial concentration was determined using a microplate-based assay [102].
Superoxide production was measured by two different assays: Lucigenin-based or Diogenes-based superoxide chemiluminescence kits (National Diagnostics, Atlanta, GA) [19, 29, 40, 41]. 100,000 PMNs adhered to 96-well white plates for 15 min at 37°C in HBSS containing 1% serum. Cells were stimulated by Pseudomonas aeruginosa strains (10 MOI), PMA (100 nM) or left unstimulated. Chemiluminescence was measured by a Varioskan Flash microplate luminometer (Thermo Scientific, Waltham, MO, USA) for 90 min. Data are shown as kinetics of representative curves (relative luminescence units, RLU) or integral superoxide production by analyzing accumulated luminescence for the entire duration of the measurement.
Myeloperoxidase activity was measured by hydrogen peroxide-dependent oxidation of Amplex Red as described [19]. Undiluted neutrophil supernatants were mixed with assay solution containing 100 μM Amplex Red (Sigma, St. Louis, MO) and 100 μM hydrogen peroxide (Sigma, St. Louis, MO). Production of the fluorescent product was measured in 96-well black plates using a fluorescence microplate reader (Varioskan Flash, ThermoScientific, Waltham, MO, USA) for 30 min at 560 nm excitation and 590 nm emission wavelengths. Calibration was achieved using an MPO standard [19].
P. aeruginosa PAK strains were grown overnight in LB liquid medium, washed and resuspended in RIPA Lysis and Extraction buffer (ThermoFisher Scientific, Waltham, MA USA) before sonication. Bacterial lysates were collected as supernatants after centrifugation (14,000 g, 20 min, 4°C). Protein concentrations were determined with a Pierce BCA Protein Assay Kit (ThermoFisher Scientific, Waltham, MA USA). Equal amounts of proteins from bacterial lysates were loaded onto Novex 8–16% Tris-Glycine Gel together with molecular weight standards and run for 120 min at 110 V. Samples were blotted onto nitrocellulose membrane using the iBlot dry blotting system (Life Technologies, Carlsbad, CA). Membranes were blocked in 5% milk for 1 hr and probed with the primary antibody (anti-P. aeruginosa flagellin antibody, 1:250, 250 ng/ml, mouse IgG1, hybridoma clone: 18D7, Invivogen, San Diego, CA) overnight at 4°C. After three washes, the secondary antibody was added (HRP-labelled goat anti-mouse IgG, 1:2000, ThermoFisher Scientific, Waltham, MA, USA) for 1 hr at room temperature. Following three repeated washes, blots were probed with the Amersham ECL Western Blotting Detection Kit (GE Healthcare Life Sciences, Pittsburgh, PA, USA), and chemiluminescence was recorded with Konica Minolta SRX-101A developer using HyBlot CL Autoradiography films (Denville Scientific, Holliston, MA).
Results were analyzed by Student's t-test or one-way ANOVA. Each experiment was independently performed at least three times with PMNs isolated from different donors. Statistically significant differences were considered as *, p<0.05; **, p<0.01; ***, p<0.001.
|
10.1371/journal.ppat.1003074 | Transcription of a cis-acting, Noncoding, Small RNA Is Required for Pilin Antigenic Variation in Neisseria gonorrhoeae | The strict human pathogen Neisseria gonorrhoeae can utilize homologous recombination to generate antigenic variability in targets of immune surveillance. To evade the host immune response, N. gonorrhoeae promotes high frequency gene conversion events between many silent pilin copies and the expressed pilin locus (pilE), resulting in the production of variant pilin proteins. Previously, we identified a guanine quartet (G4) structure localized near pilE that is required for the homologous recombination reactions leading to pilin antigenic variation (Av). In this work, we demonstrate that inactivating the promoter of a small non-coding RNA (sRNA) that initiates within the G4 forming sequence blocks pilin Av. The sRNA promoter is conserved in all sequenced gonococcal strains, and mutations in the predicted transcript downstream of the G4 forming sequence do not alter pilin Av. A mutation that produces a stronger promoter or substitution of the pilE G4-associated sRNA promoter with a phage promoter (when the phage polymerase was expressed) produced wild-type levels of pilin Av. Altering the direction and orientation of the pilE G4-associated sRNA disrupted pilin Av. In addition, expression of the sRNA at a distal site on the gonococcal chromosome in the context of a promoter mutant did not support pilin Av. We conclude that the DNA containing the G-rich sequence can only form the G4 structure during transcription of this sRNA, thus providing a unique molecular step for the initiation of programmed recombination events.
| To evade the host immune response, pathogens have evolved mechanisms to provide genetic diversity in targets of immune surveillance. Organisms that express these diversification systems are under strong evolutionary pressure to provide subpopulations of preexisting variants and often rely on cellular recombination machinery to catalyze dedicated high-frequency reactions without disturbing genome integrity. Previously, we defined a guanine quartet (G4) structure in the strict human pathogen Neisseria gonorrhoeae that is required for initiating the homologous recombination reactions leading to pilin antigenic variation (Av). G4 structures have been implicated in many biological processes, however the mechanisms allowing their formation within a chromosome have not been elucidated. In this work, we show a direct link between transcription of a small RNA (sRNA) that initiates within the G4 structure forming sequence and pilin Av and conclude that the process of transcription is necessary for G4 structure formation. sRNAs have emerged as important regulatory molecules in both eukaryotes and prokaryotes, and this is a novel activity of a sRNA in a bacterium. We anticipate that the reliance of G4 structure formation on transcription is a mechanism used by other biological systems that rely on this alternative DNA structure.
| Neisseria gonorrhoeae is an obligate human pathogen and the causative agent of the sexually transmitted infection gonorrhea. Gonococci generally infect the urogenital tract and the infection typically presents as urethritis in men and cervicitis in women, but many women can be asymptomatic carriers [1]. The N. gonorrhoeae type IV pilus is essential for establishing infection [2]. Pili assist in epithelial adherence, gonococcal cell aggregation, and also mediate twitching motility [3]–[5]. Protective immunity never develops, partially because the bacterium can evade host immune selection by varying the expression of surface antigens including lipooligosaccharides, the opacity family of outer membrane proteins, and pili [6]–[10].
Pilin antigenic variation (Av) is a high frequency diversification system that operates via a specialized recombination pathway, and utilizes enzymes that participate in general recombination and repair pathways as well as enzymes that do not participate in either pathway [11]. N. gonorrhoeae possesses one pilin expression locus (pilE) and up to 19 silent pilin loci (pilS) residing in up to 6 discrete locations in the genome [12]. Pilin Av occurs as a result of non-reciprocal DNA recombination between any pilS copy and pilE, leading to the expression of a new variant protein [7]. These new variants can be fully functional, poorly expressed or not expressed, and when the pilus is not expressed or poorly expressed there is an obvious change in colony morphology on solid growth medium [13]. The difference between piliated and nonpiliated colony morphologies thus allows the ability of a strain to undergo pilin (Av) to be easily assayed [14].
A transposon-based genetic screen previously identified insertions in a DNA region upstream of pilE that blocked pilin Av, but did not alter pilin expression [15], [16]. A subsequent targeted genetic screen in this region identified 11 GC base pairs that when individually mutated completely blocked pilin Av, and a 12th GC base pair that when mutated retained residual pilin Av activity [11], [17]. Mutation of an adjacent 13th GC base pair had no effect on pilin Av but mutation of this GC base pair in addition to the 12th resulted in a complete loss of pilin Av, suggesting that the 13th GC base pair could partially substitute for the 12th [11], [17]. The organization of these 12 GC base pairs conforms to a guanine quadruplex or guanine quartet (G4) forming sequence [17]. Biophysical studies demonstrated that this G-rich sequence formed a G4 structure, and mutations that block pilin Av also inhibited structure formation [17]. Growth of N. gonorrhoeae on N-methyl mesoporphyrin IX, a compound that specifically binds G4 structures but not double or single stranded DNA [18], decreased the frequency of pilin Av [17]. Furthermore, point mutations in N. gonorrhoeae that block pilin Av and G4 structure formation prevented single stranded nicks from being detected in both the G4 forming sequence and complement strand [17]. The results from these studies suggested that for pilin Av to occur, the pilE G4 sequence must form a G4 structure [17]. In order for a G4 structure to form, duplex DNA must first be converted into single stranded DNA. In this study, we have now established that pilin Av absolutely requires transcription of a non-coding sRNA which originates within the pilE G4 sequence, thus providing a mechanism for the initiation of this conversion event.
The region upstream of pilE that contains the pilE G4 is devoid of predicted genes and open reading frames. However, we identified a putative promoter sequence adjacent to the pilE G4 [19] (Fig. 1A & B). The most conserved bases of the −10 element [20], all the bases in the −35 element, and the relative location of this putative promoter to the pilE G4 are conserved in all sequenced N. gonorrhoeae strains and Neisseria meningitidis strains, except N. meningitidis strain FAM18, which expresses a different pilin class that cannot antigenically vary and does not have a pilE G4 sequence [21], [22].
To test whether the putative pilE G4-associated promoter was functional in N. gonorrhoeae, 5′ Rapid Amplification of cDNA Ends (5′RACE) adapted for bacterial small RNAs, was used to identify transcripts associated with the pilE G4 [23]. 5′RACE analysis detected a transcript initiating within the pilE G4 forming sequence and three adjacent start sites were identified (Fig. 1B). The existence of this RNA was confirmed by RNAseq analysis of N. meningitidis strain alpha14 (N. Heidrich and J. Vogel, personal communication). Moreover, this RNAseq analysis showed that levels of the pilE G4-associated sRNA in N. meningitidis are 125× lower than the corresponding pilE mRNA transcript suggesting that this sRNA is non-abundant (N. Heidrich and J. Vogel, personal communication). We have been unable to reliably detect this sRNA molecule by Northern Blot using conditions that allow other sRNA molecules to be detected (data not shown).
To determine whether the pilE G4-associated sRNA is required for pilin Av, promoter loss-of-function mutants were constructed and analyzed by the kinetic pilus-dependent colony morphology phase variation assay, which is a surrogate measure for pilin Av [16] (Fig. 1C & D). Mutation of the −10 sequence or both the −10 and −35 promoter elements caused a complete block of pilin Av, suggesting that transcription of the pilE G4-associated sRNA is required for pilin Av (Fig. 1D). Point mutations in the −10 element, a 5 bp insertion between the −10 and −35 elements, or mutation of the −35 element showed decreased pilin Av, suggesting that the level of transcription correlates with the amount of pilin Av (Fig. 1D and Fig. S1A & B). We also tested whether pilin Av was enhanced by substitution of the −10 element with a stronger promoter that was closer to the consensus sequence, but there was no change in the levels of pilin Av [20] (Fig. 1B & D). 5′-RACE analysis of the −10/−35 promoter mutant resulted in a loss of the 5′RACE product (data not shown). Taken together, these results are consistent with a model where pilin Av requires transcription of a sRNA which originates in the pilE G4 sequence, but that transcriptional initiation is not a rate limiting step in this process. However, we cannot ascertain how much more the stronger promoter contributes to transcription in N. gonorrhoeae.
Not surprisingly, a 5 bp insertion upstream of the −35 element, which is not part of the sRNA promoter, had no effect on pilin Av. However 5, 10, and 46 bp insertions downstream of the G4 forming sequence, but within the predicted sRNA transcript, also had no effect on pilin Av (Fig. 2, and Fig. S1B, C & E). Furthermore, deletion of 5, 9, 10 and 32 bps [15], which are also potentially part of the sRNA transcript, had no effect on pilin Av (Fig. 2 and Fig. S1C, D & F). Since these insertions and deletions downstream of the pilE G4 have no effect on pilin Av, and there is no ribosome binding site or open reading frame within the sRNA sequence, we conclude that the active form of this sRNA is non-coding.
We further characterized the sRNA promoter using in vitro transcription with Escherichia coli sigma70 RNA polymerase (Fig. 1E). A 74 bp run-off transcript was produced from a 180 bp double stranded DNA template having the wild-type and stronger promoters, where the stronger promoter showed increased transcript levels (Fig. 1E). These results are consistent with a lack of a transcriptional stop sequence within the fragment. No transcript was observed when the −10 element and/or the −35 element were mutated or when 5 bps of heterologous sequence was inserted between the −10 and −35 sequences (Fig. 1E). Primer extension analysis of cDNA synthesized from in vitro transcribed RNA confirmed the products initiated from the wild-type and stronger promoters, but did not detect a product initiated from the mutated promoter (Fig. S2). These transcription profiles are consistent with the pilin Av defects observed in gonococci containing these mutations.
RecA, RecG, and RuvB are recombination factors required for pilin Av [16], [24], [25]. RecA is a DNA recombinase, RecG is a 3′ to 5′ helicase involved in branch migration of Holliday junctions and RuvB along with RuvA is a 5′ to 3′ helicase that is also involved in branch migration of Holliday junctions. Previously, the Holliday junction processing double mutant recG/ruvB was found to be synthetically lethal upon recA induction in gonococci [25]. This lethal phenotype was rescued by a pilE G4 mutation that does not allow formation of the G4 structure which demonstrated that formation of the G4 structure was upstream of RecG, RuvB, and RecA during pilin Av [17].
To determine whether transcription of the pilE G4-associated sRNA occurs before the action of RecG, RuvB, and RecA, a −10/−35 G4-associated sRNA promoter mutation was introduced into the recG/ruvB double mutant, recA inducible strain. Gonococcal colonies in the absence of RecA induction do not show a growth defect because the bacteria cannot undergo pilin Av (Fig. S3A). Upon RecA induction, the −10/−35 promoter mutation rescues the synthetically lethal phenotype of the recG/ruvB double mutant showing that transcription of the pilE G4-associated sRNA, like formation of the G4 structure, acts before RecG, RuvB and RecA during pilin Av (Fig. S3B).
To confirm that the pilin Av defects observed in the pilE G4-associated sRNA promoter mutants were a direct consequence of transcriptional defects at the pilE G4 locus and to test whether RNA polymerase was playing an active role in pilin Av, we expressed the pilE G4-associated sRNA with a T7 promoter and T7 RNA polymerase. First, the sRNA −10 promoter element was replaced with a minimal T7 promoter and tested for the ability to drive transcription using T7 RNA polymerase in vitro (Fig. 3A). Next, T7 RNA polymerase was inserted downstream of a lac promoter at the neisserial intergenic complementation site (NICS) [26] in a gonococcal strain where the pilE G4-associated sRNA −10 promoter element was replaced with a minimal T7 promoter (Fig. 3B). Colony morphologies examined after 46 hours of growth demonstrated that expression of T7 RNA polymerase alone in the parental strain (T7 Pol.) does not affect pilin Av, whereas replacement of the sRNA −10 promoter element with a minimal T7 promoter (T7 Prom.) produced an Av deficient (Avd) phenotype because colonies showed a stable piliated morphology after many generations (Fig. 3C). This Avd phenotype was restored to wild-type levels of pilin Av by induction of T7 RNA polymerase expression with IPTG (T7 Prom./NICS::T7 Pol.) (Fig. 3C). These observations were confirmed by the pilus-dependent kinetic colony phase variation assay (Fig. 3D). Since only T7 RNA polymerase can restore pilin Av in the sRNA minimal T7 promoter mutant, we conclude that expression of the pilE G4-associated non-coding sRNA is essential for pilin Av, but that the identity of the polymerase providing the transcription is not important.
Since transcription of the pilE G4-associated sRNA allows for melting of duplex DNA containing the pilE G4 sequence which then enables formation of the G4 structure, we reasoned that this sRNA might act in-cis. Therefore, we determined whether expression of the pilE G4-associated sRNA in-trans could complement a promoter mutant at the endogenous locus. The G4 sRNA region with a wild-type or stronger promoter was cloned and inserted at the ectopic NICS locus on the chromosome [26] (Fig. S4A). Kinetic pilus-dependent phase-variation assays demonstrated that expression of the sRNA at the ectopic locus does not complement a promoter mutant at the endogenous locus indicating that the pilE G4-associated sRNA acts in-cis and provides an essential step for the formation of the pilE G4 structure (Fig. S4B).
Because the pilE G4 sRNA acts in-cis, we determined whether the orientation and direction of the sRNA at the endogenous locus was also required for its function. Mutants with an inverted pilE sRNA, a pilE sRNA in reverse, and a pilE sRNA that is both in reversed and inverted were created (Fig. 4A&B). None of the mutants allowed pilin Av, indicating that both correct orientation and direction are required for the sRNA and G4 structure to function (Fig. 4C).
We have established a direct link between transcription and G4 structure formation in the process of pilin Av. We show that transcription of the cis-acting pilE G4-associated non-coding sRNA is absolutely required for N. gonorrhoeae pilin Av and is therefore required to initiate the homologous recombination reactions leading to pilin Av. From this work, we can propose a working model for the initiation of pilin Av (Fig. 5). We propose that initiation of pilin Av begins with transcription of the pilE G4-associated sRNA. Pilin Av was not enhanced by substitution of the −10 promoter element with a stronger promoter or a phage promoter, which suggests that pilin Av has a maximal level of efficiency that is not influenced by increased pilE G4 formation potential. However, since we have not been able to detect the sRNA directly, this conclusion is based on the assumption that these promoters function similarly in N. gonorrhoeae as they do in E. coli. It was surprising that that transcription of the sRNA by T7 polymerase in N. gonorrhoeae resulted in wild type levels of phase variation. This result rules out a direct role for RNA polymerase in G4 structure formation and suggests that it is the act of transcription and/or the sRNA product that are critical for G4 structure formation and pilin Av.
It is also possible that other processes involved in pilin Av downstream of transcription may be rate-limiting, thus negating any effect of increases in transcription rates. We reason that as transcription proceeds through the pilE G4 sequence, an RNA∶DNA hybrid could form between the sRNA and the C-rich strand leaving the G-rich strand unpaired to aid G4 structure formation, possibly aided by an as yet unknown protein (Fig. 5). How the formation of a G4 structure then initiates gene conversion is presently unknown, but the affinity for the pilE G4 structure by RecA, and the ability of the pilE G4 to stimulate RecA-mediated strand exchange [27], supports a model where the G4 structure acts to recruit recombination factors to the pilE region of the chromosome. In addition, it is likely that the G4 structure would cause a replication fork collapse when the replisome tries to synthesize past the G4 structure [11].
The lack of any observable pilin Av when the orientation and direction of both the sRNA and G4 structure were changed shows that this element does not act like enhancer elements found in eukaryotic cells. We did not observe any growth defect in strains where the G4-forming sequence was moved from the lagging strand onto the leading strand. This result suggests that either the structure doesn't form often enough to produce an observable reduction in viability when the G4 structure interrupts replication, or there are proteins that can remove the G4 structure to prevent a replication stop. The RecQ helicase is a likely candidate for an enzyme that could remove the G4 structure to prevent a replication blockage (Laty A. Cahoon, Kelly A. Manthei, Ella Rotman, James L. Keck and H. Steven Seifert, unpublished).
Regulatory cis and trans-encoded small RNAs have been identified in bacteria, but the pilE G4-associated sRNA does not fit either the cis or trans-encoded sRNA paradigm established for bacteria. Cis-encoded sRNAs are transcribed antisense to the genes they regulate whereas trans-encoded sRNA share limited complementarity. Once base paired with their target mRNA transcript, these regulatory sRNAs can influence the stability of, or modulate translation of, their target [28], [29]. The protein factor Hfq generally binds trans-encoded sRNAs and mediates base-pairing of the sRNAs to their target mRNAs; sRNAs that are not bound to Hfq are extremely labile [30]. Generally, cis-encoded antisense RNAs (asRNA) do not require Hfq [31]. In N. gonorrhoeae, Hfq is an RNA chaperone that functions as a pleiotropic regulator of RNA metabolism [32]. An hfq mutant was found to decrease the amount of pilE protein by an indirect mechanism that has yet to be determined. Since the pilE G4-associated sRNA is not an asRNA or a trans-encoded RNA and does not share complementarity to any known mRNA target, it is unlikely that this sRNA binds Hfq, but this has not been experimentally addressed. Some sRNAs have also been shown to bind and regulate proteins, such as the E. coli 6S RNA that interacts with RNA polymerase [29].
The pilE G4-associated cis-acting sRNA is transcribed in a region devoid of open reading frames and is not complementary to any expressed gene. In addition, the sRNA is most likely non-coding since mutations (5–46 bp insertions or 5–32 bp deletions) just downstream of the G-rich region have no effect on pilin Av (Fig. 2 and Fig. S1). It is possible that this sRNA performs other functions in this bacterium and could be processed or form a structure on its own. While we cannot rule out the possibility that the pilE G4-associated sRNA regulates other molecules, it is likely that the unwinding of the duplex DNA containing the pilE G4 sequence during transcription, and formation of the RNA∶DNA hybrid with the C-rich strand, allows formation of the G4 structure which is required for pilin Av (Fig. S3). It is interesting to note that two other predicted G4 forming sequences in the gonococcal genome also have putative promoters in a similar relative location. Formation of a G4 structure via transcription and occlusion of the C-rich stand has also been previously proposed to act during immunoglobulin class switching [33], [34]. Since both pilin Av and immunoglobulin class switching are high frequency recombination systems, there may exist other diversification systems that use transcriptionally induced alternative DNA structures to initiate programmed recombination reactions, and these may represent a specialized class of transcription-associated recombination systems [35].
While G4 DNA structures have been implicated in many biological processes in eukaryotic cells, less is known about their function in prokaryotic cells. In eukaryotes, G4 structures have been implicated in telomere metabolism [36], the regulation of genes required for normal cell growth and differentiation [37], potentiating genomic instability [38], the regulation of RNA transcription and translation [39], and mediating immunoglobulin gene diversification [33]. Proteins that bind or resolve G4 structures in both eukaryotes and prokaryotes have also been identified [40]–[42]. In prokaryotes, potential G4 forming sequences have been identified by bioinformatics, but functional analysis of these sequences has yet to be reported [43]. Our previous report demonstrating that the pilE G4 is essential for pilin Av in N. gonorrhoeae was the first study to show the requirement of a G4 structure for any biological process in a bacterial cell [17]. This work develops the mechanism of action further by showing that transcription of a cis-acting sRNA is critical for G4 structure formation to direct this specialized diversity generation system.
E. coli One Shot TOP10 competent cells (Invitrogen) were grown in Luria-Bertani (LB) broth or on solid media containing 15 g/L agar at 37°C and used to propagate plasmids. E. coli selected for plasmids containing kanamycin or erythromycin resistance were selected on media containing 50 or 100 µg/ml of the respective antibiotic. Gonococcal strains were grown on GC Medium Base (Difco) plus Kellogg supplements (GCB) [22.2 mM glucose, 0.68 mM glutamine, 0.45 mM co-carboxylase, 1.23 mM Fe(NO3)3; all from Sigma] at 37°C in 5% CO2, when applicable 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) was added for induction. Gonococcal transformants were selected on media containing 50 µg/ml kanamycin or 2 µg/ml erythromycin.
Gonococci were revived from frozen stocks, then after 24 hours of growth a single piliated colony was passaged onto solid media to obtain single isolated colonies. Colony variation was scored every 2 hours after 18 hours of growth up to 28 hours by observing the number of non-piliatied (P-) sectors arising over time. 10 colonies per strain were scored per assay. A colony that showed no P- sectors was given a score of 0, one P- sector was given a score of 1 and so forth. Colonies with 4 or more P- secors were given a score of 4.
A plasmid containing the region between USS2-pilArev2 from FA1090 (1-81-S2) pilE::mTn#9 recA6 [17] was digested at PacI and Mlu1 sites. Annealed linkers with PacI (TopA/BotA), an overlapping overhang and MluI ends for the −10 promoter mutant (10KOTop/10KOBot), stronger −10 mutant (10strTop/10strBot), T7 promoter mutant (T7G4PTop/T7G4PBot), or PacI and MluI ends for the 5 bp insertion between the −10 and −35 mutant (5in-10-35Top/5in-10-35Bot), were ligated to the PacI/MluI digested plasmid and transformed into TOP10 competent cells (Table S1). The −35 promoter mutant was created by digesting the USS2-pilArev2 plasmid (above) at MluI and EcoRV sites. Then the −35KO linker (35KOTop/35KOBot) was ligated to the digested plasmid. To increase the amount of homologous DNA available for recombination, a plasmid containing the −35 mutation was PCR amplified with GCUUSS2 and a −35 lengthening primer using KOD (Novagen) and cloned into pCRBluntII-Topo (Table S1). To create the −10 and −35 promoter mutant, a plasmid containing the −35 promoter mutation was digested with PacI and MluI. Then, the linker used to create the −10 promoter mutant was ligated to the digested plasmid (Table S1). Then gonococcal strain FA1090 (variant 1-81-S2) recA6 was transformed as described previously [44] and selected for kanamycin resistance. Kinetic phase variation assay was performed as described [16].
The USS2-PilArev2 plasmid was digested with PacI and MluI. Annealed linkers with MluI (G4Bot/G4Top), an overlapping overhang and PacI ends for the 10 bp insertion or deletion downstream of the G4 (in10bpDTop/in10bpDBot and del10bpDTop/del10bpDBot, respectively), the 5 bp insertion or deletion downstream of the G4 (in5bpDTop/in5bpDBot and del5bpDTop/del5bpDBot, respectively), were ligated to the PacI/MluI digested plasmid (Table S1). The 5 bp insertion upstream of the sRNA promoter mutant was created by digestion of the USS2-pilArev2 plasmid at unique MluI and EcoRV sites follow by ligation to the linker (in5bpUTop/in5bpUBot) (Table S1). Then gonococcal strain FA1090 (1-81-S2) recA6 was transformed as described [44] and selected for kanamycin resistance. Other mutants (Fig. 2, the 9 bp deletion and 46 bp insertion downstream of the G4) were identified during screens for the desired mutations listed above. Point mutants (Fig. 2) were identified in a previous screen [17] and backcrossed into the parental strain and selected for kanamycin resistance.
For E. coli RNA polymerase (holoenzyme, sigma saturated, Epicentre) in vitro transcription reactions, template DNA was PCR generated using KOD, gonococcal DNA having the desired pilE G4 sRNA region mutation, and primers RS1For and pilARevnested2Bot (Table S1). Then using 100 ng of purified template, in vitro transcription reactions were performed as specified by the manufacture. After 1–2 hours at 37°C, reactions were DNase I treated and run on a 14% polyacrylamide denaturing gel (UreaGel System, National Diagnostics). 0.5 µl of the above reactions were used for primer extension analysis, cDNA was synthesized using 3 pmol of 5′FAM-labeled PEx1 primer and Superscript III, EtOH precipitated, and sent for fragment analysis (Northwestern Genomics Core Facility). For T7 RNA polymerase in vitro transcription reactions, template DNA was PCR generated using KOD, gonococcal DNA having the replacement T7 promoter at the pilE G4-associated sRNA locus, and primers RS1For and T7G4invitro (Table S1). Then using 100 ng of purified template, in vitro transcription reactions were performed (Maxiscript Kit, Ambion).
Creation of the pilE G4-associated sRNA T7 promoter replacement mutant is detailed above. To create the IPTG inducible T7 polymerase strain, we used KOD and primers PacIT7F engineered to have a Pac I site and T7R to amplify T7 polymerase from BL21 (DE3) E. coli (Table S1). This PCR product was digested with PacI and cloned into the PacI and PmeI digested pGCC4 vector [26]. This construct was allowed to recombine into strain FA1090 (1-81-S2) at the NICS by liquid transformation and was selected for on media containing erythromycin [26], [45], [46]. Then chromosomal DNA from this strain was transformed into FA1090 (1-81-S2) recA6, and the pilE G4-associated sRNA T7 promoter replacement mutant. These mutants were then assayed for pilus-dependent colony phase variation [16].
The −10 and −35 pilE G4-associated sRNA promoter mutant plasmid detailed above was allowed to recombine into the Holliday junction processing mutant by liquid transformation [25], [46]. After limited exposure to IPTG, colonies were selected for kanamycin resistance, tested for erythromycin resistance associated with the ruvB functional deletion and chloramphenicol associated with the recG functional deletion. Gonococcal transformants containing the −10 and −35 sRNA promoter mutations were verified by sequencing analysis.
The region from the edge of the transposon insertion in RS1 which does not affect pilin Av [15] to 25 bps upstream of the pilE G4-associated sRNA −35 promoter element were PCR amplified using KOD and primers G4compR1PacI engineered to have a Pac I site and RS1For (Table S1). This PCR product was digested with PacI and cloned into the digested PacI and PmeI pGCC4 vector. This construct was allowed to recombine into the pilE G4-assoicated sRNA −10 promoter mutant strain at the NICS as described previously [44] and selected for erythromycin resistance. Transformants were verified by sequencing both the ectopic and endogenous pilE G4-assoiciated sRNA site. These mutants were then assayed for pilus-dependent colony phase variation [16].
The sRNA inverted, reverse, reverse-inverted and respective −10 and −35 promoter element mutants were created by ligating a PacI digested 433 bp synthesized dsDNA fragment (IDT) to the PacI/EcoRV digested USS2-pilArev2 plasmid (Table S1). Then gonococcal strain FA1090 (1-81-S2) recA6 was transformed as described [44] and selected for kanamycin resistance.
|
10.1371/journal.pcbi.1003883 | Stochastic Modeling of Mouse Motor Activity under Deep Brain Stimulation: The Extraction of Arousal Information | In the present paper, we quantify, with a rigorous approach, the nature of motor activity in response to Deep Brain Stimulation (DBS), in the mouse. DBS is currently being used in the treatment of a broad range of diseases, but its underlying principles are still unclear. Because mouse movement involves rapidly repeated starting and stopping, one must statistically verify that the movement at a given stimulation time was not just coincidental, endogenously-driven movement. Moreover, the amount of activity changes significantly over the circadian rhythm, and hence the means, variances and autocorrelations are all time varying. A new methodology is presented. For example, to discern what is and what is not impacted by stimulation, velocity is classified (in a time-evolving manner) as being zero-, one- and two-dimensional movement. The most important conclusions of the paper are: (1) (DBS) stimulation is proven to be truly effective; (2) it is two-dimensional (2-D) movement that strongly differs between light and dark and responds to stimulation; and, (3) stimulation in the light initiates a manner of movement, 2-D movement, that is more commonly seen in the (non-stimulated) dark. Based upon these conclusions, it is conjectured that the above patterns of 2-D movement could be a straightforward, easy to calculate correlate of arousal. The above conclusions will aid in the systematic evaluation and understanding of how DBS in CNS arousal pathways leads to the activation of behavior.
| Brainstem and thalamic regulation of arousal has been studied experimentally since the mid 20-th century. Today, Deep Brain Stimulation (DBS) is used in the treatment of movement disorders, chronic pain, clinical depression, amongst others. At present, the proper choice of DBS parameters (frequency and strength of the electric stimulation), and how those parameters should be modified as conditions change, are not well understood. In this work, using motor activity as the observed response, a statistical framework is developed for such study, and a quantitative relationship between parameter values and response is established. Within this framework, a possible correlate of arousal, the rapid onset of spatial (two-dimensional) movement, is uncovered and also studied. One long-term hope for techniques such as DBS are that they could assist in the treatment of disorders of consciousness, by supplementing or replacing (e.g., in Traumatic Brain Injury) what should ordinarily be the appropriate endogenous stimulation.
| Deep Brain Stimulation is currently being used in the treatment of Parkinson’s Disease, Disorders of Consciousness (DoC) and clinical depression [1–3]. The possibility that Deep Brain Stimulation (DBS) could be used to enhance brain arousal is a subject of immense interest, e.g., in traumatic brain injury (TBI). For example, in a human patient who had suffered DoC for more than seven years, DBS of the central thalamus was used successfully to aid in the recovery of his consciousness (Schiff et al (2007)) [4], Schiff (2010) [5]. From a fundamental neuroscientific point of view, this has been conceptualized as an elevation of generalized CNS arousal (Pfaff, 2006) [6]. In the mouse, locomotion is the most elementary of behavioral responses, and in this work we utilize such movement patterns as the basis of our inference relating DBS parameter changes to behavioral effects (Leshner and Pfaff (2011) [7], Benjamini et al (2011) [8], Quinkert et al (2010, 2011, 2012) [9–11]).
In the present study, the stimulations occur in the central thalamus (cf.Schiff et al [4]), over fixed 10 min intervals, every three hours, eight per day (four in light, four in dark). There are both stimulated and control (electrodes implanted but nonstimulated) mice. What is observed are the x- and y-coordinates of location, per second, over three days, with there being 12 hours of light, 12 hours of dark. The range of DBS parameters consist of three amperages and four frequencies, and were applied to each mouse over the three days (reported by Quinkert et al [9]). A common motor activity summary statistic is Total Activity, i.e., the total distance traveled over some fixed time interval (e.g., 10 min (600 sec)), or equivalently, Mean Activity (or Mean Speed): Total Activity divided by the number of time points (e.g., 600). Two questions that arise in the use of such statistics are, first, is there is a loss of important information in such Mean (or Total) Activity summarization; for instance, are important angular changes in direction or differences in the spatial range of movement, lost? Secondly, how does one calculate a standard error (or make a probabilistic assessment) for any statistic derived from such time-varying location data, doing so in a manner that can be justified? One needs to appropriately account for both local time-correlations, as well as broader circadian changes, otherwise the calculations may not be representative.
In a control mouse, or a stimulated mouse if there were no DBS effect, our basic assumption is that the time-varying processes constructed from the motor activity, are piecewise stationary. We will verify and apply piecewise stationarity for division into 3-hr segments, although longer periods could also be justified. We show that the autocorrelations within such a stationary segment die out after 20 minutes; segments of motor activity separated by 20 min can be assumed to be uncorrelated (or, in the present context, it is reasonable to assume independence). In the case of the stimulated mouse, we show that autocorrelations also die out after 20 min for regions sufficiently separated from a stimulation interval. The first stage of the analysis is to show that the stimulation is effective for at least one of the DBS parameter values. In this analysis, all calculations are on segments separated by 20 min; that is, our statistics for each animal are based on the ±80 minutes, centered at each stimulation time, leaving at least 20 min between any of the time intervals on which calculations are to be made. The statistic calculated on each (independent) segment results in a null hypothesis of no effect. A False Discovery Rate (FDR) thresholding then establishes that there is at least some DBS effect.
Once such a DBS effect is established, piecewise stationarity is used in a slightly different manner, since stationarity is being briefly perturbed by the stimulations to new steady-states, with possibly different stationary mean and/or the covariance structure for the 10 min stimulation intervals, than for nearby non-stimulated time intervals. That is, after establishing that there is some DBS effect, the determination of the relationship of DBS parameters to motor activity will need to be based on the collection of 10 min stimulation intervals, with the means and/or covariance structures for distinct such intervals possibly varying with the DBS parameter values. Differences in the stationarity structure for the different 10 min stimulation intervals, as we will see, result in significant differences in the variances for the calculated statistics, and hence does not allow for a traditional ANOVA formulation or nonparametric method. We will also consider differences in Mean Activity between light and dark, and piecewise stationarity will be the basis of that model.
As part of our analysis, we analyze speed and angular patterns and also decompose the movement into its randomly occurring epochs of 0-, 1-, and 2-dim movement. One-dimensional (linear) movement appears to be not unlike background noise in the present context, which is why the decomposition was formulated. The main ideas of the present paper important to neuroscience are: (Hypothesis I) The stimulation is effective; (Hypothesis II) It is 2-D movement, not 1-D movement, that occurs in response to stimulation, and there is a detectable relationship between DBS amplitude and frequency and the resulting movement; and, (Hypothesis III) It is 2-D movement, not 1-D movement, that differs between light and dark; and, stimulation in the light initiates a manner of movement (2-D movement) more commonly seen in the (non-stimulated) dark. To address these three hypotheses, we consider seven motor processes and their resulting statistics. For each, we calculate a forward 10 min mean (i.e., forward in time from each given sec) and a right minus left difference in 2 min means (at each sec); the later can detect rapid changes. The focus of several (3–4) of the statistics (those that are one-dimensional) is not on their power of information extraction, but rather the opposite; these statistics have basically no power and should be subtracted off, in order to obtain better statistics (those that are purely two-dimensional).
All animal procedures were in compliance with National Institutes of Health guidelines and approved by the Rockefeller University Institutional Animal Care and Use Committee.
Details of methodology regarding neurosurgery and behavior have been published (Quinkert et al [9]). Briefly, mice were singly housed with food and water available and were subjected to a 12 h light/dark cycle. Stainless steel monopolar electrodes (0.3 mm diameter (Plastics One)) were insulated using polymide with 0.2 mm stripped from the electrode tips, with electrodes implanted bilaterally in the central thalamus. Stimulation was programmed and delivered by a four-channel stimulus generator (Multichannel systems STG2004). Stimulation epochs lasted for 10 min and occurred every 3 h over the course of 3 days. All stimulations were biphasic with a pulse width of 0.1 ms on both anodic and cathodic phases of the pulse. Three pulse amplitudes were applied, each for one day: 75, 100 and 125 μA. Four pulse frequencies were selected from 50, 125, 175 and 225 Hz. The mice were euthanized and then a histological assessment of electrode placement was made, following data collection. Here, data are analyzed from novel points of view.
Nine mice were studied, with electrodes implanted in each; stimulation was applied in six mice (Mice 1–6), with three others used as controls (without stimulation, Controls 1–3). Four of the six stimulated mice (Mice 1–4) showed significant responses to the stimulations, whereas Mice 5–6 did not appear to respond at all; it was established via histological investigation that differences in electrode placement explained the nonresponsiveness of Mice 5–6. They were included in the initial analysis as a matter of completeness. The number of stimulations (Nstim) was 8 × 3 = 24 for mice 2, 4, 5 and 6, and was 23 for mice 1 and 3 (the recording of the 24-th stimulation period was not complete). Home cage activity data was collected by a 3D infrared monitor (Accuscan Instruments), which records the locations at the times of a detected change. The times between changes varied from the millisecond scale to that of multiple seconds; the data were interpolated to the one second scale, for computational purposes.
In the present mouse experiments, what is observed are the time-evolving locations (x- and y-coordinates) of the animal. The stimulation data is observed over 3 days, with N = 3*24*3600 = 259200 seconds. The stimulation times are denoted as {Sk, k = 1, …, Nstim}, with Nstim being the number of stimulations. From the location data r(ti = (x(ti), y(ti)), ti = 1, …, N, one can calculate the (discrete-time) Velocity, Speed and Angle Direction processes:
r _ ( t i ) = ( x ( t i ) , y ( t i ) ) V _ ( t i ) = ( V x ( t i ) , V y ( t i ) ) = r _ ( t i + 1 ) − r _ ( t i ) ,, t i = 0 , … , N − 1 M ( t i ) = x 2 ( t i ) + y 2 ( t i ) , U _ x = ( 1 , 0 ) , θ ( t i ) = angle ( V _ ( t i ) , U _ x ) (1)
The angle θ(ti) is defined with respect to the positive x-axis (Ux) and is uniquely defined in [−π, π), with −π identified with π (i.e., the angles are on the unit circle). Since the location data are fully recoverable from the latter two and the initial location:
r _ ( t i ) = ∑ j = 0 i − 1 M ( t j ) × ( c o s ( θ ( t j ) ) , s i n ( θ ( t j ) ) ), (2)
the Magnitude and the Direction of Angle processes are the basis for our analysis. Hence, we model the changing patterns of the following motor activity processes:
M(ti): Speed, or Magnitude of the velocity, and by summing over any time interval one can obtain the Total Activity (i.e., Total Distance).
θ(ti): Angle of direction as a function of time (relative to the fixed x-axis). Moreover, one can decompose these angles into two groups:
θ(P)(ti): those that are Multiples of π/2 (e.g., continuation in same direction, a reversal or a perpendicular move), including movement parallel to the walls; and
θ(NP)(ti): those that are Non-Multiples of π/2 (here, e.g., movement into the interior, non-parallel to a wall);
Movement Pattern over a w = 30 sec window, calculated in a forward direction starting at each second, and decomposed into three groups (uniquely defined at each time point): (D(0)(ti) = 0, for zero-dimensional movement)
D(1)(ti): length of line segment containing the 1-dim movement (if points in time all lie on a line, and are not constant). One can determine if the movement over the 30 sec window beginning at ti, is one-dimensional, and if so, to calculate the length of its 1-D domain.
D(2)(ti): area of the two-dimensional convex hull generated by the points (if they do not lie on a line). Because the two-dimensional path can, and often does, cross itself, defining the 2-D domain is not straightforward or necessarily well defined. A natural definition is the use of the convex hull. The convex hull of the points (x(ti+r), y(ti+r), r = 0, 1, …, w − 1), over a moving window of length w = 30 seconds, is determined and its two-dimensional area D(2)(ti) is calculated.
ID(ti): identifies at each time whether the movement is zero-, one- or two-dimensional, by 0, 1 or 2 (based upon (i) and (ii), above).
The Speed M(ti), when non-zero, can be decomposed into the Speed at times at which the movement is two-dimensional or one-dimensional: M(ti) = M2D(ti) + M1D(ti)
M2D(ti) = M(ti) if ID(ti) = 2, and 0 otherwise;
M1D(ti) = M(ti) if ID(ti) = 1, and 0 otherwise;
One can define A(ti) to be the Total Activity over the 10 min window (w1 = 600) starting at ti (i.e., the sum of M(tj) over the interval). For simplicity, we will use the Mean Activity A ¯ ( t i ) rather than Total Activity, in that the derivation of standard errors is more direct. The difference is merely one of scale. Based upon (d) above, summing over M2D(ti) and M1D(ti), respectively, A ¯ ( t i ) can be decomposed into the sum of 2-Dim and 1-Dim Mean Activity: (zero-dim movement adds zero)
A ¯ ( t i ) = 1 w 1 × ∑ j = 0 w 1 − 1 M ( t i + j )= 1 w 1 × ∑ j = 0 w 1 − 1 M 2 D ( t i + j ) + 1 w 1 × ∑ j = 0 w 1 − 1 M 1 D ( t i + j ) = A ¯ 2 D ( t i ) + A ¯ 1 D ( t i ) (3)
As we will see, it is the 2-Dim activity that differs most significantly between light and dark, as well as that which predominates in response to stimulation. That is, we will show that it is the statistics that extract 2-Dim information that are informative concerning responsiveness to DBS stimulation, and not those that are 1-dimensional.
In Figs. 1–2, there are two panels (A and B) in each, one for high (A) and low (B) parameter values; Fig. 1 is during the dark, Fig. 2 is during light, and they are for Stimulated Mice 1–2, respectively. In the first row of each panel are displayed the actual time-varying two-dimensional (x- and y-) position (per sec), over a sequence of four 10 minute intervals, starting 20 min prior to a stimulation, and including the 10 min stimulation interval and the 10 min interval following it. In the second and third rows are the individual x- and y-coordinate patterns, from which one can infer movement parallel or non-parallel to a wall. In the remaining rows are the above-described time-evolving motor processes, appropriately plotted; in row 4 are the time changing patterns of D(1)(∙) (length, red) and D(2)(∙) (area, blue); in row 5 are the analogous plots for Speed: M1D(∙) (1D, red) and M2D(∙) (2D, blue); and, in row 6, are the time-evolving θ(P)(∙) (multiple of pi/2, green) and θ(NP)(∙) (non-multiple of pi/2, blue). The seventh motor process, Mean Activity M(∙), being the sum of M1D(∙) and M2D(∙), was not plotted, for simplicity. The figures very much depict the motivation for the 1D and 2D decompositions, and their use in quantifying the stimulation response.
We will utilize two basic statistical calculations, defined below in expression (4): (1) a forward mean over a moving window of time (forward meaning that the value at time ti is for the window starting at ti; the window width being w1 = 600 sec); and (2) a difference in means (i.e., that to the right minus that to left), for a moving window of time (window width being w2 = 120 sec, on each side). The latter statistic acts as a high-pass filter, detecting the rapid onset of movement. We denote the two statistics, respectively, as X¯ ( ∙ ) and X¯ R L ( ∙ ):
X¯ ( t i ) = 1 w 1 × ∑ j = 0 w 1 - 1 X ( t i + j ) and X¯ R L ( t i ) = 1 w 2 × ( ∑ j = 0 w 2 - 1 X ( t i + j ) - ∑ j = - w 2 - 1 X ( t i + j ) ) (4)
where X generically represents any of the following seven motor activity processes:
A 1 D ( t i ) , D ( 1 ) ( t i ) , θ ( P ) ( t i ) , θ ( N P ) ( t i ) , A ( t i ) , A 2 D ( t i ) , D ( 2 ) ( t i ) (5)
The first three are 1-Dim statistics (1-D Activity, 1-Dim Length, Multiple of pi/2 angle). The next two are 2-dimensional Non-Multiple of pi/2 angle, Total (1-D and 2-D, combined) Activity, but are incomplete in certain ways (Results). The final two 2-dimensional statistics (2-D Activity, 2-Dim Area) are those that are of greatest potential interest.
In Fig. 3, for Stimulated Mouse 1, the time-varying (per sec) statistics were averaged over a moving 10 min window starting at each second, moving second by second; the statistics were: mean activity, fraction of non pi/2 and fraction of pi/2 angles, mean area (of 2-D movement) and the mean length (of 1-D movement). The red asterisks denote the stimulation times, and are plotted at the height of the recorded response at the stimulation time to accentuate the magnitude (or lack of) in response to the stimulus. For the angle processes, the statistics calculated are fractions, rather than means.
Let X(ti) generically represent any of the following seven Motor Activity Processes given in expression (5), above. Our basic assumption is that, in the non-stimulated animal, or in the stimulated if there were no DBS effect (Hypothesis I), there is piecewise stationarity (in time) for motor processes under consideration, calculated from the location data r(ti = (x(ti), y(ti)), ti = 1, …, N. Below, we show such piecewise stationarity for the speed (M(ti)) process, but others (e.g., the angle process) can similarly be shown. Specifically, we assume that time can be decomposed into a finite set of time segments, for which the process of interest is assumed to be representable as a strictly stationary process on each segment. On different time segments, the structural parameters of stationarity (mean, variance, autocovariances, spectral density) are allowed to differ. We justify below the use of a decomposition into 3 hr time segments (although longer segments could be justified). Once it is shown that there is (some) DBS effect, piecewise stationarity will no longer hold in the same form, in that the local stationarity at the stimulation intervals is being perturbed by the stimulations to new brief steady-states on these 10 min stimulation intervals, or for possibly longer (as it returns to its original steady-state). In testing Hypothesis II, only the 10 min stimulation intervals will be used (they are now stationary at the perturbed stead-state, potentially different for different stimulation levels); in testing Hypothesis III, the stimulation intervals will be excluded. In modeling piecewise stationarity, as in stationarity, one can model from either the time- or frequency-domain. In the present work, we have chosen to use a time-domain approach, primarily because it made the analysis for the three hypotheses, as a whole, more unified. There has been a great deal of work on modeling time-varying spectra (see Ombao et al (2001) [12], Huang et al (2004) [13]).
We first establish the overall pattern of changing stationarity. In Fig. 4, there are four panels (A–D). In A, left, for a single control mouse, the local mean is calculated for the speed (M(ti)) process, over a moving window of width 10 min (blue), 1 hr (green) and 3hr (red); on the right, are the time-varying means for all three controls, over a moving window of width 3 hrs. Mouse movement typically consists in random bursts of movement, of random lengths, interspersed with low or no movement periods, of random lengths. One can view such patterns as being doubly stochastic, with the first level describing whether there is or is not movement. The mean patterns reflect such behavior, where the values locally can be quite variable. In B, there are two rows. The first displays the sample autocovariance functions (over 1 hr = 3600 sec)) for all three controls, starting at the 4th hr and at the 6th hr, in light (left) and dark (right), six functions in total. The second displays the same, calculated at the 8th and 10th hrs. The 4 and 6 hrs represent early behavior in the 12 hrs, whereas the 8 and 10 hrs represent late behavior for the 12 hrs. Panel C displays the two autocorrelation functions calculated from the mean of the, respective, early (blue) and late (red) autocovariance functions. The horizontal bands (± 1 . 96 / 3600 × 6), which at a single lag is a confidence band about zero, are a standard time series practice to assess a “loss of correlation in time.” The result of Panels A–C of Fig. 4 is that, for the control mouse, the means are relatively stable during light and during dark, differing for the two; the same is true for the autocovariance structure, being relatively stable during light and during dark, but differing for the two. The autocorrelation patterns indicate that it is reasonable to assume that the statistics calculated on intervals that are separated by 20 min or more can be assumed to be uncorrelated (or, as we will, as independent). In the bottom panel (D) of Fig. 4, the autocovariances and autocorrelations are examined, for the stimulated mice 1–4, at times in between the broad range of stimulation intervals that are separated by 3 hrs. The autocovariances were calculated for 1 hr, starting 80 min after a stimulation interval. Again, one sees the same decay to negligible levels after 20 min, and hence, even in the stimulated case, intervals sufficiently separated can be assumed to be uncorrelated (again, in the present case, we will assume independence). Specifically, in Hypothesis II, where statistics are only calculated on the 10 min stimulation intervals, separated by 3 hrs, it is reasonable to assume uncorrelation (or, again, independence). Moreover, in Hypothesis III, in comparing 1D and 2D Mean Activity between light and dark, we will make calculations on three 3hr segments, separated by 1 hr, within light and within dark each, excluding the stimulation intervals and an additional 20 min following the stimulation. Hypothesis III concerns whether the means of the two (light, dark) are also different, in addition to their autocovariances.
In Fig. 5, top row are speed (per sec) data for a Control mouse (hence, non-stimulated). The left column displays 12-hrs in light and the right column 12-hrs in dark. In the second row are sample autocovariance functions of the top row data (per sec), calculated over 3-hr periods, starting at hrs 4, 5, 6 and 7. The most common approach to testing stationarity, has been spectral (squared modulus of the Fourier Transform), using the cumulative periodograms. The third row displays the estimated log spectral densities (using Thomson’s multitapering in the construction), for the four 3-hr time intervals. From the auto covariance functions, a dark versus light comparison, shows variances that differ (by a factor of more than five). However, within light and within dark, individually, the four variances show virtually no difference. Because of this, within light and within dark, we normalize the cumulative periodograms to a maximum of one (i.e., dividing by the variance). To test the equivalence of the four spectra, the Diggle-Fisher test (1991) [14] was performed, which is as follows. If the different sample spectra were all estimating the same true spectra, then shuffling the four spectral values at a given frequency, should statistically produce equivalent estimates. This is the basis of the test. One calculates the maximum difference at each frequency in the cumulative periodograms, and then the maximum of those over all frequencies. This is done for the actual observed cumulative periodograms and compared to the results for all the shuffling. The resulting histogram of the maximum difference (over all frequencies) of the shuffled cumulative periodograms, over 1000 shuffles, are displayed in the fourth row. The red asterisks denote the observed spectral maximum differences. The P-values for testing the hypothesis of stationarity during light is .54 and during dark is .75. Various other tests of stationarity have been proposed (e.g., Priestley and Subba Rao (1969)) [15], often as analogues of a Kolmogorov-Smirnov like test, which have proven difficult to use in practice.
Our basic assumption is that, over a 3-hr period, there exists a strictly stationary process that describes the chosen motor process (e.g., the speed or angle processes), and that it satisfies a mixing condition, specifically ϕ-mixing (see Billingsley (1968)) [16], i.e., that there exists a function ϕ(∙), such that limn→+∞ϕ(n)=0, and for any two events, F1 and F2, F2 dependent upon the information up to time m, F1 dependent upon the information up to time m+n, n ≥ 0 and P(F2) > 0, they satisfy ∣P(F1∣F2)−P(F1)∣ ≤ ϕ(n). We also assume that its autocovariances are absolutely summable (and hence a continuous spectral density exits). Phi(ϕ)-mixing is a very weak assumption which describes the rate at which time-dependency dies out. From Figs. 4–5, this is a very reasonable assumption of the dying out of the dependencies. Since X¯(∙) and X¯ R L ( ∙ ) are finite linear filters of X(∙), they also are strictly stationary and satisfy the same (form of) mixing condition. Let FX¯ ( t j ) denote the marginal distribution of X¯ ( ∙ ) at a (arbitrary) single time point tj, which by strict stationarity is the same at all times tj (in the segment of stationarity). Under the assumption of ϕ-mixing, the empirical distribution function Fn(∙): F n ( x ) = 1 n ∑ j = S k - 80 * 60 S k + 80 * 60 I [ X¯ ( t j ) ≤ x ] , where n is the number of terms in the sum, is asymptotically equivalent to FX¯ ( ∙ ) (see Billingsley (1968)) [16]. Specifically, there is uniform weak convergence, with s u p x n ( F n ( x ) − FX¯ ( x ) ) converging to a Gaussian process (indexed by x). Hence, we have that limn→+∞Fn(x)=FX¯(x), uniformly in x. As a consequence, probability calculations under Fn are asymptotically equivalent to those under FX¯.
Since Sk is a fixed time (the k-th stimulation time onset), if there were no effect due to the stimulation at time Sk, then X¯ ( S k ) is random with the probability distribution FX¯. Evidence against the hypothesis that there is no effect due to the stimulation at time Sk, can hence be measured by the probability of observing a value greater than or equal to X¯ ( S k ) under Fn, i.e., a P-value for each k, k = 1, …, Nstim. This comparison can be interpreted as a permutation test except that the permutations are restricted to being translations (or rotations if viewed on a circle). This restriction is a direct consequence of the test adhering, as is necessary, to the piecewise stationarity. The same results apply to the empirical distribution function for X¯ R L. Thus, for Hypothesis I, the question is (where X¯ represents any of the seven motor processes): Is X¯ ( S k ), for a fixed k, k = 1, …, Nstim, significantly greater than most X¯ ( t j ), for tj in the ±80 min (160 min) period centered at Sk? If so, this is (probabilistic) evidence that the stimulation caused a change in motor activity. The same question applies to X¯ R L ( S k ). In each case, one calculates the proportion of the values that lie above X¯ ( S k ) and X¯ R L ( S k ), respectively. Representative P-values (k = 1, …, NStim) are presented in Fig. 6 for Mouse 1 and Control 1. If one wishes to make an assessment of the effect due to a given amperage level (i.e., the level over a given day), one can apply a False Discovery Rate (FDR) procedure (presented in Results) (Benjamini and Hochberg (1995) [17], Benjamini and Yekutieli (2001) [18]).
In the case of Hypothesis II, once Hypothesis I has been affirmed (that there are motor effects due to the stimulations), then one must establish the standard error for the statistics, using only that particular 10 min stimulation interval, since the structure on that interval can now be quite different from that of neighboring non-stimulation intervals or other 10 min stimulation intervals (corresponding to different DBS parameter values). Hence for the Mean Activity (and its 2D and 1D components), obtained by summing over the 10 min stimulation interval, one needs to base its standard error calculation on the stationarity of the speed over same 10 min stimulation interval. For a stationary time series with a covariance function that dies out sufficiently fast so that the spectral density exists and is continuous (which occurs under our assumptions), the asymptotic variance of a sample mean (sample size n) of the process is:
lim m → + ∞ ( γ ( 0 ) + 2 ∑ h = 1 m − 1 ( 1 − h m ) γ ( h ) ) (6)
and the standard approximation to this, is to replace the covariances with their sample estimates, using a number of terms (m) of an order less than n. A standard practice is to use as the number of terms (m), the square-root of n (see Shumway and Stoffer (2011) [19]). Hence, the approximation can be applied to the various motor processes (expression (5)), e.g., A ¯ ( S k ), A ¯ 2 D ( S k ) and A ¯ 1 D ( S k ), A ¯ 1 D ( S k ), k = 1, 2, …, Nstim. One might ask if it is more accurate to fit a time series ARMA(p, q) model to each stimulation 10 min interval, estimate the ARMA parameters and to then use covariance estimates based upon these finite number of parameters. For finding standard errors of most time series parameter estimates, that approach is superior; but for the sample mean, because of its linear construction, there is no improvement, in that the use of the sample covariances in the above construction, produces an asymptotically efficient estimator (Grenander and Rosenblatt (1957) [20], Priestley (1981) [21]). In testing Hypothesis II, the above standard errors for the motor processes are utilized (Results, Fig. 8).
In order to identify the effect of dark versus light on one- and two-dimensional movement patterns, we break, for each of the 3 days, the 12 hrs (720 min) of light and dark each into three 3hr segments, with 1 hr between: 30 to 210, 270 to 450, 510 to 690. We do this for each of the nine mice. This breaks the 3 days into a sequence of eighteen 3hr segments, each separated by 1hr. In addition, certain times are excluded: those times during a stimulation and the 20 min following a stimulation (for the Stimulated Mice), and the ±30 min at a light/dark transition are excluded by construction (they are within the 1 hr separating the segments). On each segment we calculate the 1D and 2D Mean Activity (A ¯ 1 D, A ¯ 2 D). We assume piecewise stationarity for the 3 hr segments. The sample autocovariances and means are calculated on each 3 hr segment. We do this for the 1D Mean Activity and the 2D mean Activity, separately. The variances of the means for each 3hr segment are calculated, as they were for the 10 min stimulation intervals, by applying expression (6) above. We then average these 3hr-based means over the dark and over the light, and take the difference. What we wish to test is whether or not there are differences in the means in light versus dark, for 1D and for 2 D Mean Activity. Let Y1Dim, D−L and Y2Dim, D−L denote these differences in means. Thus, we obtain values Y1Dim, D−L, EstVar(Y1Dim, D−L), Y2Dim, D−L and EstVar(Y1Dim, D−L). The test statistics are:
Y 1 D i m , D − L / EstVar ( Y 1 D i m , D − L ) and Y 2 D i m , D − L / EstVar ( Y 2 D i m , D − L ) (7)
For each of the nine mice, a P-value can be calculated for the dark/light comparison, for 1D and 2D Mean Activity, and FDR analysis of the P-values is applied. These values are given in Fig. 9A–B, as part of the testing of Hypothesis III.
The important ideas of the present paper are: (Hypothesis I) The stimulation is effective. The methodology confirms that the resulting movement at the stimulation times (under appropriate DBS parameters) is in fact stimulation-driven, delineating it from merely being coincidental endogenously-driven movement; (Hypothesis II) It is 2-D movement, not 1-D movement, that occurs in response to stimulation, and, the effects due to the three amperages (75, 100, 125 μA) are statistically increasing in value and distinguishable. Moreover, there is a significant synergism at the combination of 125 μA and 125 Hz. In terms of the four frequencies, the effect due to 50 Hz was less than that for each of 125, 175, 225 Hz. There is not a significant response in 1-D movement to the stimulation; and, (Hypothesis III) It is 2-D movement, not 1-D movement, that differs between light and dark, and finally, stimulation in the light initiates a manner of movement (2-D movement) more commonly seen in the (non-stimulated) dark.
In order to establish the above ideas, we utilized two basic statistical calculations, defined above in Methods: (1) a forward mean over a moving window of width 10 min; and (2) a difference in right minus left means with moving window of width 2 min. The calculations are applied to each of seven motor activity processes:
A 1 D ( t i ) , D ( 1 ) ( t i ) , θ ( P ) ( t i ) , θ ( N P ) ( t i ) , A ( t i ) , A 2 D ( t i ) ,D ( 2 ) ( t i ) .
The first three are 1-Dim statistics (1-D Activity, 1-Dim Length, Multiple of pi/2 angle); we show that these statistics, are best removed from the resulting statistics, leaving only 2-dimensional components. The next two are 2-dimensional (Non-Multiple of pi/2 angle, (Total (1-D and 2-D, combined) Activity), but lack certain strengths, e.g., the Total Activity still has the 1-dim activity as a component, and the angle calculation has 2-dim information but has no velocity magnitude information. The final two 2-dimensional statistics (2-D Activity, 2-Dim Area) are those that have the greatest strength. (For the angle processes, the statistics calculated are fractions, rather than means.)
For Hypothesis I, to test that there is a response to (at least some of) the stimulations, the asymptotic distributions of our test statistics, under the assumption of piecewise stationarity, were derived in Methods. Statistically, one is going to calculate a single value, based upon that particularly given 10 min interval of stimulation. The starting point of each stimulation interval is surrounded by a ±80 min larger interval, with these larger intervals all separated from one another by at least 20 min. Hence, calculations on each (based upon results of Methods) are independent of one another. One wishes to show that the value of the statistic calculated for the 10 min stimulation interval is significantly different from the same calculation at an arbitrary point in the surrounding interval. The mouse is being stimulated by environmental cues all the time. One needs to show that the motor activity at the stimulation time was not just coincidental endogenously-drive movement, but rather significantly different from such. One cannot permute because of time dependency, and bootstrapping in a stationary context is difficult and involves various heuristic choices (Kunsch (1989) [22], Lahiri (2003)) [23]). However, one can imagine, for any value in the surrounding 160 min interval, that one makes the same calculation for a 10 min interval starting at any point in the 160 min. Calculations at the limits of the 160 min interval, are made in a wrap-around manner (i.e., the interval is viewed as a circle). Such wrapping around is a standard time series/Fourier procedure that has no asymptotic effect; so doing allows us to make the calculations and keep the different 160 min intervals sufficiently separated. The sample cdf of all such translations (rotations) is asymptotically derived in Methods, under the stated conditions of ϕ-mixing and absolute summability of autocovariances, which are very weak assumptions. That is, the appropriate test statistic is the restriction of permutations to just those that are translations (i.e., rotations, if viewed as a circle); only the translations adhere to the local time invariance of stationarity. In Fig. 6, representative calculations, including the final P-values of the test statistics, are displayed for one stimulation mouse (Panels A–B) and one control mouse (Panels C–D). There are four rows for each mouse. In the first row are shown the raw data on which each of four statistics are to be calculated; the second row shows the collection of values for each statistic, based upon the translations, plus the observe value of the statistic is displayed as a red asterisk. The third and fourth rows show, respectively, the resulting probability histogram and cumulative distribution function, with the observed value displayed as a red asterisk and the P-value is indicated.
In Fig. 7, Hypothesis I is tested, and the P-values for all mice and motor process statistics are summarized; the results of Fig. 6 are contained within these. Because the calculations are all made on distinct intervals separated by 20 min or more, as described in Methods, the calculations can be assumed to be uncorrelated or, more specifically, independent. For a given mouse, there are 24 (or 23 for two) stimulation intervals; each stimulation interval produces a null hypothesis for no DBS effect at those particular parameters (Amperage, Hz, light/dark). For a chosen statistic, a P-value is obtained for each interval, for each mouse. In this multiple testing setting, we use a False Discovery Rate (FDR) thresholding at q = .05, to assess the evidence of significance. First, though, the first eight stimulation intervals are for amperage 75μA. This value was included as a baseline, with little expectation of response, but rather to be used as a reference point for the primary two: 100, 125 μA. If one is assessing the strength of evidence, it does not seem appropriate to include these eight in any overall assessment, but to be considered separately. [It is analagous to including experiments with a placebo to determine if there is some physiological response to an agonist.] For simplicity and compactness of display, we have shown, in some of the subplots, the results simultaneously for all seven of the motor processes. In Fig. 7, displayed are the sorted P-values (uncorrected) for m null hypotheses, and the thresholding function (j/m) * (q/cv); in the independence case, cv = 1, and in the correlated case, cv is the sum of reciprocal indices (Benjamini and Hochberg (1995) [17], Benjamini and Yekutieli (2001) [18]).
As stated above, our calculations are independent for distinct intervals. In Fig. 7, first row, we display the three amperages 75, 100 and 125 μA, separately. One can argue that the three are different experimental conditions. In the second row, middle plot, the 100 and 125 μA null hypotheses are combined. As has been stated previously, one aspect of the work is to show that the three 1-dimensional statistics do not reflect DBS responsiveness, and should in fact be removed within appropriate statistics. The two statistics: Non-multiple of pi/2 (contains 2D information, but lacks information about velocity magnitude) and the Mean Activity (combining 1D and 2D) both lacked strength. The remaining two statistics are those of primary interest: 2D Mean Activity (A ¯ 2 D ( S k )) and the Mean 2D Area (D ¯ ( 2 ) ( S k )). We are not trying to choose between the seven statistics; each identifies separate and distinct information. However, we include a plot (second row, rightmost) of the null hypotheses for the two primary statistics, which, since they are correlated, we use FDR in the correlated case. In the second row, leftmost plot, we consider the 3 controls. In the third row we display the results for the two nonresponsive mice (Mice 5–6), on the left. The other two subplots consider the Stimulated Mice 1–4, and the second form of the calculations: Difference between the Right and Left Means, at each point; this can be viewed as a high pass filter or as a change-point detector. An expression of the strength of evidence in a given subplot is the number of significant hypotheses. One statement of the strength of evidence for at least some detectable DBS effect, would be the combined 100, 125 μA plot for the Stimulated Mice 1–4, which is the middle plot of row 2. There, the four 2-dimensional statistics each have between 20 and 25 significant tests out of the combined 62 (32+30). In all of the cases: Stimulated Mice 1–4, Controls 1–3, and Nonresponsive Mice 5–6, the three 1-dimensional statistics never go below the threshold function.
The results in Fig. 7, with respect to Hypothesis I, are that: (i) there is no measurable response, for any of the seven processes, to the stimulations at the lowest ampere level of 75 μA; this was not unexpected, in that it was chosen to hopefully identify a baseline level; (ii) the (total) Mean Activity, Mean Activity 2D, the Fraction of Non-Multiples of pi/2 Directions and the Area 2D, represent measurable stimulation responses at the 100 and 125 μA; and, (iii) the Mean Activity 1D, Fraction of Multiples of pi/2 Directions and the Length 1D, show virtually no response.
Once one has shown that there are motor responses specifically due to stimulation, one then proceeds to test (Hypothesis II) that these responses can be related to the stimulation parameters. To calculate a standard error for the stimulated response, one is restricted to using only the data in the 10 min stimulation interval itself, in that the distribution of the test statistic is now known to be different outside the stimulation interval. A time series method was presented in the Methods to calculate the standard errors for the statistics over each of the 10 min stimulation intervals (intervals separated from another by 3 hrs). One can combine these means and (unequal) standard errors across animals, to obtain overall means and standard errors. The availability of these standard errors allows one to make multiple comparison calculations, in order to assess differences in the responses with respect to amplitude (μA) versus frequency (Hz) and versus light or dark (L/D). Because the standard errors differ significantly across amplitude and frequency, traditional methods such as Analysis of Variance or the nonparametric Kruskal-Wallis test, both of which require a constant variance, cannot be applied. There is a certain degree of robustness, but the differences here are significantly beyond that (a factor 5–10, at times) (Scheffe (1959)) [24]. An alternative is still available. The statistical basis for our analysis is a multiple comparisons test, Dunnett’s C test, for which the variances are allowed to be unequal, as are the sample sizes (Dunnett (1980) [25]. In Fig. 8A, the data, consolidated over Mice 1–4, are displayed for four different statistics: 1D and 2D Mean Activities, Mean 2D Area and the RL Mean Difference of 2D Area. In Fig. 8B, a summary of the multiple comparisons (at α = .05) is presented for the the 2D and 1D Mean Activities. The results of the other two 2D statistics are similar to those of the 2D Mean Activity.
In summary, for Hypothesis II, the mean activities due to the three amperages were statistically distinguishable (and increasing with respect to amperage), whether one collapsed over L and D or compared within L or within D. From Hypothesis I, one knows that 75 μA could not be delineated from non-stimulation, and now it is shown, statistically, that 100 μA results in an increase in activity, and 125 μA in an even greater increase. As for comparisons of frequencies, they differed as to whether L and D were combined or not. A general statement (at α = .05) is that in dark, 50 Hz was statistically distinguishable from all three higher values (125, 175 and 225 Hz). If one considers nonlinear interactions between amperage and frequency, there is a significant synergistic effect at the combination of 125 μA and 125 Hz, an important, practical conclusion. Again, all of the above results were based upon multiple comparison procedures.
Hypothesis III is that stimulation initiates a pattern of movement that is more common to dark. Specifically, we show that 2-D movement bursts are more natural in the dark and that stimulation of sufficient strength in the light initiates a 2-D movement burst of the form that occurs in the non-stimulated dark state. The analysis for this hypothesis is based upon a comparison of the data, for all nine mice. To compare the 2-Dim and 1-Dim Mean Activities during light and during dark, test statistics were constructed (in expression (7)) (comparisons of the means during dark with those during light) and their standard errors were calculated, based upon estimated autocovariance functions (see Methods). P-values were calculated based upon the test statistics. For two-dimensional movement, using nine mice (six stimulated (two, non-responsive) and 3 non-stimulated controls), eight of the nine were statistically significant at.05 for the Day-light comparisons. For one-dimensional movement, only one of the nine was significant (at.05) (Fig. 9A). The strength of the evidence in such a multiple testing setting (nine null hypotheses each for 1D and 2D) was evaluated by a FDR thresholding at q = .05, with eight of the nine hypotheses being significant for 2D, and none of the nine hypotheses were significant for 1D (Fig. 9B).
The evidence is very strong for the day/light difference in two-dimensional movement. For one-dimensional movement, it is a matter of interpretation; at most, it would suggest a very weak circadian effect. Finally, in Fig. 9C, evidence is given that stimulation during light initiates movement representative of non-stimulated nocturnal movement (which from Fig. 9A–B is 2-D movement). Plotted are the fractions of 2-D movement during non-stimulated light and non-stimulated dark (for Mice 1–6, Controls 1–3), and the fractions during stimulated light and stimulated dark (for Mice 1–4). The plots reveal that similar changes occur during stimulated light (Mice 1–4) and non-stimulated dark, as compared to non-stimulated light, providing additional evidence that stimulation in light initiates movement that is more naturally nocturnal.
We have shown that Deep Brain Stimulation does initiate motor activity in response to stimulation, distinguishing it from what otherwise might have been just the coincidental occurrence of the continuously occurring stop-and-start movement of the mouse. We established that there is an increasing level of response to an increasing level of amperage, and increased response to the three higher frequencies. Significant synergism at the 125μA, 125 Hz combination was uncovered. The responses were shown to be those corresponding to 2-dim movement, not 1-dim movement. That 2-dim movement is much more common in dark than in light, was quantified. An important identification was that DBS stimulation in light initiates a level of 2-dim movement similar to the level of 2-dim movement in dark, generally (i.e., in non-stimulated dark). [We have also calculated from the incremental (per sec) changes in angle and the time-evolving winding number, and verified that stimulation initiates spatial movement and not mere spinning in place. (not included in the Results)]
The statistically discoverable principles of the impact of DBS on Generalized Arousal (GA) behavior are not well understood, yet, at the same time, its use in the treatment of a diverse range of diseases is rapidly expanding. The present paper makes an important and vital contribution by identifying, with statistical criteria, the relationship of DBS parameters to induced motor activity (which serves as a correlate for behavior).
In the present work, we have established that DBS does stimulate movement and have quantified the degree of responsiveness to the stimulation parameters (amperage and frequency). Stimulation at 100 μA produced a significant increase in activity above 75 μA (shown to be equivalent to baseline) and 125 μA a significant increase above 100 μA. A key conclusion was that there is a highly significant synergism at the combined 125μA and 125 Hz levels; 125 μA was the highest current, but 125 Hz was mid-level in the (50, 125, 175, 225 Hz) range. This could have a significant impact on the practical use of DBS as a treatment for a variety of diseases. A key concept in the present modeling was the identification of the importance of two-dimensional versus one-dimensional movement. It is 2-D movement, not 1-D, which responds to DBS stimulation. It is 2-D movement, not 1-D, which differs between light and dark. The proportion of 2-D movement in the dark is much greater than the proportion in light. The present experiments revealed that the proportion of 2-D movement initiated by stimulation in light during was similar to that of 2-D movement in non-stimulated dark. These factors, as whole, suggest that stimulation activates in light a manner of movement (2-D movement) that is more commonly, nocturnal. Not only do the above conclusions have important practical consequences for brain arousal, but the methodology developed to draw such inferences, itself, has broad potential.
The methodology is applicable to studies, broadly, for which the data consists of measurements of animal motor activity over time. One cannot apply traditional statistical methods to time-dependent processes, unless the time-varying structure is taken into account. Moreover, as a general statement, if one observes a time-dependent process, for which the time-dependency itself is changing (e.g., circadian rhythm), then, without additional knowledge, very little can be inferred about the underlying structure. If one observes a large number of animals (e.g., 15–20), under identical conditions, one can potentially avoid the time-varying issue. However, if a large number of animals is not observed, then a much different approach is required. One must begin the modeling at the level of the individual animal, and build up from there. The key assumption (justified in Methods) is that of local stationarity, specifically, piecewise stationarity. In such a setting, methods such as shuffling or permutation tests are not valid and bootstrap methods are difficult to implement. If one uses the data in a manner that does not adhere to the time-dependency, one can often end up mistakingly inferring that there is an experimental effect, when in fact what was implicitly being tested (and rejected) was that the data was IID (which it is not).
In the present approach, a method was developed to compare the activity in the stimulation interval to that in neighboring intervals, in a manner consistent with the inherent local stationarity of mouse motor activity, which is highly influenced by light and dark. The developed methods, utilizing piecewise stationarity, allowed one to calculate statistics of motor activity over a segment of time, and, most importantly, to obtain accurate and justified standard errors for those statistics, again for an individual animal. The methods then allowed one to combine across individual animals, reaching the level of desired inference (drawing conclusions based upon the full data). Because the variances at different times of day and/or different DBS parameters, are significantly different (e.g., at times, a factor of 5–10), Analysis of Variance and nonparametric tests are not applicable, but other multiple comparison tests (Dunnett’s C test) are applicable.
The conclusions of the present paper will aid in our understanding of the manner by which the CNS arousal pathways initiate various forms of behavior. In addition, the methodology developed for this work provides the experimentalist with justified methods for testing hypotheses in the common neuroscience framework in which animal motor activity is measured.
|
10.1371/journal.ppat.1003523 | Clostridium difficile 027/BI/NAP1 Encodes a Hypertoxic and Antigenically Variable Form of TcdB | The Clostridium difficile exotoxin, TcdB, which is a major virulence factor, varies between strains of this pathogen. Herein, we show that TcdB from the epidemic BI/NAP1/027 strain of C. difficile is more lethal, causes more extensive brain hemorrhage, and is antigenically variable from TcdB produced by previously studied strains of this pathogen (TcdB003). In mouse intoxication assays, TcdB from a ribotype 027 strain (TcdB027) was at least four fold more lethal than TcdB003. TcdB027 caused a previously undescribed brain hemorrhage in mice and this correlated with a heightened sensitivity of brain microvascular endothelial cells to the toxin. TcdB003 and TcdB027 also differed in their antigenic profiles and did not share cross-neutralizing epitopes in a major immunogenic region of the protein. Solid phase humoral mapping of epitopes in the carboxy-terminal domains (CTD) of TcdB027 and TcdB003 identified 11 reactive epitopes that varied between the two forms of TcdB, and 13 epitopes that were shared or overlapping. Despite the epitope differences and absence of neutralizing epitopes in the CTD of TcdB027, a toxoid form of this toxin primed a strong protective response. These findings indicate TcdB027 is a more potent toxin than TcdB003 as measured by lethality assays and pathology, moreover the sequence differences between the two forms of TcdB alter antigenic epitopes and reduce cross-neutralization by antibodies targeting the CTD.
| During the past decade, the C. difficile BI/NAP1/027 strain has emerged and in some settings predominated as the cause of C. difficile infection. Moreover, in some reports C. difficile BI/NAP1/027 has been associated with more severe disease. The reasons for association of this strain with more severe disease and relapse are poorly understood. We compared the toxicity and antigenic profiles of the major C. difficile virulence factor, TcdB, from a previously studied reference strain and a BI/NAP1/027 strain. The results indicate TcdB027, the toxin from the BI/NAP1/027 strain, is more lethal and causes more extensive brain hemorrhaging than TcdB003, the toxin produced by a reference strain of C. difficile. Furthermore, the results show that the antigenic carboxy-terminal domain (CTD) encodes at least 11 epitopes that differ between the two forms of TcdB. In line with this, experiments demonstrate that antiserum against the CTD does not cross-neutralize TcdB003 and TcdB027 toxicity against CHO cells, and TcdB027 appears to be devoid of neutralizing epitopes in this domain. These findings indicate differences in TcdB003 and TcdB027 contribute to increased virulence of C. difficile BI/NAP1/027 and reduce the likelihood of acquired immunity providing cross-protection against infection by these strains.
| Clostridium difficile is the leading cause of hospital-acquired diarrhea in developed countries [1], [2], [3], [4]. This spore-forming anaerobic bacterium contaminates hospital environments and infects patients undergoing antibiotic therapy within health care facilities [2], [5], [6]. Despite these problems, historically, treatment with antibiotics such as metronidazole and vancomycin has been an effective means of treating this disease [7], [8]. Yet, disturbing trends of increased morbidity and mortality, as well relapse of C. difficile infected patients have become apparent over the past decade [9], [10], [11], [12], [13], [14], [15]. These trends correlate with the emergence of the BI/NAP1/027 strain of C. difficile [10], [12], [16], [17]. Although an absolute association between BI/NAP1/027 strains and increased disease severity has not been made in all cases [18], [19], [20], [21], extensive clinical surveillance over the past ten years has shown a strong correlation between BI/NAP1/027 frequency and mortality rate [22], [23]. This C. difficile strain has now been found in a majority of states in the US and is prominent both in Europe and Canada [16], [24]. To date, many factors such as antibiotic resistance, sporulation ability, and toxin production have been proposed to contribute to the potential difference in virulence of historical ribotypes and C. difficile 027 [13], [25], [26], [27], [28], [29]. Yet, the relevance of these factors is still greatly debated [30], [31], leaving us with a poor understanding into how this emergent strain correlates with increased mortality.
C. difficile produces two large clostridial toxins, TcdA and TcdB, which cause extensive tissue damage and are major virulence factors in human disease [32], [33], [34]. Our work has focused on understanding how variations in the toxins produced by historical and epidemic strains change the extent of C. difficile virulence [35], [36]. Of particular interest are the differences in the sequence and activities of TcdB, which has been implicated as a critical C. difficile virulence factor [37], [38]. We hypothesize that variation between TcdB from previously predominant ribotypes and BI/NAP1/027 strains, is a major contributing factor to the increased virulence of the recently emerged forms of C. difficile.
TcdB (∼270 kDa; 2366 amino acids; YP_001087135.1) is a single chain polypeptide toxin where the glucosyltransferase domain is located at the N-terminus (GTD: 1–543), followed by an autoprocessing site between amino acid 543 and 544 which is subject to intramolecular cleavage by the cysteine protease domain (CPD: 544–807), a hydrophobic transmembrane domain (TMD: 956–1128), and a putative receptor binding domain at the C-terminus (CTD: 1651–2366) [39], [40], [41], [42], [43], [44], [45]. The gene encoding TcdB is located within a pathogenicity locus on the chromosome of C. difficile along with genes encoding TcdA (enterotoxin; YP_001087137.1), TcdE (YP_00108136.1), and regulators of toxin gene expression (TcdC, YP_001087138.1 and TcdR, YP_00108134.1) [46]. While the sequence of TcdA, TcdE, TcdR, and TcdC are almost identical between ribotype 012/003 and BI/NAP1/027 strains, TcdB is more variable (96% similarity, 92% identity) [35]. These differences in the sequence of TcdB may explain the observations of Wren and colleagues, who found that TcdB from a BI/NAP1/027 strain (TcdB027) is more potent on cultured cells than TcdB from a historical ribotype 012 strain [47]. In line with this we also found that TcdB027 causes more extensive and broader tissue pathologies than TcdB from the commonly referenced strain, VPI 10463 (TcdB003), in a zebrafish embryo model [35]. As a possible underlying mechanism for these differences in activity, we found previously that TcdB027 is translocated into cells more rapidly and is autoprocessed more efficiently than TcdB003 [35].
The greatest sequence variation between the two forms of TcdB is found in the C-terminal domain (CTD), which we define as the region of the toxin between amino acid 1651 and the terminal residue at position 2366. There is an overall 88% sequence identity between TcdB0271651-2366 and TcdB0031651-2366. The CTD of TcdB encodes combined repetitive oligopeptides (CROPs), which are thought to be responsible for the recognition of glycans on target cells [39], [48], and as such the CTD is often referred to as the receptor binding domain. However, the role of the CTD as the receptor binding domain is still very much debated as no receptor has been identified, and studies in TcdA have shown that this region contributes to, but is not required for cellular uptake of the toxin [49]. The CTD is also antigenic and known to contain neutralizing epitopes [50], [51]. Yet, whether sequence differences in the CTD of TcdB027 and TcdB003 alter the tropism or antigenic profiles of these two forms of the toxin is not known.
In the current study, we examined differences in the lethality and in vivo pathologies of TcdB027 and TcdB003. The data indicate TcdB027 exhibits a lethal dose substantially lower than TcdB003. We also show that while both toxins caused pronounced hemorrhaging in major organs, TcdB027 caused brain pathologies in vivo, as well as an increased cytotoxicity on brain microvascular cells in vitro. This study also characterized the influence of the CTD on this cell tropism and the possible contribution of sequence variation to changes in antigenicity. The data suggest that the CTD may not occupy the same role in TcdB027 as TcdB003, and identifying these key differences is a critical step toward understanding the virulence and systemic effects of C. difficile associated disease.
In previous work we found that that TcdB027 is more cytototoxic and causes broader tissue damage in a zebrafish embryo model than TcdB003 [35]. To determine how this difference in activity might impact systemic damage and lethality between the two forms of the toxin, in the first set of experiments in this study we determined and compared the lethal doses of TcdB003 and TcdB027 in a murine systemic intoxication model. The previously published lethal dose of 220 µg/kg (i.p.) for TcdB003 [32] was used to establish a range of toxin concentrations for these treatments, but the lethality we observed via i.v. injection was much higher than previously reported. As a result, the initial doses of 100 µg/kg (data not shown), 50 µg/kg, and 25 µg/kg of TcdB003 were much more potent than anticipated, and resulted in a very rapid time to death (Fig. 1A). Therefore, the remaining mice were subjected to much lower doses of 5 µg/kg and 2.5 µg/kg of TcdB003. Based on the results of the TcdB003 treated mice, the TcdB027 group started with a dose of 10 µg/kg and was continued with 1∶2 dilutions down to 625 ng/kg of TcdB027. After the mice were injected with TcdB003 or TcdB027, they were followed for up to 7 days and the survival curves of the data from these experiments are shown in Fig. 1B.
The data shown in Fig. 1 indicate mice injected with TcdB027 succumb to the toxin at a lower dose than that observed in mice injected with TcdB003. Within 26 h of treatment all of the mice administered 5 µg/kg of TcdB027 died or reached a moribund condition. In comparison, mice administered the same dose of TcdB003 did not succumb to the toxin until after 40 h and as long as 57 h with a median survival of 48 hr (Fig. 1C). At the next lower dose (2.5 µg/kg), no mice survived TcdB027 treatment, while all of the mice treated with TcdB003 survived (Fig. 1D). Based on these outcomes we estimated the LD50 of TcdB027 to be between 625 ng/kg and 1.25 µg/kg of body weight. In comparison, a higher range for TcdB003 was estimated and fell between 2.5 µg/kg and 5 µg/kg of body weight. Thus, in line with previous studies demonstrating more potent effects on cultured cells and zebrafish embryos, TcdB027 also appears to be more toxic than TcdB003 in a rodent model of intoxication.
The results shown in Fig. 1, combined with our earlier findings in the zebrafish model [35], all point to the fact that TcdB027 is more toxic than TcdB003. Recent work by Steele and colleagues detected TcdA and TcdB circulating in the bloodstream of piglets infected by C. difficile, and this correlated with systemic effects that could be blocked by passive administration of antibodies against the toxins [52]. This led us to question whether TcdB027 might also cause more extensive systemic damage than TcdB003 due to its higher potency. To assess this, mice were administered TcdB003 (2.5 µg/kg to 50 µg/kg) or TcdB027 (625 ng/kg to 10 µg/kg) and tissue pathologies were examined. Tissues and organs from mice administered sublethal doses of the toxins did not reveal pathologies that differed from that of control (Fig. 2A). In contrast, abnormal tissue histologies were found in several of the major organs examined from mice intoxicated with lethal doses of TcdB. Mice treated with either TcdB003 or TcdB027 showed pronounced liver damage with extensive blood-pooling, parenchymal cell loss, and evidence of hemorrhage, which can be visualized by the appearance and expansion of the dark red patches as the survival time progresses (Fig. 2A). To a lesser extent, acute hepatocellular coagulative necrosis and hemorrhage in the spleen along with follicular necrosis and possible apoptotic cells was also detected (data not shown). The severity of the observed pathologies was more related to the length of time of toxin exposure rather than toxin concentration. Figure 2A shows representative liver sections from TcdB003 and TcdB027 treated mice, illustrating that the damage is the more extensive in mice receiving the minimum lethal dose and surviving for the longest period of time.
Despite the difference in lethality, the majority of the in vivo effects of TcdB003 and TcdB027 were identical, with the exception of moderate to severe hemorrhage detected in the brain of TcdB027 treated mice. Indeed, brain hemorrhage was the most obvious difference between mice exposed to the two forms of TcdB. The brains of mice treated with TcdB003 displayed only small lesions while the brain hemorrhage of TcdB027-treated mice was profuse with large multi-focal areas of blood accumulation within the cerebellum and cerebrum (Fig. 2B). These data suggest there may be a loss of endothelial integrity in mice challenged with TcdB, as well as a significant difference in the in vivo targeting and tropism of TcdB003 versus TcdB027.
Experiments were next performed to determine the toxicity of the two forms of the TcdB on endothelial cell lines as a possible correlation with the differences in the amount of brain hemorrhage. We first wanted to determine whether endothelial cells displayed increased sensitivity to TcdB compared to the epithelial-like cells (e.g. CHO cells) that are normally used in cytotoxicity assays. Rat Aortic Endothelial Cells (RAEC) exposed to TcdB003 and TcdB027 displayed very similar cytotoxic doses (Fig. 3A). The concentration needed to cause toxicity in 50% of culture cells (TCD50) for TcdB003 was 6.07±1.41×10−12 M and 2.74±1.16×10−12 M for TcdB027. Since the major differences in pathology between TcdB003 and TcdB027 occurred in the brain, we next tested rat brain microvascular endothelial cells (RBMVEC) for differences in sensitivity to the two forms of TcdB. Interestingly, there was a 10-fold difference in the cytotoxicity of TcdB027 on the RBMVECs, with the TCD50 being 6.32±1.16×10−13 M compared to the TCD50 of 8.46±1.12×10−12 M for TcdB003 (Fig. 3B). These data indicated that TcdB was highly cytotoxic on endothelial cells, as the previous published observations of TcdB003 and TcdB027 toxicity on CHO cells is 2.53×10−11 and 2.37×10−13 respectively. Additionally, the RBMVECs had a greater susceptibility to TcdB027, which correlates with the brain pathologies in Fig. 2B.
To further study the differences in the cell and organ targeting between TcdB003 and TcdB027, we focused on the CTD, which is thought to be important in facilitating cell interactions [39], [53]. We hypothesized that if this region is indeed important in cell targeting, then the sequence differences between TcdB003 and TcdB027 in this region could be an important factor in the distinct cell tropism and animal pathologies between the toxins. We also predicted that these differences could change the profile of antigenic epitopes, and perhaps neutralizing epitopes, in the CTD. We designed a set of experiments to address both of these possibilities.
In order to evaluate differences in the CTD of TcdB003 and TcdB027 we expressed and purified protein fragments representing this region of each toxin. These fragments consisted of the final 721 amino acids of the TcdB protein, including the CROP region along with approximately 206 residues amino terminal to the CROP region. Based on previous sequence comparisons, there are 89 residues that differ between CTD003 and CTD027[35].
Initially, each CTD was used as an antigen to immunize rabbits for the collection of CTD antisera, which were then used in TcdB neutralization assays to further determine the impact of the CTD on the activity of both TcdB003 and TcdB027. We first investigated the impact of αCTD003 on the cytotoxicity of both TcdB003 and TcdB027 and found that treatment with αCTD003 neutralized the cytotoxic and cytopathic effects of TcdB003 (Fig. 4A). However, αCTD003 caused no detectible reduction in the cytotoxicity of TcdB027 (Fig. 4A). ELISA analysis confirmed that while αCTD003 was only able to neutralize TcdB003 in cell culture, the polyclonal serum could recognize both TcdB003 and TcdB027 in vitro (Fig. 4B). When the αCTD027 antibody was used in the neutralization assay, we found no protection against either TcdB003 or TcdB027, although the serum strongly reacted with both forms of the toxin as determined by ELISA (Fig. 4A and 4B).
The data shown in Fig. 4 suggested that CTD027 and CTD003 differ in their profile of neutralizing epitopes (i.e. sequences where antibody binding blocks intoxication). It was also possible that TcdB027 shared the same sequences of TcdB003 neutralizing epitopes, but, unlike TcdB003, TcdB027 did not depend on these regions for cellular intoxication. To address this alternative explanation, serum against CTD003 was incubated with a 100-fold excess of CTD003 or CTD027, and the mixture was tested for its ability to neutralize cytotoxicity of TcdB003. We reasoned that if CTD027 contains sequences that are targets for antibody-mediated neutralization of TcdB003 then the preincubation with CTD027 should prevent the antiserum from neutralizing TcdB003. As expected, the addition of CTD003 in the neutralization assay resulted in the inhibition of antibody activity and a return to full cytotoxicity of TcdB003 (Fig. 4C and 4D). In line with the possibility that TcdB027 contains sequences that are neutralizing epitopes in TcdB003, preincubation with CTD027 also blocked the neutralizing effects antiserum against TcdB003 (Fig. 4C and 4D).
The data from the analysis of antiserum against the two forms of TcdB suggested there is likely to be shared epitopes between the two proteins, but the extent of shared and unique epitopes was difficult to predict. In order to begin to identify shared and unique epitopes between TcdB027 and TcdB003 we used solid phase peptide based ELISAs to map antibody reactive sequences in the CTD of TcdB. In all, 358 decamer peptides, overlapping by 8 residues and covering the entire CTD003 sequence, were synthesized and tested for reactivity to CTD003 and CTD027 sera. Sera was collected from rabbits immunized with CTD003 or CTD027 (n = 2), and when we compared the peptides recognized by αCTD003 to those recognized by αCTD027 we found an overall difference in the pattern of peptides recognized by antisera from the 2 groups (Fig. 5). Each serum sample was analyzed individually, and the average response of αCTD003 and αCTD027 to the CTD003 peptides is shown in Fig. 5. The analysis identified identical epitopes, overlapping epitopes, and epitopes unique to each form of the toxin. The analysis identified approximately 7 regions that were recognized only by αCTD003 (Fig. 5). The analysis also found 4 regions recognized by only αCTD027 and 13 regions where there was overlap or exact matches in the epitopes recognized by both sera (Fig. 5). The majority of the peptides identified are localized in the CROP domain, and many of the epitopes that differ in recognition between αCTD003 and αCTD027 are located sequentially, within the first seven repeats of the CTD. As summarized in Fig. 5, peptides recognized by only the αCTD003 serum were variable regions between the two toxins, with as many as 6 amino acid differences as in the case of peptide 21. In contrast, the peptides recognized by only αCTD027 were highly conserved between the two forms of TcdB, with only one peptide (#7), with a single amino acid change. These data suggest that sequence variation of TcdB027 impacts antibody recognition of sequential epitopes and may contribute to differences in conformational epitopes as well.
The observation that the CTD of TcdB027 is a poor target for the production of antibodies that prevent toxicity on CHO cells, raised concerns about the overall antigenicity of TcdB027. The majority of the amino acid sequence variation between TcdB003 and TcdB027 occurs in the CTD, so we reasoned that producing antibodies using the holotoxin as an antigen could have better potential to be broadly neutralizing. Both TcdB003 and TcdB027 were inactivated using formaldehyde to create ToxoidB003 and ToxoidB027. These toxoids were used as antigen to immunize mice and test for protective antibodies against TcdB. After two subsequent boosts, serum was collected from the mice, and the neutralizing effects were tested in vitro. The data in Fig. 6A shows that the mouse antiserum toward ToxoidB027 protected against the cytotoxic effects of both TcdB003 and TcdB027, while anti-Toxoid003 was not cross-neutralizing and only maintained the cell viability of the CHO cells treated with TcdB003. The immunized mice were next tested for protection from TcdB in vivo, using a 2-fold minimum lethal dose of TcdB003 or TcdB027. Consistent with the in vitro neutralization data, all mice immunized with ToxoidB027 were completely protected from i.v. challenge of both TcdB003 and TcdB027 (Fig. 6B and 6C). Immunization with ToxoidB003 provided only a slight, yet significant protective effect, increasing the median survival from 15 h to 24 h in mice injected with TcdB003, but only from 9 h to 13 h in mice challenged with TcdB027 (Fig. 6B and 6C). Eventually, all of the ToxoidB003 mice succumbed to the effects of TcdB027, and only two ToxoidB003 mice were fully protected from TcdB003 (Fig. 6B and 6C). Whereas the antisera to the CTD of TcdB027 showed no effect, antibodies to the toxoid form of TcdB027 successfully inhibited toxicity, suggesting that the protective effect against TcdB027 is better conferred by the full-length toxin rather than the CTD in this system.
C. difficile infection is a complex illness commonly involving colitis and, in more severe cases, systemic complications [54], [55], [56]. In the current study we sought to determine how systemic complications vary between two forms of TcdB. To focus on the systemic events mediated by the different forms of TcdB, we bypassed the intestinal stage of this illness by directly administering toxin intravenously. This analysis found that TcdB027 was more lethal and caused more pronounced systemic damage than TcdB003. Further studies revealed this effect correlated with differences in the extent of specific cellular tropisms between the variants of TcdB. Assessing the CTD of TcdB found that this region may contribute to not only differences in tropism, but also accounts for a variability in the antigenic make-up of this domain. Collectively, the data support the notion that TcdB027 is not only more potent than TcdB003, but may have sequence alterations that prevent cross neutralization.
Several recent observations led us to predict that the increased virulence of C. difficile BI/NAP1/027 is due to altered TcdB activity. First, the sequence of TcdB, but not TcdA, varies between the two strains [35], [57]. Second, in cell culture systems, TcdB027 is more potent on a broad range of cell types [35], [47], [57]. Thus, we hypothesized that TcdB027 could have a lower lethal dose and cause more extensive tissue damage in vivo. Our findings support this hypothesis. When experiments compared the lethal doses of TcdB027 and TcdB003 the BI/NAP1/027 toxin was found to be 4 times more lethal than the ribotype 003 toxin (Fig. 1). More importantly, TcdB027-treated mice died much more quickly and, in some cases, in less than half the time than TcdB003-treated mice. In regards to the pathologies, TcdB027 clearly caused brain damage that was less prominent in mice treated with TcdB003 (Fig. 2). These findings provide insight into the differences in the in vivo effects of TcdB027 and TcdB003, and this variation in toxicity could contribute to more severe disease caused by recently emerged strains of C. difficile.
Very little is known about the underlying mechanisms of C. difficile-induced systemic damage and complications. The extent to which the pathologies observed in toxin-treated mice reflect systemic complications in humans is not known and there is clearly a need for more studies in this area. However, several reports make it reasonable to suspect the toxins contribute to the systemic complications in this disease [54], [55], [56]. The idea that toxin enters the bloodstream during disease is supported by recent work using a piglet model of C. difficile infection where TcdA and TcdB were detected in the bloodstream of the infected animals [52]. Other work has demonstrated that serum IgG, and not mucosal IgA, against the toxins correspond with protection against illness and relapse [58], [59], [60] further supporting the notion of systemic effects of these toxins. Thus, the more extensive systemic damage caused by TcdB027 may explain in part why C. difficile NAP1/BI/027 is associated with more severe disease.
Our previous studies found that TcdB003 is cardiotoxic and targets cardiomyocytes with an equal efficiency to TcdB027 [35], [61]. In vivo and in vitro data support the notion that the two forms are TcdB are very similar in their cardiotoxic effects, but the sequence differences in TcdB027 allow the toxin to target other tissues an cell types more effectively than TcdB003. Consistent with this idea, the TCD50 for TcdB027 and TcdB003 was found to be very similar on aortic endothelial cells, but substantially lower for TcdB027 on brain microvascular endothelial cells. Thus, the evidence to date supports a model where both forms of TcdB are cardiotoxic, but TcdB027 is more potent on other tissue and cell types.
The fact that TcdB027 is a more potent toxin than TcdB003 is now well established by several in vivo and in vitro analyses [35], [47], including the ones used in this study. Yet, the sequence changes accounting for these differences in activity have not been defined. There are 198 residue differences between TcdB027 and TcdB003 and each of the residues known to be critical for TcdB activities are conserved between the two forms of this toxin. In previous work we found that TcdB027 undergoes more complete autocleavage because it is able to engage intramolecular substrate more effectively than TcdB003 [36]. This implies the conformation of TcdB027 may be different than that of TcdB003. We have also shown that TcdB027 undergoes dramatic pH-dependent conformational changes more extensively and at a higher pH than TcdB003 [35]. Again, this is unlikely to be related to a single residue change and could be the result of the collective sequence differences.
The finding that antibodies against the CTD neutralized TcdB003 but not TcdB027 on CHO cells could be the result of TcdB027 using an alternative means of cell recognition. Interestingly, Olling et al. have reported that the CROP domain of TcdA is involved in cellular uptake of the toxin, but it is not entirely responsible for cell recognition and binding [49]. In a like manner, it is plausible that the role of the CTD has become less significant in TcdB027 and variations have little effect on the toxin. If so, TcdB027 could bind cells by an alternative manner, which helps explain the current data that TcdB027 has a broad effect in mice, as well as previous data that shows extensive necrosis in a zebrafish model of intoxication.
The data from the peptide arrays showed αCTD003 reactivity with many epitopes in which the sequence varied in TcdB027. Whether these sequence variations evolved as a way of allowing TcdB027 to avoid immune recognition or if this is a means of TcdB027 altering its activity, is not yet clear. If the former is true, it could be possible that a change to one single epitope could be responsible for the lack of neutralization of TcdB027. However, work by Torres and Monath suggests that while the CTD is quite antigenic, antibodies to a single peptide epitope fail to prevent cytotoxicity of TcdB [50]. Finally, in further support of the idea that the two toxins are not identical in their overall structure, three of the epitopes recognized by serum against TcdB027 were not recognized by serum against TcdB003 despite the fact that these sequences were the same (Fig. 5). The conformational differences in the two forms of TcdB could determine whether identical sequences are antigenic.
It is also important to consider this variation in the context of virulence of C. difficile, as well as vaccination. Our previous work suggests that TcdB027 enters cells more rapidly and efficiently than TcdB003 [35]. Given that the CTD is believed to facilitate interactions with the cell surface, it is possible that antigen recognition occurs, but the toxin overcomes this by utilizing a more effective mechanism of cell entry. Arguing against this possibility is the fact that we did not detect even a minor change in the rates of TcdB027-induced cell rounding or the overall level of cell killing. It's also important to note that our experiments involved preincubating TcdB027 with the antiserum. Therefore, if the toxin overcame the neutralizing effect by more efficient cell entry, we would expect to see at least a nominal change in toxicity, but this doesn't appear to be the case. We believe the reasonable explanation is that the neutralizing epitopes of TcdB027 are sufficiently altered to avoid toxin neutralization or that the toxin has a different mechanism of interacting with and entering the cell. These data also suggest successful vaccines targeting TcdB will need to include antigens from multiple forms of this toxin or, alternatively, be designed to target highly conserved neutralizing epitopes shared among variants of TcdB.
Although further studies are needed, the toxoid of TcdB027 could provide a vaccine that generates a broadly neutralizing response. Given that the CTD027 did not generate an antibody response that protected CHO cells from TcdB027, and past studies have found that TcdB toxoid is not a highly effective vaccine [62], [63], we were surprised to find the toxoid of TcdB027 stimulated a potent neutralizing response in mice. It has been known for many years that anti-serum does not cross neutralize TcdA and TcdB, making it reasonable to consider the possibility that anti-serum to the variant forms of TcdB also do not cross neutralize. This does not appear to be the case. As shown in Fig. 6, mice vaccinated with the toxoid form of TcdB027 were completely protected against both TcdB003 and TcdB027. In line with a prior study by Wang et al. [64], the toxoid of TcdB003 evoked only marginal immunoprotection against TcdB, and we found this to be true for mice challenged with either the historical or ribotype 027 form of the toxin. This raises the possibility that converting TcdB003 into a toxoid alters the protein in a way that reduces immunogenicity, but sequence differences in TcdB027 make this form of the toxin more effective as a toxoid.
Overall, these findings demonstrate critical differences between TcdB produced by ribotype 003 and ribotype 027 strains of C. difficile. The sequence variations in TcdB027 impact the toxin's cytotoxicity, lethality, and antigenic make-up, and likely contribute to the overall heightened virulence of C. difficile BI/NAP1/027 strains.
The animal immunization and toxin challenge studies were performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All animal procedures reported herein were approved by the Institutional Animal Care and Use Committee and Institutional Biosafety Committee at OUHSC (IACUC protocol # 09-092-I and 11-016-I). The procedures precluded the use of anesthesia for in vivo lethal challenge assays. To minimize pain and distress, the mice were monitored at least twice daily and any animals with signs of distress such as labored breathing, lethargy, inability to eat or drink, ruffled fur, disorientation, or loss of 20% body weight were euthanized immediately. This method was approved by the IACUC and monitored by a qualified veterinarian.
C. difficile VPI 10463, a ribotype 003 strain (produces TcdB with identical sequence to the 630/ribotype 012 strain), and C. difficile BI17 6493, a ribotype 027 strain (a gift from Dr. Dale Gerding), were used as sources of to purify TcdB003 and TcdB027 respectively.
Female BALB/cJ and C57B/6J mice (Jackson Laboratories), aged 8 weeks, were purchased from The Jackson Laboratories (Bar Harbor, ME) and handled in accordance with IACUC guidelines at University of Oklahoma Health Science Center.
Rat Brain Microvascular Endothelial Cells (RBMVEC) and Rat Aortic Endothelial Cells were a generous gift from the laboratory of Dr. Eric Howard (University of Oklahoma Health Sciences Center) and have been described previously [65], [66]. CHO-K1 cells were purchased from American Type Culture Collection (ATCC). RBMVEC and RAEC were grown in DMEM containing 10% FBS while CHO cells were grown in F12-K with 10% FBS. All cell types were used between passage 15–30, and were maintained in tissue culture treated T-75 flasks (Corning) at 37°C in the presence of 6% CO2.
C. difficile was cultured using the dialysis method as previously described [35] and TcdB was isolated using anion-exchange (Q-Sepharose) chromatography in 20 mM Tris-HCl, 20 mM CaCl2, pH 8.0, following a thyroglobulin affinity chromatography protocol to first remove TcdA [67]. Purification of TcdB was confirmed by visualization of a single 270 kDa band by SDS-PAGE, and LC/MS/MS analysis (University of Oklahoma Health Science Center).
Toxoid versions of TcdB003 and TcdB027 were prepared by mixing 500 µl of TcdB (0.4 µg/µl) into 500 µl of 8% formaldehyde with 8.5 mg of lysine to help prevent precipitation and aggregation of the formalinized protein [68], [69], and incubating at 37°C overnight. The volume was then brought up to 10 ml with PBS, yielding 20 µg/ml of ToxoidB in 0.4% formaldehyde with 0.425 mg/ml lysine. Both toxoid preparations lacked toxic activity as confirmed by the absence of cytopathic effects on CHO cells.
The CTD-encoding region of tcdb gene (YP_001087135.1: nucleotides 4961–7111) from the strain VPI 10463 was codon optimized and cloned into pET15b (Genscript). The CTD of the tcdb gene (YP_003217086.1: nucleotides 4961–7111) from the NAP1 strain was cloned from a pET15b plasmid containing full-length tcdb that had been codon optimized by Genscript. The CTD gene was amplified using primers 5′-GATCATATGCTGTATGTGGGTAACCG-3′ and 5′-AACGGATCCTTATTCGCTAATAACCA-3′ containing BamHI and Nde1 sites for cloning into pET15b. The CTDs were expressed using Escherichia coli BL21 star DE3 (Invitrogen) at 16°C overnight and then purified by Ni2+ affinity chromatography (HisTrap, GE Life Sciences) resulting in proteins representing TcdB1651–2366 from both TcdB003 and TcdB027.
To determine the differences in the minimum lethal dose of TcdB003 and TcdB027, 100 µl of TcdB003 or TcdB027 dilutions in phosphate-buffered saline was injected intravenously into the tails of BALB/cJ mice using a 27-gauge needle. Twenty mice were given TcdB003 in groups of 4, receiving doses of 2 µg, 1 µg, 500 ng, 100 ng, and 50 ng. Twenty additional mice were injected with doses of 200 ng, 100 ng, 50 ng, 25 ng, and 12.5 ng of TcdB027 (n = 4). The animals were monitored for up to 7 days post challenge for toxin effects and mortality, and mice were euthanized if they became significantly distressed or moribund. Survival was graphed using Kaplan-Meier analyses on GraphPad Prism (GraphPad Software, Inc., La Jolla, CA).
Immediately after death, the mice were dissected and major organs and tissues were submerged in formalin fixative overnight. Tissue sectioning, slide preparation, H&E staining, and pathology analysis was performed by the Department of Comparative Medicine at OUHSC.
Two rabbits per group were immunized with 0.1 mg of the CTD fragment of TcdB003 or TcdB027 in complete Freund's adjuvant on day 1 and boosted with 0.1 mg in incomplete Freund's adjuvant on days 14, 21, and 49. Blood samples were collected on days 0, 35, and 56. These experiments were carried out by Cocalico Biologicals Inc. (Reamstown, PA).
BALB/cJ mice (20 mice each for ToxoidB003 and ToxoidB027) were injected in equal portions subcutaneously and intraperitoneally with 2 µg of toxoid in PBS emulsified 1∶1 in 100 µl of complete Freund's adjuvant on day 1 and boosted with 2 µg in incomplete Freund's adjuvant on day 10. Control mice were similarly immunized and boosted using an unrelated peptide. Blood samples were collected via tail bleeds on day 0 and 24, and each bleed was tested by ELISA to evaluate toxoid response.
After completion of the immunizations, the mice were subjected to i.v. challenges of TcdB003 and TcdB027. Each immunization group (ToxoidB003, ToxoidB027, control) contained 20 mice, and 9 from each group were injected via the tail vein with a 2-fold lethal dose of either TcdB003 or TcdB027. The previously established minimum lethal dose was used to set the 2×LD100 at 200 ng per mouse for TcdB003 and 50 ng per mouse for TcdB027. The remaining 2 mice from each group were euthanized and exsanguinated for serum collection. The animals were monitored for up to 7 days post challenge for toxic effects and mortality, and mice were euthanized if they became significantly distressed or moribund. Survival was graphed using Kaplan-Meier analyses and compared with the Log-rank test on GraphPad Prism (GraphPad Software, Inc., La Jolla, CA).
Direct antigen ELISAs were used to measure the antibody reactivity in animal sera. 1 µg of purified TcdB or CTD fragment was coated per well in polystyrene plates at 4°C overnight. The plates were washed and blocked with 0.1% BSA in PBS for 1 h at room temperature. Then, the rabbit sera diluted at 1∶100 and 1∶1000 in PBS-Tween with 0.1% BSA was added in triplicate and incubated for 2–3 h at room temperature. Plates were washed with PBS-Tween and incubated with anti-rabbit IgG conjugated to alkaline phosphatase (Jackson ImmunoResearch Laboratories, Inc) at a dilution of 1∶5,000 for 3 hours at room temperature then washed and developed with p-Nitrophenyl Phosphate substrate (Sigma). Plates were read at 405 nm using a Tecan-infinite plate reader (Tecan Group, Ltd.). Plates were read when the positive control reached an OD of 1.0 and the assay was considered invalid if the negative control was over OD 0.2.
Cells were seeded in 96 well plates at a density of 1–2×104 cells per well in DMEM or F12-K (ATCC) containing 10% FBS (ATCC). For TcdB sensitivity measurements on endothelial cells, dilutions of TcdB003 or TcdB027 were added to each well in triplicate, and the cells were incubated 24 h and cell viability was measured by CCK-8 (Sigma). In order to measure neutralization of TcdB, a 1∶10 dilution of serum raised in rabbits against the CTD or mouse serum to the toxoid was preincubated with 37 pM TcdB003 or TcdB027 alone, or with 3.7 nM CTD003 or CTD027, for 1 h at 37°C in F12-K media (ATCC). CHO cells were treated with the toxin/antiserum mixture or toxin alone and incubated at 37°C for up to 24 h. Cells were analyzed under the microscope for cell rounding at 2–4 h and cell viability was measured at 24 h using a CCK-8 assay according to manufacturers instructions (Sigma).
The 358 decapeptides overlapping by 8 amino acids covering the length of the CTD region from TcdB003, were covalently synthesized on polyethylene solid phase supports (pins) as previously described and used to assay antibody specificity with a modified ELISA assay [70]. Blocking was performed in 3% milk in PBS for 1 h at room temperature, then the peptides were incubated in 100 µl/well of sera diluted 1∶100 in 3% milk-PBS with 0.05% Tween for 2 h at room temperature. The pins were washed 4 times for 8 min with mild agitation in PBS-Tween and then incubated with 100 µl/well of a 1∶5,000 dilution of anti-rabbit IgG conjugated to alkaline phosphatase in 3% milk-PBS with 0.05% Tween at 4°C overnight (Jackson ImmunoResearch Laboratories). Next, washes were performed as previous and the peptide ELISAs was developed using 100 µl/well of a 1 mg/ml solution of p-nitrophenyl phosphate dissolved in 150 mM carbonate buffer pH 10.4 containing 100 mM glycine, 1 mM MgCl2 and 1 mM ZnCl2. The absorbance was read at 405 nm using a Tecan-infinite plate reader (Tecan Group, Ltd.), and the results were normalized to the standard positive control peptide having an OD of 1.0. Positive epitopes were defined as at least two consecutive peptides with an OD greater than 2 standard deviations above the mean of pre-bleed serum.
Relevant SwissProt accession numbers are P18177 (TcdB003/CTD003), P16154 (TcdA003), C9YJ35 (TcdB027/CTD027), C9YJ37 (TcdA027),
|
10.1371/journal.pcbi.1002528 | Fine-Tuning Tomato Agronomic Properties by Computational Genome Redesign | Considering cells as biofactories, we aimed to optimize its internal processes by using the same engineering principles that large industries are implementing nowadays: lean manufacturing. We have applied reverse engineering computational methods to transcriptomic, metabolomic and phenomic data obtained from a collection of tomato recombinant inbreed lines to formulate a kinetic and constraint-based model that efficiently describes the cellular metabolism from expression of a minimal core of genes. Based on predicted metabolic profiles, a close association with agronomic and organoleptic properties of the ripe fruit was revealed with high statistical confidence. Inspired in a synthetic biology approach, the model was used for exploring the landscape of all possible local transcriptional changes with the aim of engineering tomato fruits with fine-tuned biotechnological properties. The method was validated by the ability of the proposed genomes, engineered for modified desired agronomic traits, to recapitulate experimental correlations between associated metabolites.
| Considering cells as biofactories, we aimed to optimize their internal processes by using existing design principles acquired from engineering. Herein, we present a synthetic biology approach based on experimental and computational methodology that integrates genomic, transcriptomic, metabolomic and phenomic data to formulate a kinetic and constraint based model of tomato agronomic and fruit quality characteristics. The model has been used for exploring the landscape of all possible local transcriptional changes with the aim of engineering tomato fruits with improved biotechnological properties. The methodology was validated by the ability of the proposed engineered genomes with modified desired agronomic traits, to recapitulate correlations between associated metabolites that are found experimentally in a number of examples.
| Considering a cell as a DNA-based molecular factory [1] and applying principles drawn from industrial engineering provides new approaches to optimize cellular performance (Figure S1). This approach adopts the new philosophy implemented nowadays by large industries that is known as Lean Manufacturing (LM). LM consists in the implementation of standards based on elimination of bottlenecks and processes without mark-up and minimization of pathways and excessive costs. This approach can be applied to the emerging fields of systems and synthetic biology, and allows translating engineering concepts into biotechnology [2]–[4]. Our main goal is to optimize the phenotypic response of a natural plant biofactory, exemplified here by the edible tomato fruit, by using a combined experimental and computational synthetic biology approach. The approach involves re-designing the fruit factory from within; i.e., by modeling and identifying the important genes and intermediates for a given trait of agronomical interest.
Previous works have considered modeling the global metabolism [5], transcription [6]–[11] or the integration of both in microbial organisms [12]–[14] from the point of view of systems biology. Many groups, using a re-designing strategy that is characteristic of synthetic biology, have implemented genome-scale re-designs and explorations of the gene knockout landscape both in prokaryotes [15]–[17] and eukaryotes [18]. More recent reports have tackled the prediction of phenotypes from metabolic data based on statistical models for microbes [12] and plants [18]–[20]. The next logical and desirable development should consist in modeling phenotypes of interest in a complex organism from metabolic and gene expression data. For that purpose we have chosen tomato: a model plant for fleshy fruit -this being a natural biofactory of nutrients and healthy compounds, and a plant of agronomic interest with well-developed genetics and genomics (http://solgenomics.net) and with extensive work on metadata analysis [21]–[23]. We have assumed that at least in part the genetic program of the fruit at the ripe stage should have an impact on the metabolite content and also in other high order fruit traits. In this study, we have used omic data that have been experimentally obtained by means of transcriptomics, metabolomics and phenomics for a large number of recombinant inbred lines (RILs) derived from a cross of Solanum lycopersicum×S. pimpinellifolium. Following the LM approach, we have developed here a novel in silico optimization method that extensively explores single and multiple genetic perturbations to render a series of desired tomato phenotypes; i.e., show agronomical properties of biotechnological interest. Recently, large efforts in genome-scale modeling have been reported [24], [25] (e.g., genome wide selection methods). Herein, techniques based on reverse engineering were applied to a large set of experimental omics data to obtain a kinetic model based on ordinary differential equations (ODEs) that describe the steady state concentration of mRNAs. This model has the advantage of quantitatively characterizing the kinetic parameters describing molecular interactions that are essential for simulating the genetic perturbations involved in redesigning genomes. Hence, this model describes the fruit metabolic profile from gene expression data for an autonomous subset of genes with potential effect on transcription regulation. By capturing relationships between metabolic profiles and high-throughput phenomic data, our model was extended to predict changes in agronomic properties that would be produced by specific changes in genetic expression (Figure S2).
Finally, in order to close the design cycle imposed by LM, the genetic modifications suggested by our computational approach were experimentally verified. This was done by demonstrating the predicted ability of the in silico modified fruit genomes to reproduce the correlations between metabolites empirically found in the fruit. We propose that the principles and practices learned from these engineering success cases can help to formulate a model to guide the design of new organisms with biotechnological applications.
We have extended our recently developed inference methodology, InferGene [7], to obtain a gene regulatory model coupled to metabolism that allows us analyzing optimality in terms of specified agronomic and organoleptic properties of the tomato fruit (Figure 1). For this, we have taken advantage of an experimentally characterized subset of the metabolome of 169 tomato RILs, which includes the accumulation levels of 67 metabolites in the fruit and that contribute to the flavor (sugars, acids and some volatiles), aroma (volatiles) and other quality traits (such as color and healthy carotenoids and vitamins). Moreover, we have also used the information on transcript levels from fruits for a subset of the 50 RILs analyzed at the metabolic level, to select 5592 non-redundant genes that were consistently expressed in those fruit samples (see Methods).
Transcriptomic and metabolomic data from these 50 RILs were normalized by the LOWESS method [26] and used to construct a model that predicts components of the fruit quality metabolome from transcriptome data; i.e., level of a given metabolite is effectively determined by the expression of a minimal set of genes. The size of the space of possible gene-predictors was reduced in one order of magnitude by using a CLR method (Dataset S1). After that, LASSO method was used to find a minimal set of potential predictor genes for each metabolite; subsequently, multiple regressions were obtained to estimate the effective kinetic parameters of a linear model based on ODEs that integrates transcription and metabolism processes in steady state (Figure 2) [7]. Values were used as optimal threshold in order to limit the number of possible gene-metabolite interactions and minimize the distance between the predicted and measured metabolic profiles over the training set in terms of average Pearson correlations (blue bars in Figure 2C; r = 0.85, 167 d.f., ). Hence, on average, each metabolite required 18 genes for explaining its behavior, thus a total of 959 genes was required to describe our tomato fruit metabolome. This subset of genes constitutes the effective transcription network. We performed a 5-fold cross-validation test to rule out dependence of the testing set, this reducing the metabolite average prediction (red bars in Figure 2C; r = 0.42, 167 d.f., p = 0.067 with a mean false positive rate (FPR) of 14% and a 56% mean positive predictive value (PPV) of predictors (bootstrap test, and , respectively).
The next step was to construct an effective gene regulatory model able to predict autonomously the transcriptional processes that, by means of the model previously described, would generate a quantitative metabolic response. In this way changes at the transcriptional level resulting from the proposed genetic perturbations could be translated and predicted effectively into metabolic changes. For doing that, we used the microarray data obtained from fruits of 50 of the RILs to infer a network of gene-gene interactions. The CLR method provided the first sets () of predictor genes for each gene considered. Afterwards, LASSO method reduced the number of regulations per gene to a scale-free space following a power-law with exponent () and an average of 26 interactions per gene. High values of similarity between the predicted and measured gene expression (blue bars in Figure 2D) were computed for the whole training set (, 48 d.f., ) while for a 5-fold cross validation the average similarity (red bars in Figure 2D) was r = 0.59 (48 d.f., ) with a mean FPR of the 25% and a 63%mean PPV of predictors (bootstrap test, and , respectively).
We addressed the question of whether the agronomic/phenotypic properties of the tomato fruit could be determined by their metabolite composition. For that, we studied the relationship between agronomic properties and metabolic composition across 169 tomato RILs. We applied LASSO method to select a set of metabolites that may act as predictors for each agronomic property (Dataset S1). Our model included 47 metabolites observing considerably high Pearson correlations between the measured and predicted phenotypic responses over the 169 RILs for number of fruits per plant and fruit harvested across two different seasons, (Figure 2A; r = 0.62 and r = 0.73 respectively, 167 d.f., in both cases). A reduction to r = 0.46 (167 d.f., ) and r = 0.62 (167 d.f., ) in the median correlation was computed in a 10-fold cross validation, with 84% mean PPV in both cases (bootstrap test, ), and mean FPR of 33% and 35% (bootstrap test, in both cases), respectively. Average fruit weight and pH required as many as 44 metabolites as potential predictors with high reliability levels. Reliability was assessed by comparing the corresponding predicted and measured values for the 169 RILs (Figure 2A; r = 0.85 and r = 0.80, 167 d.f., in both cases). A 10-fold validation only reduced those similarities to r = 0.73 and r = 0.63 (167 d.f., in both cases), with mean FPRs of 37% and 22% (bootstrap test, and , respectively), and mean PPVs of 81% and 88% (bootstrap test, and ), respectively. Additionally, to test how the metabolome contributed to an accurate prediction of tomato phenotype, we studied the relationship between agronomic properties and gene expression of the core of 959 genes across the 50 tomato RILs [27]. Note that we select this reduced set of genes as a core of potential predictors to avoid model over-fitting due to the low number of RILs with the transcript levels measured. Imposing the same criteria that was used to select metabolites as predictors, we observed that similarities between predicted and measured values of number of fruits per plant and harvested fruits increased (r = 0.80 and r = 0.81, 48 d.f., in both cases) while average of fruit weight and pH decreased (r = 0.79 and r = 0.73, 48 d.f., in both cases) (see dashed line in Figure S3A–B). Moreover, relaxing the threshold () to include possible interactions agronomic variable-genes in the LASSO method, surprisingly similarities for all agronomic variables highly decreased (r<0.65, 48 d.f., ; see Figure S3C–D). Hence, we illustrated an alternative way to described accurately phenotypic properties of tomato fruit by using gene expression profile of the reduced set of RILs.
Next, to test the specificity of the inferred model parameters, we perturbed the target phenotypic profile for each RIL adding different levels of noise. Figure 2B shows the distance between predicted and measured values (green points) and mean correlations for different noise levels. A similar approach was performed by using the metabolic and gene expression profiles (red and blue points, respectively). Correlations with significance levels higher than the indicated above were not considered in the cross-validations. In addition, we estimated a very low mean error in predicting the agronomic properties across the training set (, see Methods).
Here, our main goal is to redesign the genome of tomato to generate an engineered surrogate that, if viable, would be easier to study and of greater potential biotechnological interest. Our design approach was inspired by the practice of in silico optimization over a predictive global model. Our next step was to test the possibility of improving agronomical properties of interest. We tested several scoring functions that fall into two global types: on the one hand, agronomical variables measured experimentally such as the number of fruits harvested per plant, the average fruit weight or its pH; and on the other hand, more complex fruit attributes that could be defined according to some of the components of the metabolic profile and are related to organoleptic properties of the fruit. In this later case, we first evaluated as proof of concept: fruit acceptability according to criteria based on acidity and sugars [28], quality as defined by the contribution of specific volatiles to aroma and by a reported [28] panel assessments of the tomato fruit and consequently on organoleptic acceptance. For this latter case we assumed a strong influence of a set of metabolites to be either maximized (-ionone, -damascenone, 2-phenylethanol and benzaldehyde) or minimized (methyl salicylate, guaiacol, hexanal, 1-penten-3-one and (E)-2-hexenal) using balanced weighting factors to account for their positive or negative contribution to quality. Moreover, all single metabolites were also optimized in single target analyses. Finally, a bi-objective function that included a high trade-off was proposed to optimize fruit quality and its production. As a first approach, we re-engineered tomato genome by perturbing independently the 959 genes included in the model, then we re-computed the scoring functions for all RILs enumerating all single knockouts and finally, all gene over-expression models were obtained.
Hence, mimicking the optimization patterns typical from LM, the landscape of desired agronomic properties of tomato fruit was exhaustively explored perturbing its effective transcriptional regulatory network (TRN) with single-gene alterations. Figure 3A shows the improvement of two of the agronomic properties mentioned above (fruit acceptability and quality vs production) as result of single gene perturbations according to our model. The success of the approach is shown by the efficiency function obtained for each transcriptional perturbation computed and which is defined by the normalized ratio between the agronomic property obtained for the re-engineered TRN and that for the wild-type TRN. Both agronomic properties and efficiencies in the case of single-perturbations were computed for each of the 169 RILs, resulting in a high variability between the lineages for all knockouts and over-expressed gene re-engineered TRN cases. We corroborated that there is a highly significant linear correlation (, for fruit acceptability and quality vs production) between the average value of the improved agronomic properties and the efficiencies reached across the set of RILs for all transcriptional perturbations. Both gene knockout and over-expression models resulted in similar linear regression slopes when considering acceptability and quality vs production together (0.05 and 0.24, respectively, Figure 3A). In addition, we also explored the possibility of tuning a given agronomic property towards a defined value, as it is desired for some biotechnological applications (see Text S1); achieving also in this case high efficiency values (Figure S4 and Tables S1 and S2).
After this, we ranked the list of knockout/over-expressed genes of the TRN according to two criteria directed to maximize: (i) the mean efficiency across all lineages in the case of goals such as acceptability and quality vs production; and (ii) the average of the maximum agronomic property reached by all possible TRN reconfigurations in the case of fruit quality (Dataset S2). Specifically, Table 1 shows the top 5 genes proposed for knockouts or over-expressed depending on the fitness evaluated. Fruit acceptability could be improved to 2.91% or 8.84% using gene knockout (i.e., LE24K20) or over-expression (i.e., LE13M10) in all lineages, respectively. By contrast, quality was highly increased achieving improvement ratios of 43.34% by gene knockout (i.e., LE24K20) and 227.31% by over-expression of LE15D07. Finally, taking into account not only the quality but also fruit production, ratios decreased to 15.32% (i.e., LE13F23) and 35.94% (i.e., LE14B20) using the two types of perturbations, respectively. Notice that all these rates of improvement were achieved in the lineages that provided maximum fitness in the wild-type TRN.
Lineages exhibited variability in their resistance to be optimized and this resistance changed with each target agronomic property. Figure 3B shows a strong linear dependence between the level of the agronomic property in the wild-type TRN and the average level of the agronomic properties resulting from all single perturbations in the TRN for each RIL (linear regression slope in the range 0.99–1.12 and , ). Interestingly, we observed that the effect of predicting agronomic properties under genetic perturbations was not dependent on the lineage selected. This provided a high level of robustness when we selected the lineages to implement experimentally re-designed TRN.
We computed the average number of single-gene perturbations to overcome an efficiency threshold given in the 169 RILs and the average probability of selecting the same gene-perturbation commonly for the whole set of RILs. The right panel in Figure 3C shows that only a few gene knockouts were able to improve fruit acceptability with a high probability in all lineages whereas, on the other hand, tens of gene knockouts could be proposed for increasing fruit quality and for the quality and production. On the other hand, the left panel in Figure 3C allowed re-asserting that re-engineering the TRN by gene over-expression could result in higher increments in the agronomic properties and with a higher density of suggested perturbations across the RILs.
The next step in our study was to propose new genome re-designs including multiple perturbations. To do this, we sampled widely the landscape of the acceptability, quality and quality vs production of tomato fruits by introducing two-gene perturbations either by knockouts and over-expressions (Dataset S3). Figure 4A shows the median efficiencies reached by two-gene transcriptional perturbations based on knockouts and over-expression in order to improve the agronomic properties defined as multiple-objective. As expected, we corroborated that multiple perturbations, located in different pathways (Table 2), could improve the agronomic properties significantly better than single perturbations. Table 2 lists the best gene-pairs to be used in perturbations that maximize such agronomic properties of the fruit. Figure 4B shows the average number of single gene perturbations that are able to overcome a given efficiency threshold for the top 5 RILs when ranked for single perturbations as well as the average probability of selecting the same multiple-perturbation commonly in a set of RILs.
After generating our predictive model for the TRN and metabolism of tomato fruit, we use it to automatically design tomato genomes with extreme alterations for each of the 56 volatile compounds by introducing a set of genetic perturbations. We compared sets of genetic perturbations for all pairs of volatile compounds and then inferred their levels of correlations (see Methods). Hence, these predicted correlations were compared to the levels of correlations obtained from the experimental values for each volatile pair that often reflects their belonging or not to the same metabolic/regulatory pathway or to be or not structurally related. Figure 4C–4F shows the predictive power of our model to determine correlations between all the volatile compounds. Interestingly, selecting a correlation cut-off between 0.5 and 0.8 we obtained high performance -scores (see Methods section) ranging between 0.32 and 0.91 (Figure 4D) for gene knockouts and between 0.31 and 0.80 when model selected genes by over-expression (Figure 4F). Notice that only pairs of experimental volatile compounds with were considered. Predictions decreased when we incorporated all pairs of compounds (Figure 4E–4F) indicating that our model captured high correlations observed experimentally with more precision. Figure S5 shows the dendograms of the volatile compound obtained from the correlation of experimentally obtained volatiles levels and the dendograms obtained using as distance between volatile compounds the number of common genetic perturbations proposed by the model. We observed that perturbations proposed by gene over-expression were pivotal to predict computationally significant distances between volatile compounds (Mantel test: r = 0.54, 1540 d.f., ) thus providing high support to our model. By contrast, predicted perturbations based on gene knockout could only identify a small fraction of the entire dendogram (Mantel test: r = 0.38, 1540 d.f., ).
To give further support to our model we constructed experimentally two inbred lines (ILs) derived from another interspecific cross whose transcriptome and metabolome were also experimentally measured. Parents of these ILs are a different cultivar of tomato M82 and a S. pennelli accession and therefore represent a completely different set of gene alleles from those in RILs used to construct the model. These ILs can be used as independent and useful test case to evaluate the validity of the model. We corroborated that a significant set of genetic perturbations suggested by computational design to optimize the phenotype observed were identified as genes differentially altered in the target phenotype (Text S1 and, Figure S6).
LM is a methodology that is being implemented by large industries to optimize their production. In the process of decision making applied to the redesign of production systems, firstly, engineers evaluate systematically the addition or elimination of resources in each of the participating single processes; afterwards, multiple changes are considered trying to achieve maximum quality and production [29]. Translating this engineering approach to a cellular molecular factory and identifying the basic functional elements has allowed us to develop a design methodology that optimizes the genome, resulting in a more desirable phenotypic properties. In addition, by mimicking the methodology from LM we have provided a first robust optimization to redesign an optimal genetic network based on the systemic exploration of the effects of a large number of single gene knockout and over-expression genotypes; then, a second multiple-optimization of random paths allowed improving substantially the desired agronomical properties. The success of this approach indicates that despite the existence of molecular interactions, the model is able to overcome this limitation and results in a good predictor.
We have proposed several re-engineered genomes that improve desired agronomic properties of the fruit by targeting single or multiple genetic modifications. It has been previously reported that single under-/over-expressed of certain genes may affect fruit quality traits, being these key genes involved in the biosynthesis of a product of fruit metabolism or to a general ripening regulators (i.e., carotenoids [30]). We have explored single perturbations by gene knockout or over-expression and our results indicated that a significantly better fine-tuning could be obtained by using over-expression approaches. We observed that improvement ratios could reach even more than 4-fold the wild-type value of most of phenotypes desired by designing genomes with only two genetic perturbations (Figure 4A and Table 2). The magnitude of the predicted change sometimes may appear low but an improvement in a quantitative trait, if consistent and predictable, maybe economically important. Indeed, a good combination of high yield with even slightly increased solid solids content is a major breeding goal for processing tomatoes that it is difficult to be achieved [31] because of polygenic nature and pleiotropic relationships of both traits [32].
Although it is not the objective of this paper, it does not escape our attention that some of the perturbations proposed are consistent with the biological processes associated to the trait and therefore the model could be used to reveal the molecular underpinnings of quality traits (see experimental evidences of each gene perturbation proposed by the model in the Dataset S4). For instance the role of YABBY (a gene proposed by our model to affect quality) in controlling fruit size probably through the auxin pathway and the effect of auxin in altering fruit growth and ripening has been previously reported [33], [34]. Similarly the importance of phytoene desaturase to affect carotenoids and carotenoid derived volatiles has been reported [35]. Most of the genes proposed by the models however are new, therefore opening new avenues of research either by targeting in transgenic plants, identification of mutants in those genes by TILLING [36] or by TAL engineering [37], as well as to be used as an additional guide during plant breeding. In principle these modifications are to be implemented in red fruit or around red fruit stage either genetically or by the use of external elicitors (physical or chemical) and our model provides roadmap for those approaches. Our methodology takes advantage of our ability to predict variations in fruit cell phenotype based on changes in the transcriptome. The linear relationships shown in Figure 3 (A, C, and D) guarantees that by optimizing our effective transcriptomic, metabolic or phenotypic fitness we are also optimizing the phenotype measured experimentally of the tomato fruits. While it is true that complex multi-organ organism such as tomato rely on the coordination and transport of multiple signals and nutrients from different parts of the plants to achieve the final phenotype, and this is especially true for the fruit [19], [38], it not less true that the most important part of the fruit characteristics at ripening depends basically on the fruit program before around the ripening stage [39], [40].
The ability to target redesign crops for enhanced content of metabolites of interest has been experimentally achieved in a number of cases (for instance vitamins C [41] and E [42]) using transgenic approaches and the information of bottlenecks or limiting steps for the biochemical pathways of the compounds of interest. The most dramatic examples of this have been introducing the new trait in a genetic background with very low value for it (i.e., golden rice [43]) using ectopic expression of one or several foreign genes. The use of natural genetic variability in combination with our nonbiased (hypothesis-free) modeling approach allows us to identify new candidate genes as potential targets to engineer the plant (although the biotechnological use of more active orthologs from other organisms is not discarded in our approach). The existence of regulatory networks connecting primary and secondary metabolism in plants should also be taken into consideration in future attempts to metabolically engineer the various classes of plant secondary metabolites [44]. It is interesting that known genes in the biosynthesis path often do not co-localize with quantitative trait locus for the metabolites in the path [35] indicating that there is ample of opportunities to be explored for metabolite and quality improvement, and our model fits nicely in this gap.
The construction of the tomato RILs used in this study has been described elsewhere [45]. Triplicate samples of red ripe fruits (each representing at least 5 fruit) from each of 169 RILs were harvested and analyzed for volatile compounds as described in [46]. For method validation, red ripe fruits from five ILs with a different genetic background [47] were used. Transcript profile datasets (11876×3×50 data points) were obtained from triplicate fruit samples of 50 selected RILs using TOM2 microarray, as previously reported [48]. Data sets corresponding to the rest of metabolites and phenomic data were obtained as in [46] from triplicate samples of the 169 RILs. To decrease experimental variability, the same fruits representing each RIL were homogenized and divided in different aliquot samples for the different metabolite or transcript profiling techniques. Before use all transcriptomic, metabolomic and phenomic data were normalized and transformed to log-scale. The ILs used for model validation have been described previously [21].
An effective linear model based on ODEs each providing the steady states of tomato fruit mRNA was used to describe transcriptional gene regulations [7]. Thus, the mRNA steady state from the gene, , is given by , where represents the regulatory effect that gene has on gene . Each gene expression value is contained in a range interval defined by the minimum and maximum value of all its experimental measurements obtained from the subset of 50 RILs used for transcript profiling. is a tunable parameter that decreases the gene expression range to improve the predictive capacity of the presented model under genetic predictions. The dynamics of metabolic profile was computed by , where is the steady-state concentration from the metabolite, is the regulatory strength that gene has on metabolite . Hence, agronomic variables () were predicted by means of a linear combination of the metabolic profile, , where is the regulatory effect that metabolite has on agronomic variable . , and are the perturbation terms that allow to calibrate gene expression, metabolic profiles and predicted agronomic properties, respectively, for all RILs. Notice that degradation coefficients of genes and metabolites (, respectively) scaled time conveniently and that we assumed the model in steady state ( and ).
Our global model consists of three blocks of algebraic equations covering respectively from gene expression, through metabolic profile until agronomic properties, and in all three cases the same methodology was applied. The inference procedure consisted of two nested steps. Firstly, the network connectivity was inferred by using the InferGene algorithm [7]. This method uses mutual information with a local significance value (z-score computation) to obtain the effective regulations. Hence, the potential interaction between a predictor and a target is z-scored, constituting an estimator of the likelihood of mutual information. Subsequently, we selected a z-score threshold for a predictor cutoff. In a second step, LASSO method was used to avoid over-fitting and to estimate the kinetic parameters of each effective model. Notice that the 8.7% of the selected genes in the TRN were annotated as TFs and 16.2% as encoding enzymatic activities and, in neither case, they were over-represented since both the tomato genome and the whole array contain similar fractions of TFs (8.8%) and enzymes (17.1%).
For the construction of the effective TRN model and its later integration with the metabolism, we used steady-state mRNA expression profiles derived from RILs transcriptionally and metabolically characterized. The dataset contains pre-processed expression data from 50×3 = 150 hybridization experiments using an array with 11876 probe sets spotted, and data for levels of 67 metabolites that were quantified over the same sample set. For this study, we only considered the 5592 genes whose expression values could be consistently found in more than 80% of the microarrays. We found 1057 TFs and 1962 genes with enzymatic activity after searching for the motifs transcription regulator and enzyme activity respectively in the functionally annotated tomato genome (Tom2). Moreover, all 169 RILs (including the previous 50 ones) for which we had metabolite and phenotype data were used to train a linear model able to predict agronomic properties of the fruit from potentially predictor metabolites. In all cases transcriptomic and metabolomic data were first normalized using the LOWESS procedure [26] and subsequently converted into z-scores across the RILs. In order to calibrate gene expression and metabolite concentration, both models included a perturbation term ( and , respectively) to fit all their -genes and -metabolites for a given RIL. We assumed a constant perturbation in the gene expression prediction because of its low variation across the training set (standard deviation of for all RILs is 0.072-fold the standard deviation of gene expression, ) with respect to the mean value, . Similarly, the average error to predict the metabolic profile across the training set was increased to .
Three plain text files containing the transcriptional, metabolic and phenotypic model for tomato were constructed and are available in Dataset S1. A directed network was constructed which places genes and metabolites as nodes and effective transcriptional and gene-metabolite interactions as edges. For the transcriptional interactions, edges link genes (including TFs, enzymes and genes without ability to regulate) to other genes or to a metabolite, in the case of metabolism.
Our algorithm searches possible reconfigurations of the global effective transcription regulatory network of tomato such as that the specified agronomic properties are improved (maximized or minimized) with respect to the properties of interest obtained in a given RIL. Different properties of interest have been optimized, ranging from single metabolites defining the sweetness or sourness of the fruit, to linear combinations of a set of metabolites determining the quality in terms of flavor and taste and even further to include objective functions that try to integrate two of those goals with a trade-off and balanced weighting factors such as fruit quality and yield.
We have addressed this optimization problem using two approaches. Firstly, we exhaustively enumerated all possible single gene knockouts and over-expression for each case to be optimized under a given selective pressure of interest. Second, we ranked all possible perturbations according to the new agronomic properties they would generate. The third step was to suggest genome reconfigurations that include multiple actions: gene knockouts, over-expressed genes, or both, in order to enlarge the combinatorial space of perturbed genomes. To do that, we have used an exhaustive method aimed at finding the global optimum in the space of all possible synthetic TRN. We started from the inferred model (see Mathematical model above) and applied an optimization scheme. At each step of the optimization process, we selected each gene among the ones involved in the transcriptomic model to evaluate the effect of three possible approaches (knockout, over-expression or wild-type scenario); we updated the model with the genetic perturbation that provided the best score. Note that to simulate knockout or over-expression in the gene , we substituted its ODE by the minimum () or maximum () values respectively observed in the range of diversity of the 50 RILs.
We computed the sets of single-gene perturbations, ?, by gene knockout or over-expression that alter significantly the levels of the 56 volatile metabolites representing the volatile compounds taking into account the global model. For the sake of the model we considered only those gene perturbations that would cause significant changes in the metabolite concentration higher than 1% (). can be divided into genetic modifications that increase () or decrease () the metabolite concentrations, respectively. Hence, correlations between metabolite pairs and () were calculated as the difference between and by using the set of single-gene perturbations proposed by the modelwhere and is the maximum normalized intersection predicted between the set of gene perturbations proposed by altering positively or/and negatively, respectively. We used these correlations to compute dendograms of all volatile compounds by using the distance inferred by the model () depending on the selected by gene knockout or over-expression.
The performance of the inferred metabolite correlations was evaluated using as a reference a set of empirical correlations previously obtained among these metabolites. We used different cut-offs, k, to identify metabolite correlations (). The fraction of metabolite pairs that were correctly predicted by the model (precision, ) and the fraction of all known correlations that were discovered by the model (sensitivity, ) were used to compute a performance statistic defined as .
To estimate the range of and statistics computed in the different cross-validations of the model, a bootstrap method was used. To this end, we generated 10000 random lists (with replacement) of metabolites/genes of size equal than the set of metabolites/genes proposed by the model as predictors of agronomic properties/metabolites/genes. Each of these random lists was then compared to the actual list of predictors proposed by the model and the corresponding and values computed to construct their expected null distributions. The observed and values were contrasted against these distributions and their significance assessed.
|
10.1371/journal.pgen.1007502 | Deletion of Nkx2-5 in trabecular myocardium reveals the developmental origins of pathological heterogeneity associated with ventricular non-compaction cardiomyopathy | Left ventricular non-compaction (LVNC) is a rare cardiomyopathy associated with a hypertrabeculated phenotype and a large spectrum of symptoms. It is still unclear whether LVNC results from a defect of ventricular trabeculae development and the mechanistic basis that underlies the varying severity of this pathology is unknown. To investigate these issues, we inactivated the cardiac transcription factor Nkx2-5 in trabecular myocardium at different stages of trabecular morphogenesis using an inducible Cx40-creERT2 allele. Conditional deletion of Nkx2-5 at embryonic stages, during trabecular formation, provokes a severe hypertrabeculated phenotype associated with subendocardial fibrosis and Purkinje fiber hypoplasia. A milder phenotype was observed after Nkx2-5 deletion at fetal stages, during trabecular compaction. A longitudinal study of cardiac function in adult Nkx2-5 conditional mutant mice demonstrates that excessive trabeculation is associated with complex ventricular conduction defects, progressively leading to strain defects, and, in 50% of mutant mice, to heart failure. Progressive impaired cardiac function correlates with conduction and strain defects independently of the degree of hypertrabeculation. Transcriptomic analysis of molecular pathways reflects myocardial remodeling with a larger number of differentially expressed genes in the severe versus mild phenotype and identifies Six1 as being upregulated in hypertrabeculated hearts. Our results provide insights into the etiology of LVNC and link its pathogenicity with compromised trabecular development including compaction defects and ventricular conduction system hypoplasia.
| During fetal heart morphogenesis, formation of the mature ventricular wall requires coordinated compaction of the inner trabecular layer and growth of the outer layer of myocardium. Arrested trabecular development has been implicated in the pathogenesis of hypertrabeculation associated with ventricular non-compaction cardiomyopathy. However much uncertainty still exists among clinicians concerning the physiopathology of ventricular non-compaction cardiomyopathy, including its clinical characteristics, prognosis, classification and even the definition of hypertrabeculation. In particular, distinguishing between pathological and non-pathological subtypes of non-compaction is currently a major issue. Here we show that deletion of the gene encoding the transcription factor Nkx2-5 at critical steps during trabecular development recapitulates pathological features of hypertrabeculation, providing the first model of ventricular non-compaction cardiomyopathy in adult mice. We demonstrate that excessive trabeculation due to failure of trabecular compaction during fetal development is associated with Purkinje fiber hypoplasia and subendocardial fibrosis. Longitudinal functional studies reveal that these mice present all the clinical signs of symptomatic left ventricular non-compaction cardiomyopathy, including conduction defects, strain defects and progressive heart failure. Our results, including transcriptomic analysis, suggest that pathological features of non-compaction are primarily developmental defects. This study clarifies the origin of the pathological outcomes associated with LVNC and may provide helpful information for clinicians concerning the etiology of this rare cardiomyopathy.
| Left ventricular non-compaction (LVNC) or hypertrabeculation is a cardiac anomaly characterized by a thickened sub-endocardial layer of ventricular myocardium with prominent trabeculations and deep recesses that communicate with the left ventricular (LV) cavity [1]. The recent classification of LVNC as the third most common form of cardiomyopathy is still debated because of the strong variability in pathophysiology, clinical symptoms and genetic associations [2, 3]. LVNC has been described as a genetic disorder caused by mutations in genes encoding sarcomere, cytoskeletal, nuclear membrane, ionic channels and chaperone proteins [4, 5]. Most importantly, LVNC can be asymptomatic but complications including heart failure (HF), thromboembolism and malignant arrhythmia are often observed [6]. The relatively recent interest of clinicians in LVNC comes from the increasing number of patients presenting signs of hypertrabeculation in the last 25 years, due to significant improvements in cardiac imaging using echocardiography or Magnetic Resonance Imaging (MRI). However, the phenotypic variability of this anomaly raises major problems for clinicians concerning its pathological nature. In particular, the wide spectrum of clinical signs makes prognosis of LVNC difficult. To understand the etiology of this non-compaction the generation of adult mouse models is essential.
The anatomic resemblance between human cases of LVNC and embryonic hearts have suggested that this anomaly results from an arrest of myocardial compaction during fetal development [7]. This hypothesis is now well established in several mouse models [8–11]; however, it does not explain the occurrence of different LVNC subtypes. The presence of ventricular trabeculae is assumed to be critical for oxygenation and conduction in the developing heart [12]. Trabeculation is initiated by the formation of endocardial outpockets in which trabecular myocardium develops by proliferation of cardiomyocytes forming a two-layer myocardium with a compact and a trabecular zone. In contrast to fishes and reptiles, mammalian trabeculae are not maintained in the adult heart, their reduction occurs by a process of compaction that initiates during fetal stages and ends after birth [13]. This is thought to arise by coalescence of trabeculae, which are progressively incorporated to the compact myocardium as well as forming papillary muscles [14]; however, the mechanisms underlying compaction of the ventricular trabeculae remain unknown. Moreover, trabecular development is directly coupled with the development of the ventricular conduction system. The function of the conduction system is to generate and orchestrate the propagation of the electrical activity through the heart in order to coordinate the sequential contraction of atria and ventricles [15]. The gap junction Connexin40 (Cx40) characterized by high conductance properties is strongly expressed in ventricular trabeculae and Cx40+ cardiomyocytes form preferential conduction pathways in the embryonic heart [16].
The cardiac transcription factor Nkx2-5 is an important regulator of ventricular trabeculae and conduction system development. Nkx2-5+/- haploinsufficient mice have an abnormal electrocardiogram, with a prolonged QRS and progressive elongation of the PR interval [17, 18]. These mice display ventricular conduction defects that can be correlated with hypoplastic development of the His-Purkinje system [19]. More recently, NKX2-5 mutations have been identified in LVNC patients indicating a role of this transcription factor in compaction [8, 20]. Indeed, conditional inactivation of Nkx2-5 in the ventricular myocardium showed that this deficiency provokes a hypertrabeculated phenotype resulting from defective myocardial proliferation [10]. In order to investigate the impact of the loss of Nkx2-5 during trabecular development, we have now conditionally inactivated this gene in the ventricular trabeculae in a spatiotemporally controlled manner using an inducible Cx40-creERT2 mouse model [21]. This study demonstrates that deletion of Nkx2-5 during trabecular development recapitulates in adult mice pathological features associated with LVNC and represents a highly tractable model to study the cellular and molecular mechanisms involved in the disease as well as the pathological evolution of LVNC.
Our strategy to conditionally inactivate Nkx2-5 in the developing ventricular trabeculae is summarized in Fig 1A. We crossed mice containing Floxed-Nkx2-5-Δneo alleles with Cx40-creERT2 mice in which Cre recombinase is broadly expressed in embryonic trabecular myocardium [21, 22]. Tamoxifen was injected into pregnant females at embryonic (E10.5-E11.5) or fetal (E13.5-E14.5) stages to remove Nkx2-5 during the formation of ventricular trabeculae (Nkx2-5ΔTrbE10) or during the step of trabecular compaction (Nkx2-5ΔTrbE14) respectively.
The efficiency of Nkx2-5 deletion was verified by immunofluorescence on sections from control, Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutant hearts. Numerous Nkx2-5-negative cardiomyocytes were observed in the apex, the papillary muscles and left ventricular free wall after tamoxifen injection at embryonic or fetal stages (Fig 1B). We performed a genetic tracing analysis of Cx40-cre derived cells to evaluate the extent of cardiac tissue affected by Nkx2-5 deletion. We crossed Cx40-cre mice with Rosa26-YFP reporter mice and observed the localization of YFP-positive cells in adult hearts by immunofluorescence on sections after tamoxifen induction at E10 or E14 (Fig 1C). YFP-positive cardiomyocytes were observed throughout the left ventricle, including compact zone myocardium, after induction at E10 while YFP+ cells are restricted to the inner half of the LV wall after induction at fetal stages, consistent with previous observations [16]. Almost no YFP staining was detected in the RV, in accordance with the low level of expression of Cx40 in this ventricle. These data are consistent with a contribution of ventricular trabeculae at embryonic and fetal stages to the morphogenesis of the papillary muscles and ventricular free wall with limited contributions to the interventricular septum. Quantification of Nkx2-5-positive subendocardial cardiomyocytes shows a significant reduction in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants after two consecutive tamoxifen injections at embryonic or fetal stages (Fig 1D). Consistent with the dynamic expression profile of Cx40 and our genetic tracing results, Nkx2-5 is deleted in a large proportion of the ventricular myocardium when tamoxifen was injected at embryonic stages, while deletion is restricted to the subendocardial zone at fetal stages (Fig 1E). These data suggest that the participation of Cx40-derived trabecular cells to the ventricular wall is reduced as trabecular compaction progresses during cardiac development.
Morphological analyses using high-resolution cine-MRI showed that the papillary muscles fail to coalesce and are enlarged and separated into multiple strands in adult Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 hearts (Figs 2A, S1A and S1B). In addition numerous trabeculations were observed in the left ventricular cavities of Nkx2-5ΔTrbE10 and to a lesser extent Nkx2-5ΔTrbE14 hearts (Fig 2A). We confirmed these results by histological analysis on transverse sections of adult hearts and showed that the compact zone of the myocardium is not affected, nor is the right ventricle (Figs 2B and S1C). To identify ventricular trabeculations, we performed co-immunofluorescence using antibodies labeling endothelial cells, Endoglin (Eng) and VEGFR2, to distinguish Eng+/VEGFR2- endocardium from Englow/VEGFR2+ coronary vascular endothelial cells [23]. Excessive trabeculations evidenced by extensive invaginations of the endocardium (arrows in Fig 2C) are found in both Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 compared to control hearts. Four-chamber views of these hearts reveal a distribution of these trabeculations throughout the base-apex axis (Figs 2D and S1). To quantify these hypertrabeculated phenotypes, we measured the length of the endocardium in sections through the cardiac apex and at the mid-ventricular level. Endocardial length was greater in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 compared to control hearts (Fig 2F). As the formation of coronary vasculature accompanies postnatal ventricular compaction [24], we investigated coronary artery density by counting the number of arteries identified by Cx40-RFP and SMA staining and observed no significant differences in artery density between these mice (S1C Fig). However, we observed the presence of endocardial cells defined by a strong expression of Endoglin (arrows) in the subendocardial myocardium (Figs 2C and S1C). These endocardial cells are organized in islets, are negative for VEGFR2 and are not associated with SMCs. Ex-vivo perfusion of the fluorescent lectin WGA-Cy3 (Wheat Germ Agglutinin) into the left ventricle of control and Nkx2-5ΔTrbE10 hearts demonstrated that endocardial islets are in direct communication with the endocardium, in contrast to WGA-negative coronary vasculature (Fig 2D and 2E). These results strongly suggest that endocardial islets correspond to deep endocardial invaginations due to defective trabecular compaction. Quantification revealed a significantly larger number of endocardial islets in Nkx2-5ΔTrbE10 mutant hearts than in control or Nkx2-5ΔTrbE14 hearts (Fig 2G). Together, these results suggest that excessive trabeculations detected in Nkx2-5 mutant hearts result from defects in trabecular compaction or coalescence. Furthermore, deletion of Nkx2-5 in embryonic versus fetal trabeculae results in increased phenotypic severity.
We investigated ventricular conduction system (VCS) development after conditional Nkx2-5 inactivation, by whole-mount immunofluorescence using the VCS marker Contactin-2 (Cntn-2) in opened LV preparations of adult control, Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE4 hearts [25]. In control hearts, Cntn-2 is detected in the entire VCS including the atrioventricular bundle (AVB), left bundle branch (LBB) and Purkinje fibers (PF) network (Fig 3A). In mutant hearts after Nkx2-5 deletion at embryonic or fetal stages, the PF network is strongly reduced while the LBB and AVB develop normally (Figs 3A and S2). We used the software “Angiotool” to quantify the complexity of the PF network [26]. In a branched structure, such as a vascular tree or the PF network, this image analysis software calculates the number of branching events as well as the length and density of vessels/fibers. Using Angiotool we observed a marked reduction of PF density and branching in Nkx2-5 mutant compared to control hearts (Fig 3B and 3C). However, while the PF network appears less affected in Nkx2-5ΔTrbE14 compared to Nkx2-5ΔTrbE10 mutants (Fig 3A), this difference was not scored as significant using Angiotool.
In order to study the Nkx2-5 conditional mutant cardiac phenotype in more detail, we performed immunofluorescence on transverse sections using WGA, a lectin used to quantify cardiac fibrosis and cardiomyocyte size [27]. Extensive subendocardial fibrosis was detected in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 hearts while no evident signs of fibrosis were observed in control mice (Figs 3D, S1D and S3). Interstitial fibrosis was quantified by measuring the percentage of fibrotic area in the LV and revealed a more extended fibrosis in Nkx2-5ΔTrbE10 than Nkx2-5ΔTrbE14 mutant hearts (Fig 3F). WGA staining also revealed a hypertrophic phenotype in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants (Fig 3D-a,a’). The number of cardiomyocytes/field is significantly reduced in these two mutants compared to control (Fig 3E). This hypertrophic phenotype is independent of Nkx2-5 expression as it was observed in both Nkx2-5-positive and Nkx2-5-negative, i.e. non-deleted as well as deleted cardiomyocytes (arrowheads in Fig 3D-a’).
In summary, anatomical and morphological criteria demonstrate that conditional Nkx2-5 loss of function in Cx40 expressing trabecular myocardium results in excessive trabeculations. This phenotype is accompanied by trabecular compaction defects including absence of papillary muscle coalescence, deep protrusions of Endoglin+ cells resulting in endocardial islets, hypoplasia of the PF network, extensive subendocardial fibrosis and cardiac hypertrophy. Moreover, distinct severe and mild phenotypes were observed following Nkx2-5 deletion using Cx40-cre at embryonic and fetal stages respectively, suggesting a relationship with the extent of trabeculae-derived myocardium in which Nkx2-5 has been deleted.
To investigate the embryonic basis of the hypertrabeculated phenotype, we analyzed cardiomyocyte mitotic activity in the left ventricle of E14.5 control and Nkx2-5ΔTrbE10 hearts by PH3 immunofluorescence on sections (Fig 4A and 4B). Quantification revealed a slight excess of PH3+ nuclei in cardiomyocytes of mutant hearts, which is not restricted to the trabecular zone, suggesting a global increase of proliferation in ventricular cardiomyocytes. Using Cx40-Cre::R26-YFP genetic tracing of the same fetal hearts, we observed no difference in the distribution of YFP+ cells in mutants in either atria and ventricles (Fig 4B). Moreover, in situ hybridization using probes specific for the compact and trabecular zones, Hey2 and ANF (Nppa) respectively, revealed the formation of a clear boundary in both E14.5 control and Nkx2-5ΔTrbE10 hearts (Fig 4C). The compact-trabecular transition at fetal stages is thus unaffected in mutant hearts, suggesting that the excessive trabeculation results from both defects in cardiac proliferation and impaired ventricular compaction.
The above results show that conditional Nkx2-5 loss of function using Cx40-cre results in new mouse models of severe and mild ventricular non-compaction. We next carried out a longitudinal study of the same animals in order to document the evolution of the hypertrabeculated phenotypes with age and the consequence of these morphological defects on cardiac function. A follow up of cardiac morphology and function was performed in individual mice over a year using cardiac MRI, echocardiography or ECG recordings.
In order to compare trabecular morphology in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants, we used high-resolution cine-MRI, a highly sensitive non-invasive imaging technique. High-resolution cine-MR images revealed the presence of trabeculations in the LV of Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants compared to the smooth endocardium observed in control hearts (Fig 5A). These results are consistent with our histological findings (Figs 2B and S1). Follow-up analysis of the same animals from 3 to 10 month-old showed indistinguishable cardiac structures at all timepoints in term of number and size of papillary muscles or trabeculations, suggesting no morphological evolution of the hypertrabeculated phenotype with age. However, several cardiac parameters measured from these MR images showed significant differences between groups and/or age (Table 1). In the line with cardiac cell hypertrophy observed in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants, LV mass is increased, particularly in Nkx2-5ΔTrbE10 hearts. At the functional levels the stroke volume is lower and the ejection fraction (EF) is reduced in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 compared to control hearts revealing serious contractility defects (Fig 5B and Table 1). Interestingly, the individual follow-up showed that several mutants exhibited fluctuating EF with age (red and green dots in Figs 5B and S4); for example, an EF measured at 35% at 3 months, 33% at 6 and increased to 62% at 9 months in the same Nkx2-5ΔTrbE10 mouse. Movies created from short-axis cine series of these different hearts show irregular deformations of the LV during cardiac cycles and reveal that these defects are aggravated with age, revealing problems of contractility in these mutants (S1–S4 Movies).
To better appreciate the myocardial deformation, we carried out speckle-tracking based strain imaging analysis by echocardiography. At 3 months, longitudinal and radial strain measurements in the long axis were identical in all mice while radial and circumferential strain measurements in the short axis were elevated in Nkx2-5ΔTrbE10 hearts compared to control or Nkx2-5ΔTrbE14 mutants (Fig 6A). In contrast, both Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants presented a decreased strain in all axes in old mice compared to age-matched control animals (Fig 6B). These results highlight the defects of myocardial deformation, already detected in young mice when the hypertrabeculation phenotype is most severe. In addition, the variation of strain with age is consistent with the variability of EF measured in both mutants. Interestingly, a number of mice with a severe or mild hypertrabeculation phenotype, presented signs of HF with an EF<40% suggesting that altered cardiac function is independent of the level of excessive trabeculation. Secondary to the HF phenotype, RV dysfunction was detected in 9 month-old Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mice but not in 3 month-old mice (S5 Fig). Moreover, both mutants presented a reduced EF in old mice which is highly correlated with strain in all axes supporting the fact that strain is a good parameter to estimate cardiac function and HF (Fig 6C).
In parallel, we performed a longitudinal study in the same group of mice to investigate the cardiac electrical activity by measuring surface electrocardiograms (ECGs). The analysis of ECG traces revealed major defects in the QRS complex from 3 month-old in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants (Fig 7A). Monophasic R and R' waves were early observed in both mutants and Notched R wave was first detected in Nkx2-5ΔTrbE10 mutant mice. Fragmented QRS, the more drastic defect, was only recorded in Nkx2-5ΔTrbE10 mutant mice (Fig 7A). At 9 months, none of the Nkx2-5 mutants had a normal QRS in contrast with control mice that never showed QRS defects. Analysis of time intervals and wave amplitudes is summarized in Table 2. Consistent with observations of QRS shape, broader QRS intervals with smaller amplitude of R waves were found in Nkx2-5ΔTrbE10 mutant and to a lesser extent in Nkx2-5ΔTrbE14. These results illustrate bundle branch blocks (BBB) in both mutants. Moreover, QRS intervals duration correlates fully with EF, measured by cardiac MRI and echography, and LV mass (Fig 7D). Indeed, animals with the lowest EF had the broadest QRS intervals. Note that only cardiac activation was disturbed. The repolarization phase, represented by QT intervals and T wave amplitude, was not impacted (Fig 7A and Table 2). Moreover, increased PR intervals, indicative of 1st degree atrioventricular block in accordance with the atrial deletion of Nkx2-5, were evidenced as previously reported in Nkx2-5 mutants [10, 28]. Because ECGs were recorded under anesthesia, the β1-agonist dobutamine (DOB) was used to challenge the hearts. Heart rate was identical in all untreated animals (Table 2), and increased after treatment by 45 and 47% at 6 and 9 months respectively, in control. In contrast, Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants displayed chronotropic incompetence after DOB injection; heart rate increased by only 15–21% and 25–36% (in 6–9 month-old mice respectively) in line with a decreased cardiac reserve particularly in Nkx2-5ΔTrbE10 (Fig 7B and 7C). However, arrhythmia and a severe susceptibility to ventricular fibrillation were unveiled in both mutants but not in control mice (Fig 7C).
Together, these data demonstrate that severe and mild hypertrabeculation phenotypes are coupled with severe defects in cardiac activation and contractility. Hypertrabeculation associated with fibrosis, hypertrophy and Purkinje fibers hypoplasia is linked with early conduction defects, leading progressively to strain defects and HF. This cardiac phenotype is similar in both mutants, though delayed in Nkx2-5ΔTrbE14 mice.
In order to identify transcriptional changes associated with hypertrabeculation and functional defects, we performed a microarray analysis on LV tissue including the ventricular free wall and apex from Nkx2-5ΔTrbE10, Nkx2-5ΔTrbE14 and control adult hearts. Comparison between both mutants and the control showed 622 differentially expressed probes, corresponding to 469 genes among the 39,430 mouse genes screened. A comprehensive list of the significantly misregulated genes is presented in a supplementary file. The hierarchical clustering in both dimensions (samples and genes) showed clear differences between the 3 groups (Fig 8A). Seven clusters, from an unsupervised hierarchical clustering based on the similarities of expression, were analyzed using Gene Ontology (GO). The most significant Biological Processes (BP) are associated with cardiovascular development, differentiation, inflammation, apoptosis and ionic transport (Fig 8A) and totally support the morphological and functional data. Major genes known for their role in LVNC or other cardiomyopathies and genes previously reported in Nkx2-5 mutants were found to be deregulated in both mutants (Fig 8A).
Comparison of gene profiles revealed that conditional inactivation of Nkx2-5 in ventricular trabeculae at embryonic stage resulted in the deregulation of many more genes than at fetal stage: 601 vs 349 respectively, 26% of which were common, as shown on the Venn diagram in Fig 8B and in the lists of the supplementary file. Specifically up-regulated genes after Nkx2-5 deletion vs control were twice as numerous as down-regulated genes in Nkx2-5ΔTrbE10 adult hearts, 64% of which were up-regulated and 36% down-regulated. In Nkx2-5ΔTrbE14 hearts, 84% of them were up-regulated by the deletion of Nkx2-5 at fetal stage while 16% were down-regulated (Fig 8B). Interestingly, GO analysis of differentially expressed genes revealed that the same biological processes were involved in both mutants but the number of genes within each process was greater in Nkx2-5ΔTrbE10 hearts. Therefore these data confirm that Nkx2-5 deletion is associated with substantial transcriptional changes causing a deep remodeling of the heart, and are consistent with the observation of a stronger phenotype in Nkx2-5ΔTrbE10 hearts.
In order to obtain a comprehensive overview of biological processes associated with hypertrabeculation after Nkx2-5 deletion, an integrative network was created with cytoscape (Fig 8C). While most of the BP were common to both embryonic and fetal deletions, a few pathways related to cardiac morphogenesis and cell adhesion were specific to Nkx2-5ΔTrbE10 mutant while Nkx2-5ΔTrbE14 was associated with cardiac differentiation and hypertrophy terms. Moreover, the majority of genes were transcription factors (TF) involved in a variety of developmental processes. This network illustrates clearly the impact of Nkx2-5 deletion on fundamental processes including morphogenesis and cell differentiation.
To validate the microarray data, we performed qPCR experiments for 14 genes that were among the most up- or down-regulated genes, selected on the basis of their cardiovascular-related function. Among the up-regulated genes in Nkx2-5ΔTrbE10 vs control, v-Maf and Six1 transcripts were quantified and Prkcz as a down-regulated gene. Mcpt4 was investigated as an up-regulated gene in Nkx2-5ΔTrbE14 vs control hearts. We also analyzed genes in the common part of the Venn diagram including Bmp10, Timp1, Cd207, Vsig4, Myh7, Vav-1, Rorb, Gcgr and Vgll2 that are up-regulated in Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 vs control hearts. RT-PCR analysis confirmed the microarray results with regard to fold-changes and significance and highlighted the important differential expression of Six1 between Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants (Fig 9A). We further investigated this using a Six1-LacZ reporter line [29]. X-gal staining in the LV was restricted to the compact zone in control mice while the staining was increased and extended to the trabecular zone in Nkx2-5ΔTrbE10 mutant mice (Fig 9B). Pecam1 co-immunostaining revealed previously unreported Six1 expression in endothelial cells as well as in cardiomyocytes with high density of positive cells in the compact zone of control hearts (Fig 9C1-3). In Nkx2-5 conditional mutant hearts, an excessive number of β-gal positive cells, i.e. cardiomyocytes and endothelial cells was observed in the entire ventricular wall (Fig 9C1-3). This result validated the up-regulation of Six1 in Nkx2-5 loss of function hypertrabeculated hearts.
Together, our results show that deletion of Nkx2-5 using Cx40-cre at embryonic and fetal stages disturbs the normal development and differentiation of the cardiovascular system. Moreover, Nkx2-5 deletion in ventricular trabeculae provokes more molecular disturbances in adult hearts when it occurs at embryonic rather than fetal stages. Nevertheless, the same BP are affected in both cases, fully supporting the morphological and functional defects documented in these hypertrabeculated hearts. Interestingly, Six1 expression is found in the trabecular zone only in hypertrabeculated hearts suggesting that Six1 is a potential genetic marker of LVNC.
In this study, we have analyzed the pathological consequences of excessive ventricular trabeculation, induced by spatio-temporal deletion of Nkx2-5 during trabecular development, on adult cardiac function. The loss of Nkx2-5 at embryonic stages resulted in a severe hypertrabeculated phenotype associated with excessive subendocardial fibrosis, while these phenotypes were milder when deletion of Nkx2-5 occurred at fetal stages. Our analysis revealed that the excessive trabeculation phenotype is more severe when Nkx2-5 deletion occurs at embryonic rather than at fetal stages, consistent with progressive reduction of Cx40-derived trabecular cell incorporation in the myocardial wall during ventricular morphogenesis [16]. Moreover, this phenotype was stable in old mice, suggesting that the level of hypertrabeculation is associated with the broader extension of Nkx2-5 deletion in the ventricular myocardium rather than with aging. Another feature linked to hypertrabeculation was the subendocardial fibrosis observed in Nkx2-5 conditional mutant hearts. This result is in agreement with a recent report that subendocardial fibrosis is commonly observed in LVNC patients [30]. As in the case of hypertrabeculation, subendocardial fibrosis is more pronounced in Nkx2-5ΔTrbE10 compared to Nkx2-5ΔTrbE14 hearts. This is consistent with previous studies showing the absence of fibrosis in hearts after conditional knockout of Nkx2-5 after mid-gestation, at perinatal or postnatal stages [31, 32]. We observed that numerous genes associated with fibrosis are specifically upregulated in the LV after embryonic deletion of Nkx2-5 and less affected at fetal stages. Together, these data strongly suggest that subendocardial fibrosis originates from an early event due to a dysfunction of ventricular trabeculae development rather than as a consequence of aging, as has been suggested in human patients [30]. This data is of major interest as subendocardial fibrosis may contribute to altered longitudinal and axial contraction, measured in strain analysis, and to BBB and fibrillation [33–35]. Our mouse models of hypertrabeculation thus mirror the clinical situation and support the conclusion that the degree of non-compaction results primarily from developmental defects.
Ventricular compaction is a poorly understood step of cardiac morphogenesis. Using two independent genetic tracing mouse models, it has been recently demonstrated that embryonic cardiomyocytes of the trabecular or the compact zone mix in a hybrid zone, which covers a large proportion of the compact myocardium in the adult heart [36]. Abolishing proliferation in the compact zone leads to a hypertrabeculation phenotype that is not observed if proliferation is altered in the trabecular zone only [36]. These data suggest that non-compaction arises as a result of proliferation defects of cardiomyocytes in the compact zone. Several other mouse models including the ventricular Nkx2-5-conditional deletion support this assumption; however, most of these mutants present a very thin compact layer and are embryonic lethal [10, 13]. In contrast, in our Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutant hearts, hypertrabeculation was not associated with a reduction of the compact layer. However, we were able to reproduce excessive trabeculation by the conditional deletion of Nkx2-5 in ventricular trabeculae demonstrating that disruption of trabecular development represents another mechanism leading to hypertrabeculation. Moreover, our data suggest that hypertrabeculation results primarily from trabecular coalescence defects. This model is supported firstly by the disruption of the molecular boundary between the trabecular and compact zone shown by Six1 misexpression, which has been recently associated with right ventricular associated hypertrabeculation and fibrosis in other mouse mutants [37, 38]. While this compact-trabecular transition initiates normally in our mutants at fetal stages, proliferation defects may compromise late development of trabecular compaction. Secondly, the numerous endocardial islets detected in Nkx2-5ΔTrbE10 mutants are highly similar to the intertrabecular spaces connected to the LV observed in LVNC patients [3]. This endocardial phenotype suggests a defect in the formation of coronary vasculature from the endocardium that normally occurs at perinatal stages [24]. Moreover, concomitant to the compaction step, the acquisition of a spiral pattern of cardiomyocytes develops as a prerequisite for twisting contraction [2, 39]. Our strain analysis points out important defects in myocardium deformation in hypertrabeculated hearts suggesting that compaction defects plays a role later for the efficiency of cardiac contractions. All these results highlight the role of Nkx2-5 in the complex modifications in ventricular morphology that take place during trabecular compaction and their requirement for normal cardiac function.
A major characteristic of our Nkx2-5-hypertrabulated mouse models is the occurrence of the numerous cardiac complications such as conduction blocks, contraction defects and HF, often associated with LVNC in symptomatic patients. Because mutations in NKX2-5 are associated with a myriad of congenital heart diseases (CHD) in humans, this transcription factor has received much attention for its role in cardiac morphogenesis [20]. The atrial-specific knockout of Nkx2-5 produced atrial hyperplasia, atrial septal defects and bradycardia [40]. In comparison, deletion of Nkx2-5 in atrial myocardium induced by the Cx40-creERT2 line is heterogeneous and gives rise only to a modest effect such as a slight increase of the PR interval; in particular, no atrial anatomical defects or reduced heart rate were detected in our mutant mice. Moreover, a direct role of Nkx2-5 on cardiac contractility has already been described. Indeed, reduced and irregular ventricular contractions are recorded by echocardiography in newborn mice after conditional deletion of Nkx2-5 at E13.5 [32]. In these mouse models, Nkx2-5 perturbs calcium handling and sarcomere organization, which are known trigger to reduce contractility [10, 17, 31, 41]. In Nkx2-5 hypertrabeculated hearts, we found numerous deregulated genes, involved in calcium signaling (Ryr2, Sln or SERCA2), the contractile apparatus (Myh7, Tnnt2), transmembrane ionic channels (HCN1, HCN4, Scn5a, Cacna1g, Cacna1h, Kcne1), and cardiac development, including genes of the Notch pathway and its targets, (Notch3, Dll1, Dtx1, Hey1, BMP10) (transcriptomic analysis and qPCR experiments (S6 Fig)). These genes have been previously described to be deregulated in different Nkx2-5 mutant mice and in some cases, including Myh7, HCN4 [32], Tnnt2 [42], Scn5a [43], Ryr2 [44] and Notch signaling [9], have been associated with LVNC. It has recently been shown that cardiac contractility may also depend directly on endothelial signaling through the NRG-1/Erbb2 pathways [45]. Indeed, Pecam-1 null mice display a slight decreased EF without morphological defects in cardiomyocytes or capillary densities but resulting from a disturbed communication between endothelial cells and cardiomyocytes. In our study, hypertrabeculated hearts present vascular defects including endocardial islets, which correlate with upregulation of Endoglin transcript levels. Indeed, LVNC patients are known to present defects in cardiac perfusion and embolic thrombosis is a pathological feature of non-compaction cardiomyopathy [46, 47]. This is consistent with the microarray data in which genes involving endothelial cells signaling or vasculature are found deregulated in mutant hearts. Finally, these changes in gene expression are likely to modulate cardiomyocyte activity that impact directly on cardiac contractility and pump function as measured by reduced strain rate in line with decreased EF. Intriguingly, we found fluctuations in strain and EF at different ages during the follow-up of the cardiac function in few hypertrabeculated hearts. Recently, an “undulating” phenotype in which the associated cardiomyopathic features change has been evocated in children with LVNC [3, 48]. For instance, a dilated and hypertrophic presentation with poor cardiac function changed to a hypertrophic, hypercontractile form of non-compaction and the final destination was dilated dysfunctional form of non-compaction with HF [49]. These observations highlight the extreme variability in contractility of hypertrabeculated hearts and argue in favor of an independent form of cardiomyopathy. The measurements of strain in LVNC patients confirm that cardiac deformation represents a valuable parameter for prognosis in cases of excessive trabeculations [50].
We documented the first longitudinal study of the conduction defects in a mouse model with hypertrabeculation. In our Nkx2-5 hypertrabeculated hearts, a total or quasi-absence of PF is observed and these mice display early conduction defects, in particular LBBB consistent with this hypoplasia phenotype. Indeed, PF connect to the sub-endocardium at Purkinje-muscle junctions sites in humans to form the origin of myocardial activation [51]. Consistent with our results, disruption of the spatiotemporal expression of ion channels has been shown to induce defective propagation of impulse from endocardium to epicardium, resulting in BBB and leading to increased susceptibility to fibrillation [52]. However, these ventricular conduction defects worsened during aging suggesting that QRS shape and duration also depend on other parameters such as fibrosis and hypertrophy, which are known to impair conduction. To support this hypothesis, we found a correlation between LV mass and QRS duration. While the exact role of Nkx2-5 in ventricular conduction system development is unknown, numerous genetically modified mouse models have revealed the importance of this transcription factor for normal cardiac conduction at postnatal stages [8, 31, 32, 53]. Hypoplasia of the ventricular conduction system has not been reported in other mouse models of non-compaction or in human samples suggesting that this phenotype may be specific to this transcription factor [9, 54]. However, the presence of LVNC is significantly associated with a rapid deterioration of LV function and higher mortality when associated with abnormal ECG measurements [55]. Finally, we observed that hypertrabeculated hearts have impaired chronotropic competence and are susceptible to ventricular fibrillation under conditions of stress. This is consistent with the appearance in LVNC patients of complications such as exercise intolerance or sudden death linked to excessive trabeculation in athletes after extensive exercise [2, 3]. Our mouse data mirror the heterogeneity of the pathological outcomes among LVNC patients and suggest that LVNC has a higher incidence of cardiac death when associated with cardiac dysfunction and arrhythmias. Furthermore, these data highlight the importance of Nkx2-5 in the development of the ventricular conduction system and the role of these defects in the pathological outcome of LVNC.
Our molecular analysis identified numerous genes related to cardiac dysfunction, and highlights the pleiotropic roles of Nkx2-5 during trabecular development. Among those genes, we observed differential upregulation of Six1 between the Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mutants suggesting a role of this factor in the severity of the hypertrabeculated phenotype. Six1 is a transcription factor identified for its role for skeletal muscle and cardiac progenitor cell development [29, 37]. Our immunohistological analysis revealed upregulated Six1 expression in endothelial cells and cardiomyocytes of the trabecular zone, suggesting that deregulation of this gene is associated with cardiac pathology rather than being a direct target of Nkx2-5. Interestingly, Six1 has been recently found to be deregulated in cardiomyopathy induced by Nkx2-5 point mutation or in human patients [56, 57]. Our data suggest that Six1 may be a good marker for pathological forms of excessive trabeculation in non-compaction cardiomyopathy.
In conclusion, impaired trabecular compaction provokes a hypertrabeculated phenotype associated with a disturbed endocardial vasculature, VCS hypoplasia and subendocardial fibrosis (Fig 9D). Our data strongly suggest that complications associated with hypertrabeculation such as cardiac hypertrophy and HF arise progressively and may be a direct consequence of the conduction defects and abnormal cardiac contractility of hypertrabeculated myocardium. Interestingly, these mice present a progressive impaired cardiac function while excessive trabeculations are morphological identical throughout adult life, supporting the idea that this anomaly does not per se trigger the pathology in LVNC. This study clarifies the origin of the pathological outcomes associated with LVNC and, although further developments are clearly warranted, may provide helpful information for clinicians in the future for the diagnostic and prognostic evaluation of left ventricular non-compaction patients.
Animal procedures were approved by the ethics committee for animal experimentation of the French ministry (n° 01055.02). Cx40-creERT2, R26-YFP and Six1-LacZ mouse lines were genotyped as previously reported [21, 29, 58]. A conditional null allele of Nkx2-5 was generated as described (Nkx2-5flox) [22]. To remove the Frt-flanked neomycin cassette present in this allele, Nkx2-5flox mice were bred with a ubiquitous flipase mouse line (ACTB:FLPe) [59]. The Floxed-Nkx2-5-∆Neo allele was used in the subsequent breedings with the Cx40-creERT2 mouse line. To delete the Floxed-Nkx2-5-∆Neo gene, tamoxifen was injected intraperitoneally to pregnant females (200μl) for two consecutive days. Tamoxifen (T-5648, Sigma) was dissolved at the concentration of 20mg/ml in ethanol/sunflower oil (10/90). After tamoxifen treatment of pregnant females, newborn mice were recovered by caesarian section and given for adoption to CD1 females. Fifty mice were assigned to three groups: Nkx2-5ΔTrbE10 mice received tamoxifen injections at embryonic stages (E10.5 and E11.5), Nkx2-5ΔTrbE14 mice received tamoxifen injections at fetal stages (E13.5 and E14.5), and control mice are littermates with Nkx2-5-∆Neofl/fl:Cx40+/+ genotype.
Macroscopic examination of the internal surface of the ventricles was previously described [60]. For histological studies, adult hearts were dissected, fixed for four hours in 4% paraformaldehyde (vol/vol) in PBS, washed in sucrose gradient, then embedded in OCT and cryosectioned. To quantify interstitial fibrosis, transverse sections were counterstained with wheat germ agglutinin-Cy3 (WGA-Cy3 from Sigma-Aldrich) as described previously [27]. For immunofluorescence, sections were permeabilized in PBS 1X / 0.2% Triton X100 for 20 min and incubated for 1 hour in saturation buffer (PBS 1X / 3% BSA / 0.1% Triton X100). Primary antibodies were incubated in saturation buffer overnight at 4°C. Secondary antibodies coupled to fluorescent molecules were incubated in saturation buffer and after washes, hearts were observed under a Zeiss Apotome microscope.
For whole-mount immunofluorescence, adult hearts were pinned on petri dish to expose the LV and fixed in 4% paraformaldehyde for 2 hours at 4°C, washed in PBS, permeabilized in PBS 1X / 0.5% Triton X100 for 1h and incubated for 3 hours in saturation buffer (PBS 1X / 3% BSA / 0.1% Triton X100). The primary antibodies were incubated in saturation buffer for 24 hours at 4°C. Secondary antibodies coupled to fluorescent molecules were incubated in saturation buffer and after washes, hearts were observed under a Zeiss LSM780 confocal microscope.
The measure of the endocardial length (mm) was carried out using Image J software by drawing the contour of the endocardium (Eng1+/VEGFR2-) lining the LV on transversal sections and the mean of 3 sections per heart was calculated (n = 3 mice of each group). The number of PH3+ cells per LV section was quantified using Zeiss Zen software from LSM780 confocal images. The PH3+ cardiomyocytes are stained for Nkx2-5+ or YFP+. The percentage of PH3+ cardiomyocytes over the number of PH3+ cells was calculated from the total number of cells of 3–4 sections per heart (n = 3 embryos of each group).
Antibodies used in this study are specific to Nkx2-5 (Sc8697 Santa-Cruz), GFP (AbD Serotec), RFP (Rockland), Contactin-2 (AF1714 R&D system), Pecam-1 (MEC13.3-BD Pharmingen), Endoglin CD105 (MJ7/18-DSHB), VEGFR2 (AF644-R&D SYSTEMS), α-Smooth Muscle Actin (F3777-SIGMA).
After dissection, adult hearts were rapidly intubated with a 24G Surflo IV catheter in the aortae and perfused with 1ml of PBS heparin to remove blood, then 200μl of WGA-Cy3 was introduced in the LV through the same catheter and incubated for 30 minutes before fixation in 4% PFA. After fixation, hearts are processed as described above.
Non-radioactive in situ hybridization on sections from E14.5 hearts were performed as described [61] using Hey2 and ANF mRNA probes.
MRI was carried out every two months on the same animal groups from 2 to 12 months-old mice. The experiments were performed on a Bruker Biospec Avance 4.7 T/30 imager (Bruker Biospin GmbH, Ettlingen, Germany) (France Life Imaging network), as previously described [62]. Anesthesia was maintained during MRI with 1.5–2% Isoflurane in a constant flow of room air (270 ml/min) through a nose cone using a dedicated vaporizer (univentor anaesthesia unit, univentor high precision instrument Zejtun, Malta). Temperature was maintained at 39°C. Breath and heart rate were monitored and signals were used to trigger MRI acquisition using a monitoring and gating system (SA Instruments, Inc. Stony Brook, NY, USA). Cine-MRI (ECG gating, repetition time 5 ms, echo time 1.51 ms, flip angle 20°, slice thickness 1 mm, pixel size 0.195×0.195 mm2) were performed in short axis view. Ten phases per heartbeat were acquired from base to apex to cover the whole LV. In addition, one high resolution cine series (ECG gating, repetition time and frame rate 15 ms, echo time 1.68 ms, flip angle 30°, slice thickness 1.1 mm, pixel size 0.086×0.086 mm2) was acquired in short axis view at the mid base-apex location to evaluate trabecular morphology.
Surface ECGs were performed on anesthetized mice. An induction with 5% isoflurane was followed by maintenance at 1 to 2% in a constant flow of oxygen at 700 ml/min. ECGs were recorded every two months from 2 to 12 months using a bipolar system in which the electrodes were placed subcutaneously at the right (negative) and left forelimb (reference) and the left hindlimb (positive) for lead II and at the right (reference) and left forelimb (negative) for lead III. Electrodes were connected to a Bioamp amplifier (AD Instruments) and were digitalized through an A/D converter ML 825 PowerLab 2/25 (AD Instruments). Digital recordings were analyzed with Chart software for windows version 5.0.2 (AD Instruments). Events were registered to 100 K/s and were filtered to 50 Hz. ECG recordings were obtained for 1 min after stabilization of the signal. Post-analysis was performed for heart rate, PR, QRS, QT intervals, T, R and S durations, T, R, S and QRS amplitudes.
A Dobutamine (DOB) stress-test was performed on selected 6 and 9 month-old mice. After recording of ECG as described above, mice received an intraperitoneal single injection of DOB at a dose of 0,75 μg/g body weight or an equivalent volume of vehicle (0,9% NaCl). The DOB dose was selected after a bibliographic study on the criteria of a moderate increase of heart rate and a preliminary experiment using the dose of 0,75 μg/g in which two volumes of injection were tested. The lowest volume was chosen and DOB was prepared as follows:
DOB hydrochloride powder (Sigma-Aldrich, product number D-0676) was dissolved in ddH2O and vortexed to provide a 10 mg/ml DOB stock solution. DOB working solution (0,1 μg/μl) was prepared by 1:100 dilution of the stock solution in sterile 0,9% NaCl.
Echocardiography was performed using a Vevo 2100 ultrasound system (VisualSonics, Toronto, Canada) equipped with a real-time micro-visualization scan head probe (MS-550D) operating at a frame rate ranging from 740 frames per sec (fps). Mice were anesthetized with isoflurane (IsofloH, Abbott S.A, Madrid, Spain) at a concentration of 3.5% for induction and between 1 to 1.5% for maintenance during the analysis with 100% Oxygen. Each animal was placed on a heated table in supine position with extremities attached to the table through four electrocardiographic leads. The chest was shaved using a chemical hair remover (Veet, Reckitt Benckise, Granollers, Spain). Ultrasound gel (Quick Eco-Gel, Lessa, Barcelona, Spain) was applied to the thorax surface to optimize the visibility of the cardiac chambers. The heart rate (HR) and respiratory rate of mice were recorded during the echocardiographic study. Two mice were excluded from the study due to very low HR. Echocardiograms were acquired at baseline. Left ventricular (LV) characteristics were quantified according to the standards of the American Society of Echocardiology and the Vevo 2100 Protocol-Based Measurements and Calculations guide, as described in the following paragraphs. LV diameters were measured on a two-dimensional (B-mode) parasternal long axis and short axis view. The functional parameters of the heart were evaluated based on LV diameter measurements.
Total RNA was isolated from LVs (left ventricular wall with apex and without the interventricular septum) of 6 month-old control, Nkx2-5ΔTrbE10 and Nkx2-5ΔTrbE14 mice (n = 4 for each group) using TRIzol Reagent according to the manufacturer’s procedure (Invitrogen, Life Technologies). The concentration of RNA was determined by reading absorbance at 260 nm on NanoDrop (ND-1000, ThermoScientific). The quality of RNA was confirmed on the Agilent 2100 Bioanalyzer (Agilent technologies, Germany) with the Agilent RNA 6000 Nano chips. The samples with the RIN (RNA Integrity Number) closed to 10 were used for microarray. The SurePrint G3 Mouse GE8x60K Microarray Kit (Agilent Technologies, Santa Clara, CA) containing 62 976 oligonucleotide probes representing 39 430 genes was used. Quantile normalization was applied to sample data to correct for global intensity and dispersion. Then, a filtering at 75% was used to keep only genes expressed over the background noise, so 22 114 probes. A Significant Analysis of Microarray (SAM) 3 classes, with 10 000 permutations was applied with a False Discovery Rate (FDR) at 5% to look for mutant specific variation in gene expression in the dataset. Hierarchical clustering of the differentially expressed genes was obtained with TM4 Microarray Software Suite V4.9 (https://www.tm4.org) using average linkage clustering metrics and Pearson correlation for the distance. We identified enriched functional annotations for the clusters using DAVID Bioinformatics Resources 6.7 (https://david.ncifcrf.gov). A SAM 2 classes with a FDR at 5% was performed for Venn diagrams in TM4 and Cytoscape (http://www.cytoscape.org) was used to design the gene network with genes whose fold change is ≥ 2. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus [63] and are accessible through GEO Series accession number GSE113251 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE113251).
On the same cDNA samples used for microarray experiments, we performed quantitative real-time PCR (qPCR) analyses of genes and housekeeping genes using SYBR Green PCR master mix (Applied Bio-systems, Life technologies, UK) on the Stratagene Mx3000P (Agilent Technologies).
Data are expressed as means ± standard error of the mean (SEM). Significant differences between groups were determined using one-way or two-way analysis of variance (ANOVA) followed by Sidak post-hoc testing with Graphpad Prism software (Graphpad Prism 7.0, La Jolla, CA, USA). A p value <0.05 was considered statistically significant.
|
10.1371/journal.pntd.0004057 | Assessment of Efficacy and Quality of Two Albendazole Brands Commonly Used against Soil-Transmitted Helminth Infections in School Children in Jimma Town, Ethiopia | There is a worldwide upscale in mass drug administration (MDA) programs to control the morbidity caused by soil-transmitted helminths (STHs): Ascaris lumbricoides, Trichuris trichiura and hookworm. Although anthelminthic drugs which are used for MDA are supplied by two pharmaceutical companies through donation, there is a wide range of brands available on local markets for which the efficacy against STHs and quality remain poorly explored. In the present study, we evaluated the drug efficacy and quality of two albendazole brands (Bendex and Ovis) available on the local market in Ethiopia.
A randomized clinical trial was conducted according to the World Health Organization (WHO) guidelines to assess drug efficacy, by means of egg reduction rate (ERR), of Bendex and Ovis against STH infections in school children in Jimma, Ethiopia. In addition, the chemical and physicochemical quality of the drugs was assessed according to the United States and European Pharmacopoeia, encompassing mass uniformity of the tablets, amount of active compound and dissolution profile. Both drugs were highly efficacious against A. lumbricoides (>97%), but showed poor efficacy against T. trichiura (~20%). For hookworms, Ovis was significantly (p < 0.05) more efficacious compared to Bendex (98.1% vs. 88.7%). Assessment of the physicochemical quality of the drugs revealed a significant difference in dissolution profile, with Bendex having a slower dissolution than Ovis.
The study revealed that differences in efficacy between the two brands of albendazole (ABZ) tablets against hookworm are linked to the differences in the in-vitro drug release profile. Differences in uptake and metabolism of this benzimidazole drug among different helminth species may explain that this efficacy difference was only observed in hookworms and not in the two other species. The results of the present study underscore the importance of assessing the chemical and physicochemical quality of drugs before conducting efficacy assessment in any clinical trials to ensure appropriate therapeutic efficacy and to exclude poor drug quality as a factor of reduced drug efficacy other than anthelminthic resistance. Overall, this paper demonstrates that “all medicines are not created equal”.
| Soil-transmitted helminths (STHs) infect millions of children worldwide. To fight STH, large-scale de-worming programs are implemented in which anthelmintic drugs (either albendazole (ABZ) or mebendazole (MEB)) are administered. However, there is a wide range of other brands, which are even more accessible, but for which the efficacy and quality remain poorly explored. We evaluated efficacy against STHs and quality of two ABZ brands commonly available on the local markets in Ethiopia (Bendex and Ovis). Both brands showed high efficacy against roundworm infections and poor efficacy against whipworms. However, for hookworm infections, Bendex was significantly less efficacious than Ovis. In terms of drug quality, a significant difference was observed in the dissolution profile, with Bendex having a significantly slower dissolution rate than Ovis. Since dissolution behavior is critical for a drug to be appropriately absorbed into the helminth (through host-blood and/or parasite-cuticle) and produce therapeutic efficacy, the poor dissolution of Bendex compared to Ovis can explain the observed difference in efficacy against hookworms. Our results emphasize the importance of periodically assessing of drug quality to ensure appropriate therapeutic efficacy and to exclude poor drug quality as a potential factor of reduced drug efficacy other than drug resistance.
| Currently, there is a worldwide upscale in the implementation of programs to control and to eliminate a selection of 10 neglected tropical diseases [1, 2]. Among these, soil-transmitted helminthiasis causes the highest burden on public health. It is estimated that more than 1.4 billion people were infected with at least one of the four STH species: the roundworm Ascaris lumbricoides, the whipworm Trichuris trichiura and the two hookworm species Necator americanus and Ancylostoma duodenale, resulting in a global burden of approximately 5.2 million disability-adjusted life years (DALYs) (20% of the total number of DALYs attributable to neglected tropical diseases) [3, 4]. To control the morbidity caused by STH, mass drug administration (MDA) of a single oral dose of a benzimidazole anthelminthic drug (ABZ or MEB) is recommended in communities where the prevalence of any STH exceeds 20% [5].
To date, major pharmaceutical companies such as GlaxoSmithKline (ABZ, Zentel) and Johnson and Johnson (MEB, Vermox) are donating these medicines to WHO, which subsequently distribute these medicines to its recipient countries. In Ethiopia, the donated medicines are made available for the patients through government hospitals and health centers. The therapeutic efficacy of these products at the WHO-recommended dosages (i.e. single dose of 400 mg ABZ or 500 mg MEB) has recently been evaluated in two consecutive multinational trials [6, 7]. These trials showed that the therapeutic efficacy measured in terms of egg reduction rate (ERR), varied both between drugs and STH species: both drugs showing high efficacy against A. lumbricoides (> 98%) and poor efficacy against T. trichiura (~64%), and ABZ being more efficacious against hookworms compared to mebendazole (96% vs. 80%). In addition to these two donated brands, there is a wide range of other brands available on local markets of STH endemic countries, which are often more accessible to the local people, but for which the efficacy or quality remain poorly explored [8]. The latter is particularly important in countries where resources are limited to monitor the quality of drugs, and hence in which prevalence of substandard, falsified or illegal drugs is substantial [9–14]. Although the quality of medicines has a direct influence on therapeutic efficacy, this remains poorly studied for benzimidazole anthelminthic drugs against STH infections. Therefore, we assessed both the efficacy and quality of two brands of ABZ commonly administered for the treatment of individual STHs in Ethiopia, namely Bendex and Ovis.
The study protocol was approved by the Ethical Committees of Jimma University (Ethiopia) (reference no RPGC/282/2014) and of the Faculty of Medicine, Ghent University (Belgium) (ref. no 2013/1114; B670201319330). The study is registered under clinicaltrial.gov identifier number NCT02420574 (https://clinicaltrials.gov/ct2/show/NCT02420574?term=NCT02420574&rank=1). The school authorities, teachers, parents, and the children were informed about the purpose and procedures of the study. The written consent form was prepared in two commonly used local languages (Afaan Oromo and Amharic) and handed over to the children’s parents/guardians after explaining the aim, confidentiality and entire procedure of the clinical trial. Only those children (i) who were willing to participate and (ii) whose parents or guardians signed the written informed consent form were included in the study. Moreover, an additional separate written informed consent form for children older than 12 years was prepared, read and handed over to them and their additional written informed consent obtained (S1 Checklist).
Samples of the two ABZ brands (Bendex, India, CIPLA Ltd, batch no: x21253 and Ovis, Korea, DaeHWa Pharmaceuticals, batch no: 2020) with a label claim of 400 mg/tablet and expiry date of November 2015 were purchased from private community pharmacy in Jimma town, Ethiopia.
The quality of the drugs was evaluated by investigating three efficacy-critical quality attributes: (i) the mass uniformity, (ii) the amount of the active compound, and (iii) the dissolution of the tablets.
In total, 679 subjects were recruited of which 418 subjects were enrolled and randomized across the two brands of ABZ (nBendex = nOvis = 209). T. trichiura was the most prevalent (69.4%), followed by A. lumbricoides (53.6%). Hookworm infections were found in 28.2% of the subjects. In total 388 subjects completed the trial (nBendex = 197; nOvis = 191), resulting in a compliance rate of more than 90%. There was no significant difference in mean age (Bendex: 10.3 years vs. Ovis: 10.3 years, p = 1.00), sex ratio (Bendex: 1.07 vs. Ovis: 0.87, p = 1.00) and mean fecal egg count (FEC) (A. lumbricoides: Bendex: 8,706 egg per gram of feces (EPG) vs. Ovis: 7,935, p = 0.69; T. trichiura: Bendex: 909 EPG vs. Ovis: 769, p = 0.45; hookworm: Bendex: 355 EPG vs. Ovis: 335, p = 0.79) between the two arms at baseline. Both brands showed high efficacy against A. lumbricoides (Bendex: 98.7% vs. Ovis: 97.8%, p = 0.64), and poor efficacy against T. trichiura (Bendex: 24.4% vs. Ovis: 20.4%, p = 0.81). For hookworm infections, Ovis was more efficacious than Bendex, though the difference was marginally significant (Bendex: 88.7% vs. Ovis: 98.1%, p = 0.05). Based on the WHO criteria to classify the efficacy of anthelminthic drugs (Table 1), both brands had satisfactory and reduced efficacy against A. lumbricoides and T. trichiura, respectively. For hookworms, Ovis had a satisfactory efficacy, whereas Bendex had a doubtful efficacy. A pairwise comparison of baseline parameters and drug efficacy between Bendex and Ovis are presented in S1 Table.
Assessing the quality and in-vivo efficacy differences between different brands of ABZ tablets are very critical to avoid indiscriminate use of different brands that could influence intended therapeutic outcomes. In the present study, we evaluated comparative in-vivo efficacy and in-vitro quality of two commonly available brands of ABZ tablets (Bendex and Ovis) that are used to treat STH infection.
The in-vivo efficacy results of two brands of ABZ against A. lumbricoides and T. trichiura determined in terms of ERR, suggest the susceptibility difference between the two STHs. Therapeutic efficacy of antihelmnthics can be influenced by various factors such as infection intensity and susceptibility of parasites. Thus the reduced efficacy of both brands against T. trichiura observed in the present study could be due to high level of infection intensity [17] and/or genetic modification of beta-tubulin gene [26, 27]. Since concentration of API at the target site of the parasites could be low due to metabolism and/or limited absorption of the drug by the parasite [28, 29], the reduced efficacy of both brands against T. trichiura might also be associated with the pharmacokinetics of ABZ in the parasite. The reduced efficacy of the two brands against T. trichiura observed in the present study is comparable to the results reported in the previous studies [6, 17]. The present finding, i.e. high prevalence of T. trichiura among other STHs in the study area together with the reduced efficacy results in ERR observed for single dose of ABZ 400 mg tablets, is supported by various literature findings [30–33]. This emphasizes the urgent need for alternative drugs and/or development of novel anthelmintic drugs to tackle this efficacy problem.
Medicines quality is a critical factor that could affect efficacy of drugs against parasites [34] and for biopharmaceutical classification system (BCS) class II [35] drugs like ABZ that have low solubility and high permeability, dissolution is the rate-limiting step for drug absorption. Hence, in-vivo/in-vitro correlation between blood concentration profile and dissolution profile may be expected. Since the bioavailability of ABZ to the host is very low and also shows variability between individuals [36], a decreased dissolution could significantly worsen bioavailability, which in turn diminishes in-vivo efficacy of both the parent drug and therapeutically active metabolite (ABZ sulphoxide). Moreover, the capacity of anthelminthics to dissolve appropriately is an essential characteristic that allows proper drug uptake by the parasites and therefore assures the appropriate drug efficacy. Therefore, the four times decreased dissolution of Bendex compared to Ovis (Fig 2) which could influence both local and systemic concentration is a plausible explanation for the efficacy difference between the two brands against hookworms, which are blood sucking parasites. Previous studies already indicated the differences in uptake and metabolism of benzimidazole drugs among different helminth species [28, 29], which may explain that the efficacy difference between the two brands was only observed against hookworms.
Though mass uniformity and content of API per tablet are critical quality attributes that could influence efficacy, comparable quality of both brands with respect to mass uniformity and ABZ content observed in the present study (Table 2) explain the efficacy difference between the two brands against hookworms is not associated with mass uniformity and content of API.
Considering a single point dissolution specification for ABZ tablets as described in the USP, i.e. Q ≥ 80% dissolved in 30 min [22], there is a statistically significant (p ≤ 0.05) difference observed between Bendex (Q = 20%) and Ovis (Q = 85%) (Fig 2). Also, the area under the dissolution curve between defined time points (0, 15, 30, 45 min) or DE quantifies the poor in-vitro dissolution of Bendex. While Ovis thus complied to the USP dissolution specifications, Bendex on the contrary did not. Although an undesirable polymorphic solid state of albendazole could be one of the reasons for the difference in dissolution behavior of API [37–39], the IR spectra (Figs 3 and 4) of Bendex and Ovis showed no significant difference between the two brands. The absence of significant observed shifts in IR-absorption indicates the similarity in polymorphic form of ABZ in both brands. Thus, the significant difference in dissolution observed between the two brands could not be associated with different polymorphic forms of ABZ. Whatever the reason of dissolution difference is, e.g. excipient and processing manufacturing conditions and stability, the poor dissolution behavior of Bendex observed in the present study is in accordance with the previous reports in which 41 samples out of 72 samples of solid oral dosage forms including ABZ tablets, different generic formulations of albendazole tablets and carbamazepine immediate-release products did not comply with the established acceptance criteria [40–42]. In general, it is important to note that quality of medicines could be one of the factors influencing outcomes of clinical trials. For instance, literature indicates the association of poor quality of locally manufactured antimalarial drugs: Sulfadoxine-Pyrimethamine with clinical failure of malaria treatment in Pakistan [43]. Therefore, the results of the present study and a recent report by Newton and his colleagues [44] point to the requirement of guidelines for quality assurance of medicines used in clinical trials. Subjects lost at follow-up per STH species were the limitations of this study.
In conclusion, this study demonstrated that the two investigated brands of ABZ tablets are efficacious against A. lumbricoides and hookworm while both brands had reduced efficacy against T. trichiura. However, there was a significant difference between the two brands of ABZ against hookworm. While both brands showed comparable tablet mass uniformity and albendazole content, the in-vitro dissolution release profile between the two brands was significantly different, explaining the clinical efficacy difference observed. The results of the present study underscore the importance of assessing the chemical and physicochemical quality of drugs before conducting efficacy assessment in clinical trials to ensure appropriate therapeutic efficacy and to exclude poor drug quality as a factor of reduced drug efficacy other than anthelminthic resistance. Our in-vivo efficacy study clearly indicates the importance of appropriate quality medicines.
|
10.1371/journal.pntd.0000698 | Human Complement Regulators C4b-Binding Protein and C1 Esterase Inhibitor Interact with a Novel Outer Surface Protein of Borrelia recurrentis | The spirochete Borrelia recurrentis is the causal agent of louse-borne relapsing fever and is transmitted to humans by the infected body louse Pediculus humanus. We have recently demonstrated that the B. recurrentis surface receptor, HcpA, specifically binds factor H, the regulator of the alternative pathway of complement activation, thereby inhibiting complement mediated bacteriolysis. Here, we show that B. recurrentis spirochetes express another potential outer membrane lipoprotein, termed CihC, and acquire C4b-binding protein (C4bp) and human C1 esterase inhibitor (C1-Inh), the major inhibitors of the classical and lectin pathway of complement activation. A highly homologous receptor for C4bp was also found in the African tick-borne relapsing fever spirochete B. duttonii. Upon its binding to B. recurrentis or recombinant CihC, C4bp retains its functional potential, i.e. facilitating the factor I-mediated degradation of C4b. The additional finding that ectopic expression of CihC in serum sensitive B. burgdorferi significantly increased spirochetal resistance against human complement suggests this receptor to substantially contribute, together with other known strategies, to immune evasion of B. recurrentis.
| Borrelia recurrentis, the causal agent of louse-borne relapsing fever is transmitted to humans via infected body lice. Infection with B. recurrentis has been achieved only in humans and is accompanied by a systemic inflammatory disease, multiple relapses of fever and massive spirochetemia. A key virulence factor of B. recurrentis is their potential to undergo antigenic variation. However, for survival in the blood during the early phase of infection and for persistence in human tissues, spirochetes must be endowed with robust tools to escape innate immunity. We have recently shown that B. recurrentis acquires the serum-derived regulator factor H, thereby blocking the alternative complement pathway. Here, we show that B. recurrentis expresses in addition a novel outer surface lipoprotein that selectively binds serum-derived C4b-binding protein and C1 esterase inhibitor, two endogenous regulators of the classical and lectin pathway of complement activation. The combined data underscore the versatility of B. recurrentis to effectively evade innate and adaptive immunity, including serum resistance. Thus, the present study elucidates a new mechanism of B. recurrentis important for its evasion from complement attack and will be helpful for the development of new drugs against this fatal infection.
| B. recurrentis, the causative agent of louse-borne relapsing fever is transmitted to humans by contamination of abraded skin with either hemolymph from crushed, infected lice (Pediculus humanus humanus) or excreted feces thereof [1], [2]. The last century has seen multiple epidemics of louse-borne relapsing fever in Europe, with high mortality rates of up to 40%. Louse-borne relapsing fever has been epidemic in Africa throughout the 20th century with foci persisting in the highlands of Ethiopia [3], [4]. Clinically, louse-borne relapsing fever is characterized by a 5- to 7-day incubation period followed by one to five relapses of fever, and spirochetemia [5], [6]. Spontaneous mortality remains as high as 2–4% despite antibiotics, with patients suffering from distinctive hemorrhagic syndrome and/or Jarish-Herxheimer reactions [7].
To survive in human tissues, including blood, B. recurrentis has to escape innate and adaptive immune responses. Complement is a major component of first line host defense with the potential to eliminate microbes. However, pathogens have evolved strategies to evade complement-mediated lysis, either indirectly, by binding host-derived regulators to their surface or directly, by expressing endogenous complement inhibitors [8], [9]. In fact, we and others have recently demonstrated that tick- and louse-borne pathogens, i.e. B. hermsii and B. recurrentis, specifically bind complement regulatory proteins, i.e. CFH and CFHR-1, via their outer surface lipoproteins FhbA, BhCRASP-1 and HcpA, respectively [10]–[14]. Surface bound CFH was shown to interfere with the alternative complement pathway by inhibiting complement activation via accelerating the decay of the C3 convertase and inactivating newly formed C3b [15], [16].
However, complement may also attack pathogenic bacteria via the classical pathway, i.e. by interacting with previously bound antibodies, resulting in deposition of the membrane attack complex on the surface of bacteria and their final death [17]. The classical pathway is initiated by the binding and activation of the C1 complex, consisting of C1q, C1r and C1s. C1q can bind to clustered IgG and IgM bound to the surface of bacteria, and also directly to many bacteria through lipoteichoic acids or other structures [18], [19]. When C1q binds, its associated proteases, C1r and C1s, become activated and form the activated C1 complex, which cleaves C4 and C2 to generate the C3 convertase. The lectin pathway is initiated when mannose-binding lectin (MBL) or ficolins bind carbohydrates on the surface of a microbe [20]. A key endogenous regulator of the classical and lectin pathway is serum-derived C4b-binding protein (C4bp). C4bp is a cofactor in factor I-mediated cleavage of C4b to C4d and interferes with the assembly and decay of the C3-convertase (C4bC2a) of the classical and lectin pathway [21], [22]. It was recently shown that acquisition of the regulators CFH and C4bp on the surface of B. recurrentis and B. duttonii contributes to serum resistance in vitro [17]. However, the respective receptors on the spirochetal surface have not been identified.
It was thus the aim of the present study to identify and characterize the putative receptor for C4bp of B. recurrentis and B. duttonii. Here, we show for the first time that B. recurrentis and B. duttonii express a novel potential outer surface lipoprotein, which specifically binds C4bp and in addition C1-Inh. The finding that pathogen-bound C4bp retains its co-factor activity suggests that this process contributes to the exceptional resistance of the two spirochetes species to bactericidal activity of human serum.
Relapsing fever spirochetes B. recurrentis strains A1 and A17, B. hermsii (ATCC35209) strain HS1, B. duttonii strain LA, B. parkeri RML, B. turicatae RML (provided by Tom Schwan, Rocky Mountain Laboratories) and the Lyme disease spirochete B. burgdorferi strains ZS7 and B313, a clonal mutant of B31 lacking all linear and circular plasmids with the exception of cp32-1, cp32-2, cp32-4, cp26 and lp17 [23], [24], were cultivated in BSK-H complete medium (Bio&Sell, Feucht, Germany) supplemented with 5% rabbit serum (PAN Biotech, Freiburg, Germany) at 30°C. Bacteria were harvested by centrifugation and washed with phosphate-buffered saline. The density of spirochetes was determined using dark-field microscopy and a Kova counting chamber (Hycor Biomedical, Garden Grove, CA). E. coli JM109 were grown at 37°C in LB medium.
All human plasma and serum samples used in this study were purchased from the Heidelberg University blood bank. Human plasma obtained from 20 healthy, anonymous blood donors without known history of spirochetal infections were pooled and used as source for C4bp. Nonimmune human serum (NHS) was acquired from healthy donors with no prior history of Borrelia spp. infection. Factor B-depleted human serum was purchased from Complement Technology, Inc. (http://www.ComplementTech.com).
C4bp protein was purified from pooled human plasma by barium citrate precipitation as described [25]. Briefly, following extensive dialysis the solution was subjected to ion exchange chromatography using Q-Sepharose (GE Healthcare) and proteins were eluted with a gradient of 0 – 2 M NaCl. C4b, C1-Inh and factor I were purchased from Calbiochem. Purified C4bp, C1-Inh and BSA were conjugated to biotin with No-Weigh Biotin-NHS (Pierce Biotechnology).
Isolation of the C4bp binding protein of B. recurrentis was carried out by co-immunoprecipitation. Whole cell lysates of B. recurrentis were prepared as described elsewhere with minor modifications [10]. Briefly, cultures were grown at 33°C in modified BSK medium to the late-log phase and harvested by centrifugation at 6.000×g for 10 min at 4°C. The resulting pellets were washed twice with PBS, resuspended in ice-cold 50 mM Tris-HCl (pH 7.5), 25 mM KCl, 5 mM Mg2Cl, 1mM EGTA, 0.5% NP40 and rotated for 1 h at 4°C. Cell debris were removed by centrifugation and for pre-clearing lysates were incubated with protein G sepharose (GE Healthcare) for 1 h at 4°C. For immunoprecipitation pre-cleared B. recurrentis lysates were incubated with protein G Sepharose previously loaded with anti-C4bp antibody and purified human C4bp for 12 h at 4°C with gentle agitation. After washing in 50 mM NaH2PO4, 300 mM NaCl, 10 mM imidazole (pH 8) bound proteins were eluted with 2x SDS sample buffer (Serva) and subjected to 14% Tris/Tricine SDS-PAGE under reducing conditions. Immunoprecipitates were separated by SDS-PAGE and visualized by staining with colloidal Coomassie (Pierce/Thermofisher, Bonn, Germany). The selected protein band of 40 kDa was cored from the gel and subjected to MALDI mass spectrometric analysis as previously described [26]. Recently, the genome of the selected B. recurrentis strain A1 was sequenced [27]. The identified peptide matched an open reading frame of 1071 bp of the B. recurrentis A1 genome, named cihC. The gene encoding CihC was amplified by PCR using primers CihC F and CihC R (Table 1), cloned into pGEM-T Easy vector (Promega, Mannheim, Germany) and sequenced by using the BigDye terminator cycle sequencing kit (PE Applied Biosystems). The resulting plasmid pGEM-BrCihC was used as template for construction of expression plasmids by PCR amplification. For recombinant full-length CihC protein, primers CihC Bam and CihC HincII were used. For N- and C-terminal deletion mutants, these primers were applied in combination with CihCΔ83F, CihCΔ122F, CihCΔ160F, CihCΔ149R, CihCΔ190R, CihCΔ260R, and CihCΔ294R (Table 1) resulting in recombinant proteins CihCΔ20–260, CihCΔ83–294, CihCΔ122–294, CihCΔ20–190, CihCΔ83–149, and CihCΔ160–294, respectively. The ORF encoding CihC of B. duttonii (CihCBD) was amplified using genomic DNA of B. duttonii strain La in combination with oligonucleotides CihC Bam and CihC Hinc. After digestion with restriction enzymes BamHI and HincII, PCR fragments were ligated in frame into the His6-tag encoding sequence into vector pQE-30Xa (Qiagen, Hilden, Germany). For expression of N-terminal His-tagged fusion proteins, the plasmids were transformed into E.coli strain JM109 and recombinant proteins were purified as recommended by the manufacturer (Qiagen).
Monoclonal antibody BR2, directed against CihC was generated by immunization of Balb/c mice with whole cells of B. recurrentis A1 according to a method described elsewhere [28]. All animal research was approved in advance by the Laboratory Animal Committee of the University of Heidelberg (RP Karlsruhe 35-9185.82/A-25/07). The animals were kept in a filter cabinet and given food and water ad libitum, with all maintenance performed according to German animal welfare guidelines.
To prepare whole cell lysates Borrelia were centrifuged and washed three times with PBS. Cells were resuspended in BugBuster Master Mix (Merck) and lysed for 5–10 min on ice. Borrelial whole cell lysates (15 µg) or purified recombinant CihC proteins (200 ng) were subjected to Tris/Glycine-SDS-PAGE under reducing conditions and transferred to nitrocellulose as previously described [29]. Briefly, after transfer of proteins onto nitrocellulose, nonspecific binding sites were blocked using 5% (w/v) dried milk in TBS (50 mM Tris-HCl pH 7.4, 200 mM NaCl) for 2 h at room temperature (RT). Subsequently, membranes were rinsed two times in TBS and incubated for 1 h at RT with NHS (1:1 diluted in TBS) or purified C4bp. Membranes were washed four times with 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.2% Tween20 (TBST) and incubated for 1 h with either peroxidase-conjugated anti-C1-Inh (Linaris) or anti-C4bp antibody (Quidel, San Diego). Following four washes with TBST, blot strips were incubated with a secondary peroxidase-conjugated anti-mouse IgG antibody (Dako, Glostrup, Denmark) for 1 h at RT. Detection of bound antibodies was performed using the enhanced chemiluminescence ECL Western blotting detection reagent and ECL Hyperfilms (GE Healthcare, Amersham).
For Western blot analysis, membranes were incubated for 1 h at RT with either anti-C4c antiserum (Dako), anti-C1s (Atlantic antibody), anti-CihC (mAb BR2) or anti-flagellin (mAb LA21) monoclonal antibodies [30]. For detection of purified recombinant CihC full-length protein and deletion mutants, the anti-His6-tag (Calbiochem) antibody was employed.
Southern blotting of total genomic DNA was done as previously described [31]. Briefly, 250 ml of Borrelia cultures were centrifuged, washed twice in PBS and resuspended in 9 ml of TE (10 mM Tris pH 7.5, 1 mM EDTA) buffer. Subsequently, 20% SDS (1 ml) and 20 mg/ml proteinase K (50 µl) was added and incubated for 1 h at 37°C. NaCl (5 M) and Hexadecyl-trimethyl-ammonium-bromide (10%) was added followed by incubation for 10 min at 65°C. DNA was extracted twice with phenol-chloroform-isoamyl ethanol (25∶24∶1) and DNA was precipitated with 0.6 volume of isopropanol. The precipitates were washed with 70% ethanol and resuspended in H2O. 10 µg of total genomic borrelial DNA was prepared as agarose blocks, loaded into the agarose gels and fractionated by pulse-field gel electrophoresis (PFGE) in combination with the CHEF-DR II System (Bio-Rad, Germany). Hybridization with a random primed cihC gene probe was conducted as described [32].
Spirochetes (1×107) were washed with Tris buffer (30 mM Tris, 60 mM NaCl, pH 7.4) and incubated with mAb directed against CihC (mAb BR2) or flagellin (mAb LA21) for 1 h at RT. Spirochetes were then washed with Tris buffer/0.1% BSA, spotted on coverslips and allowed to air-dry for 1 h. After methanol fixation, samples were dried for 15 min and incubated for 1 h in a humidified chamber with Cy3-labeled rabbit anti-mouse IgG (1/200, Dianova). Cells were visualized at a magnification of 1000x using a Nikon Eclipse 90i upright automated microscope and images were obtained using a Nikon DS-1 QM sensitive black and white CCD camera at a resolution of 0.133 µm/pixel.
Cells of B. recurrentis strain A1 were treated with proteases using a modified, previously described method [33]. Briefly, intact borrelial cells were incubated with either proteinase K or trypsin to a final concentration of 0 -12.5 µg/ml. Borrelial cells were then lysed and equal volumes (20 µl) were separated by SDS-PAGE (13%). Proteins were visualized by Western blotting using specific monoclonal antibodies.
Functional activity of C4bp was analyzed by measuring factor I-mediated conversion of C4b to iC4b. Either 100 µl of CihC (0.5 µg/well) or intact B. recurrentis A1 spirochetes were coated onto microtiter plates (MaxiSorp, Nunc) and incubated with purified human C4bp (50 µg/ml) for 1h at RT and after washing, C4b (4 µg/ml) and factor I (2 µg/ml) were added and incubated at 37°C for up to 2 h. Supernatants were removed from the wells, subjected to SDS-PAGE (10%) under reducing conditions and transferred to a nitrocellulose membrane. Degradation of C4b was evaluated by using a rabbit anti-C4c antibody (DAKO) followed by a peroxidase-conjugated goat anti-rabbit IgG.
The protease inhibitory activity of C1-Inh bound to the borrelial surface was examined by detection of SDS-insoluble complexes of C1-Inh and C1s protease. To opsonize cells with C1, 108 B. recurrentis cells were incubated with 10% NHS for 1 h at 30°C. After washing, cells were treated with 1 µg biotinylated C1-Inh for 1 h at 30°C. Following three washes, cells were lysed and the borrelial whole cell preparations were subjected to SDS-PAGE (7.5%) under non-reducing conditions. Proteins were transferred to nitrocellulose membranes and probed with either peroxidase-conjugated streptavidin or goat anti-C1s (Atlantic antibody) followed by a peroxidase-conjugated rabbit anti-goat IgG.
The CihC encoding cihC gene including its native promoter region was amplified by PCR using primers CihC Prom and CihC SphI. The resulting amplicon was cloned into pBSV2 yielding shuttle vector pCihC. Transformation of B. burgdorferi B313 and characterization of transformants was previously described [11]. Expression of CihC of transformed B. burgdorferi B313 was determined by Western blot, whole cell ELISA and immunofluorescence analysis, using mAb BR2. High-passage, non-infectious B. burgdorferi strain B313 were grown in 100 ml BSK medium and harvested at mid exponential phase (108 cells/ml). Electrocompetent cells were prepared as described previously [26] with slight modifications. Briefly, 50 µl aliquots of competent B. burgdorferi strain B313 cells were electroporated at 12.5 kV/cm in 2-mm cuvettes with 10 µg of plasmid DNA. For control purpose B. burgdorferi strain B313 cells also were transformed with pBSV2 vector alone. Cells were immediately diluted into 10 ml BSK medium and incubated without antibiotic selection at 30°C for 48 to 72 h. Bacteria were then diluted into 100 ml BSK medium containing kanamycin (30 µg/ml) and 200 µl aliquots were plated into 96-well cell culture plates (Corning) for selection of transformants. After three weeks, wells were evaluated for positive growth by color change of the medium, confirmed by dark-field microscopy for the presence of motile spirochetes. The cihC gene of transformed B. burgdorferi B313 was detected by PCR using oligonucleotides CihC F and CihC SphI. Ectopic CihC expression was analyzed using immunofluorescence microscopy and ELISA in combination with mAb BR2. In addition, ectopically expressed CihC was analyzed by ligand affinity blotting and flow cytometry with regard to its capacity to acquire C4bp and C1-Inh.
Briefly, 107 B. recurrentis A1, B. duttonii La, B. burgdorferi B313/vc and B313/pCihC cells were washed twice with PBS, blocked for 15 min at RT with PBS/10% BSA, and incubated with 10 µg/ml of biotinylated C4bp or C1-Inh in FACS-buffer (PBS/1% BSA) for 1 h at RT. As a negative control, spirochetes were incubated with the same concentration of biotinylated BSA. Cells were washed three times, stained with phycoerythrin (PE) labeled streptavidin (Bio-Rad) and were then fixed with 1% paraformaldehyde overnight and analyzed using a FACS-Calibur and the CellQuest software (BD Biosciences).
The serum susceptibility of mock-transformed B. burgdorferi B313 (B. burgdorferi B313/vc) and transformed B. burgdorferi B313 (B. burgdorferi B313/pCihC) was assessed using a survival assay. Cells grown to mid-logarithmic phase were harvested, washed and approximately 3×107 spirochetes were resuspended in BSK-H medium supplemented with either 50% factor B-depleted human serum (NHS-B) or 50% heat inactivated factor B-depleted human serum (hiNHS-B). Cells were incubated in Eppendorf tubes at 30°C for 2 days. At day 0, 1, and 2, cells were washed in 0.85% NaCl, transferred to microtiter plates and incubated with SYTO9 (Molecular Probes, Invitrogen) as recommended by the manufacturer. Subsequently, relative growth of spirochetes as compared to day 0 was determined by measuring the fluorescence intensity at 530 nm (excitation 485 nm) on a microtiter plate reader (Victor2 plate reader, Perkin Elmer).
For whole cell ELISA, approximately 108 spirochetes (B. burgdorferi B313/vc and B313/pCihC) were washed twice, resuspended in PBS and immobilized on microtiter plates overnight at 4°C. The wells were washed with PBS/0.05%Tween, blocked with PBS/5% BSA and were then incubated with the CihC-specific mAb BR2 or the flagellin-specific mAb LA21 followed by a peroxidase-conjugated sheep anti mouse IgG. Substrate reaction was performed with o-phenyldiamine dihydrochloride (Sigma-Aldrich) and absorbance was measured at 492 nm.
Statistics were analyzed with the unpaired Student's t-test, P values less than 0.05 were considered significant.
The cihC gene sequences of B. recurrentis and B. duttonii reported in this paper have been deposited in the EMBL/GenBank data bases under the following accession numbers: FN552439 and FN552440, respectively.
To verify acquisition of C4bp onto the outer surface B. recurrentis and B. duttonii spirochetes were incubated with biotinylated human C4bp and analyzed by flow cytometry. Both strains were found to acquire C4bp onto their surfaces (Fig. 1A). By applying ligand affinity blot analysis for detection of C4bp-binding molecules, a protein of about 40 kDa was identified in B. recurrentis and B. duttonii, but not in B. hermsii and B. burgdorferi (Fig. 1B). In addition, B. recurrentis and B. duttonii are capable of binding the complement regulator C1-Inh (Fig. 1). To isolate and characterize the receptor for C4bp, cell lysates of B. recurrentis A1 were incubated with C4bp and added to Protein G Sepharose coupled anti-C4bp immune serum. The co-precipitating protein of approximately 40 kDa was analyzed by mass spectrometry and the peptides generated matched an open reading frame of 1071 bp on the genome of B. recurrentis A1 [27]. The open reading frame encoded for a putative lipoprotein with a calculated molecular mass of 40.4 kDa. The encoding gene was designated cihC (complement inhibition via C4bp). Pulse-field gel electrophoresis and hybridization analysis revealed that the cihC gene represents a single genetic locus in B. recurrentis and B. duttonii that maps to a larger plasmid of approximately 200 kb (Fig. 2) [27], [34]. Neither the tick-borne relapsing fever strains of B. parkeri, B. hermsii and B. turicatae nor B. burgdorferi, the causal agent of Lyme disease, hybridized with the cihC probe (data not shown). Isolation of the homologous B. duttonii gene revealed 91% amino acid sequence similarity with that of B. recurrentis (Fig. 3). Lescot et al. previously identified the cihC gene of B. duttonii as a p35-like antigen (BDU_1) exhibiting similarity to the B. burgdorferi fibronectin-binding lipoprotein BBK32. In contrast to our observation the homologous gene in B. recurrentis was not detected [27]. Interestingly, our preliminary studies indicated that recombinant CihC of B. duttonii and B. recurrentis binds fibronectin (unpublished). A BLAST search failed to detect any other protein with significant homology, indicating that the two genes/proteins are restricted to these highly related species of Borrelia.
To determine whether CihC is surface exposed, immunofluorescence assays were performed using mAb BR2 specific for CihC. B. recurrentis spirochetes were incubated sequentially with mAb BR2 and rabbit anti-mouse Cy3-conjugated antibody (Fig. 4A). Epifluorescence microscopy revealed that B. recurrentis organisms expressed CihC on their outer surface in a patch-like manner. Controls incubated with the secondary antibody alone were negative (not shown).
To further confirm surface localization of CihC, B. recurrentis organisms were treated with either proteinase K or trypsin and subjected to Western blot analysis. As shown in Figure 4B, a significant reduction was observed for CihC after 2 h of incubation with proteinase K at concentrations ≥3 µg/ml. Upon treatment of the spirochetes with trypsin, a more restricted protease, only higher amounts (≥6 µg/ml) yielded complete degradation of CihC. The mouse mAb LA21 directed against the periplasmic FlaB protein was used in this experiment as a internal control to confirm that the fragile spirochetal outer membrane was not damaged (Fig. 4B, lower panels). These data indicate that CihC is exposed at the outer surface of B. recurrentis.
To localize the putative domain(s) of CihC that bind to C4bp and C1-Inh, a number of CihC deletion mutants with distinct N- or C-terminal truncations were constructed (Fig. 5A). Protein expression was confirmed by using a His-tag specific antibody and all recombinant proteins exhibited the predicted size. Screening for C4bp and C1-Inh binding by ligand affinity blotting revealed that from the polypeptide preparations tested, full-length CihC (residues 20 to 356) and all truncated versions employing the central protein domain (amino acid residues 145 – 185) similarly retained C4bp and C1-Inh binding activity (Fig. 5B). These results suggest that CihC contained a central region that bound to both human complement regulators.
Inactivation of complement component C4b occurs by factor I mediated cleavage of the C4b alpha chain. To assess whether C4bp maintains this cofactor activity when attached to the surface of intact B. recurrentis spirochetes were coated with purified human C4bp and incubated with C4b and factor I. The supernatant was subjected to SDS-PAGE and C4b alpha chain degradation products were detected by immunoblot analysis. As shown in Figure 6 (left panel), binding of C4bp to the cell surface resulted in α4 and α3 degradation products of 15 kDa and 25 kDa, respectively. In contrast, B. recurrentis spirochetes alone did not promote factor I-mediated cleavage of C4b demonstrating that louse-borne relapsing fever spirochetes lack endogenous C4b cleaving activity. Similarly, C4bp bound to immobilized recombinant CihC protein efficiently mediated C4b processing via factor I, as indicated by the appearance of a α4 fragment (Fig. 6, right panel). B. recurrentis or CihC preincubated with C4bp and C4b in the absence of factor I did not promote cleavage of C4b (data not shown). These findings demonstrate that CihC-associated C4bp retains its cofactor activity and may lead to accelerated disintegration of the C3 convertase (C4bC2a) of the classical complement activation pathway.
The protease inhibitor C1-Inh reacts with its complement target proteases such as C1s and C1r to form high molecular weight SDS resistant complexes [35]. We examined the formation of these covalent C1-Inh-protease complexes as an index for the protease inhibitory activity of CihC-associated C1-Inh. Intact B. recurrentis cells were preincubated in NHS as source for C1 and biotinylated C1-Inh was applied. Subsequently, cells were washed extensively to remove unbound C1-Inh, lysed and subjected to immunoblotting. As shown in Figure 7A, biotinylated C1-Inh acquired by B. recurrentis formed complexes on the spirochetal surface as indicated by the occurrence of a high molecular weight band at >170 kDa. To identify the constituent protease of these complexes, immunoblot analysis using C1s-specific antiserum was employed revealing C1s is a component of the >170 kDa large complex (Fig. 7B). Exogenously applied biotinylated C1-Inh as well as serum-derived C1-Inh formed the respective complexes with C1s. These data suggest that C1-Inh bound to the surface of B. recurrentis retains its functional activity and thus, by inactivating C1s protease, exhibits complement inhibitory activity.
To test whether CihC of B. recurrentis plays an important role in mediating complement resistance, the serum-sensitive B. burgdorferi B313 mutant strain was transformed with the shuttle vector pCihC containing the complete cihC gene (B. burgdorferi B313/pCihC); for control, the pBSV2 vector alone (B. burgdorferi B313/vc) was employed. Expression and surface localization of CihC was determined by whole cell ELISA (Fig. 8A) and immunofluorescence (Fig. 8B) analyses using the CihC-specific mAb BR2. Moreover, to ascertain whether the ectopically expressed CihC is capable of recruiting C4bp and C1-Inh to the surface of B313/pCihC flow cytometry was performed (Fig. 8C). The B313/pCihC-transformed isolate but not the mock-transformed B313/vc isolate of B. burgdorferi strongly expressed CihC and acquired both complement regulators.
To compare the susceptibility of B313/pCihC and B313/vc to complement-mediated killing, both specimens were subjected to a human serum sensitivity assay. In order to avoid killing of Borrelia strains via the alternative pathway of complement activation a factor B-depleted human serum was employed. Accordingly, spirochetes were incubated in factor B-depleted human serum (NHS-B) or heat-inactivated factor B-depleted serum (hiNHS-B) and spirochetal growth was monitored by uptake of a nucleic acid dye. B313/pCihC and the mock-transformed strain multiplied during the 48 h time interval when incubated with heat-inactivated factor B-depleted serum (Fig. 8D). However, when exposed to NHS-B only B313/pCihC spirochetes could replicate indicating that ectopic expression of CihC renders serum-sensitive B.burgdorferi B313 more resistant to complement-mediated lysis. These data suggest a decisive role for CihC in serum resistance of B. recurrentis and B duttonii.
Bacteria have evolved multiple strategies to interfere with complement-mediated clearance of pathogens by blocking distinct steps of the lytic cascade. Recently, we provided evidence that the louse-borne relapsing fever spirochete B. recurrentis selectively inhibits activation of the alternative complement pathway by specifically binding the endogenous complement inhibitor CFH via its lipoprotein HcpA [11]. We now demonstrate for the first time that B. recurrentis also expresses a surface receptor specific for C4bp and C1-Inh, two major serum-derived inhibitors of the lectin and classical complement pathways, termed CihC. Genetic and molecular analyses revealed that CihC of B. recurrentis is a potential lipoprotein and that B. duttonii harbors a homologue of CihC [27]. Upon binding to the pathogen's surface or to recombinant CihC, C4bp retained its cofactor activity for factor I-mediated C4b inactivation. Together with the fact that B. recurrentis also expresses HcpA, the presented data suggest that the potential of louse-borne relapsing fever spirochetes to interfere with both, classical and alternative pathways, contributes to their high resistance and pathogenicity in humans.
The correlation between serum resistance of bacteria and cell surface binding of functionally active C4bp has been reported before for a number of pathogenic microorganisms, including the spirochetes B. recurrentis, B. duttonii and B. burgdorferi s.s. (strain IA), the causative agent of Lyme disease [17]. Moreover, when incubated with human serum, Yersinia enterocolitica, Bordetella pertussis, Neisseria gonorrhoeae, Candida albicans, Moraxella catarrhalis, Escherichia coli K1, Streptococcus pyogenes and Yersinia pestis were also shown to acquire C4bp [36]–[43]. However, the respective receptors for C4bp have only been identified for some bacteria, e.g. Streptococcus pyogenes, Yersinia enterocolitica, and Moraxella catarrhalis, but not for B. recurrentis, B. duttonii and B. burgdorferi. The present data provide evidence that the receptor for C4bp of B. recurrentis, CihC is a surface exposed putative lipoprotein. Preliminary Southern Blot analysis and BLASTN search on databases revealed a putative homologue of cihC only in B. duttonii but not in other spirochetal species suggesting that the gene encoding C4bp receptor is unique to these two Borrelia species. To determine whether the cihC gene is located either on the chromosome or any of the linear plasmids, PFGE and Southern blotting were performed. The cihC gene was localized to a 190 kb linear plasmid adjacent to the previously identified factor H binding hcpA gene. Similarly, the B. hermsii gene encoding the factor H binding protein FhbA maps to the large linear plasmid of 220 kb [13]. However, further studies are required to resolve this issue for other bacterial pathogens.
To localize the peptide domains of CihC relevant for binding of C4bp and C1-Inh, truncated N- and C-terminal deletion mutants were generated and used for functional analyses. C4bp and C1-Inh binding was not abrogated by N-terminal (amino acid residues 20–121) or the C-terminal (amino acid residues 191–356) deletion mutants of CihC indicating that both, C4bp and C1-Inh, bind to the central domain of CihC. In related studies, Streptococcus pyogenes was previously shown to bind C4bp through the N-terminal highly variable region of M-proteins Arp and Sir and similar results were also obtained with the FHA receptor for C4bp of Bordetella pertussis [38], [39], [41], [44]. However, the reason for the differential binding domains of the various pathogen receptors for C4bp is not known at present.
Preliminary data indicate that binding of C4bp to CihC ectopically expressed by B. burgdorferi B313 cells is independent of ionic strength suggesting a hydrophobic interaction between the receptor and its ligand. Similar findings have been reported before for other pathogens. Thus, interaction of the Y. enterocolitica Ail receptor with C4bp was also found to be less sensitive to salt [45]. Moreover, C4bp receptors like Por1A of N. gonorrhoeae [42], [46], UspA1/2 of M. catarrhalis [36], OmpA of E. coli [43] and the M-proteins of S. pyogenes bind C4bp in a nonionic fashion. However, further analyses including C4bp deletion constructs are required to solve this issue for CihC.
The present study adds another facet on the versatility of relapsing fever spirochetes to persist in human blood and to evade innate and adaptive immunity. The best-known immune evasion strategy of relapsing fever Borrelia is antigenic variation, i.e. the ability to respond to newly generated specific antibodies with a switch to an altered variable major outer surface protein (Vmp). Essentially, the pathogen always stays one step ahead of antibodies. However, while antigenic variation is restricted to Vmps, other surface exposed proteins are stable and antigenic, e.g. the surface-exposed lipoprotein FhbA of B. hermsii [12], [13], [47], [48]. In this context it could be speculated that upon binding to CihC, C4bp and C1-Inh inhibit the lectin and classical complement pathway, including the formation of the lytic membrane attack complexes.
In addition to the observed anti-complement activity, B. recurrentis-exposed C4bp may exhibit another biological activity relevant for spirochetal serum resistance. This is indicated by the fact that C4bp circulates in plasma as a complex with protein S that in turn binds to negatively charged phospholipids on membranes [49]–[51]. Thus, it is possible that C4bp also promotes adhesion and subsequently hematogenous dissemination by simultaneously binding to B. recurrentis and endothelial cells. This assumption is supported by the recent observation that the related fibronectin and glycosaminoglycan binding protein, BBK32, of B. burgdorferi mediates endothelial interactions in vivo, thereby facilitating microvascular interactions [52]. Similarly, CihC of B. recurrentis and B. duttonii bound fibonectin and thus could also be involved in the dissemination process of relapsing fever spirochetes.
The assumption that CihC of B. recurrentis and probably also CihC/BDU_1 of B. duttonii are critically involved in their escape from complement-mediated lysis is further supported by the present finding that ectopic expression of CihC in the serum-sensitive B. burgdorferi strain B313 led to a significant increase in resistance to complement mediated lysis. Moreover, binding of C1-Inh, the major inhibitor of several pathways of inflammation in humans, to CihC could be observed. However, the actual role of CihC in the pathogenesis of louse-borne relapsing fever will only be elucidated by in vivo studies in a relevant mouse model [53]. Complement resistance in cihC transformed B. burgdorferi strain was detected in the presence of non-immune factor B-depleted human serum indicating that the lectin/classical pathway of complement activation may be triggered by Borrelia structures other than specific antibodies. Indeed, we have shown that C1q and the C1 complex can bind to the surface of B. recurrentis in the absence of specific antibodies. Moreover, recognition molecules specific for the lectin pathway (i.e. MBLs and ficolins) could also bind to borrelial carbohydrates and activate MASPs [20], [54]–[58]. MASP-2 is the enzyme component that, like C1s in the classical pathway, cleaves the complement components C4 and C2 to form the C3 convertase C4bC2a, common for activation of both the lectin and the classical pathways. However, it remains to be determined whether C4bp and C1-Inh binding significantly increases B. recurrentis spirochetes resistance against complement attack in humans.
In summary, this study is the first to show that B. recurrentis and most probably B. duttonii express a potential lipoprotein receptor, which selectively binds C4bp and C1-Inh, the endogenous regulators of the classical and lectin complement pathway. Together with the fact, that both spirochetal species also carry a specific receptor for the serum-derived complement inhibitor of the alternative pathway, CFH, the present data emphasize the versatility of B. recurrentis and B. duttonii to evade lectin/classical and alternative pathways of complement activation. Elucidating the pathological processes underlying relapsing fever will be helpful to design novel regimens for therapeutic treatment of spirochete-induced relapsing fever and to develop potential vaccine candidates.
|
10.1371/journal.pntd.0001658 | Methodology for Definition of Yellow Fever Priority Areas, Based on Environmental Variables and Multiple Correspondence Analyses | Yellow fever (YF) is endemic in much of Brazil, where cases of the disease are reported every year. Since 2008, outbreaks of the disease have occurred in regions of the country where no reports had been registered for decades, which has obligated public health authorities to redefine risk areas for the disease. The aim of the present study was to propose a methodology of environmental risk analysis for defining priority municipalities for YF vaccination, using as example, the State of São Paulo, Brazil. The municipalities were divided into two groups (affected and unaffected by YF) and compared based on environmental parameters related to the disease's eco-epidemiology. Bivariate analysis was used to identify statistically significant associations between the variables and virus circulation. Multiple correspondence analysis (MCA) was used to evaluate the relationship among the variables and their contribution to the dynamics of YF in Sao Paulo. The MCA generated a factor that was able to differentiate between affected and unaffected municipalities and was used to determine risk levels. This methodology can be replicated in other regions, standardized, and adapted to each context.
| Yellow fever (YF) is an infectious disease, transmitted by mosquitoes, and very common in North and Middle East region of Brazil, where cases of the disease are reported every year. Since 2008, outbreaks of the disease have occurred in regions of the country where no reports had been registered for decades, which has obligated public health authorities to redefine risk areas for the disease. The aim of the present study was to propose a methodology of environmental risk analysis for defining priority municipalities for YF vaccination. The municipalities were divided into two groups (affected and unaffected by YF) and compared based on environmental parameters related to the disease's epidemiology. Statistical analysis was used to identify associations between the variables and virus circulation, as well as, to evaluate the relationship among the variables and their contribution to the dynamics of YF. The MCA generated a factor that was able to differentiate between affected and unaffected municipalities and was used to determine risk levels. This methodology can be replicated in other regions, standardized, and adapted to each context.
| Brazil has an extended enzootic or endemic area for sylvatic yellow fever (YF), where cases of the disease are annually reported. The highest frequency of the disease occurs between January and April, when high levels of rainfall and an increase in the vector population coincide with greater agricultural activity [1]–[5].
In Brazil, endemic cases of the disease were limited to the northern, middle, western, and pre-Amazon regions until 1999 [4], [6]. Since then, YF has progressively expanded its territory, and a gradual increase of reported cases has been observed near the traditional boundaries of endemic zones. This expansion highlights the need to redefine the areas of risk [6]–[9].
Until 2008, four distinct epidemiologic area types for YF were acknowledged in Brazil: endemic areas (where vaccination against YF was recommended), transition areas (also known as epizootic or emergence areas), potential risk areas (where vaccination against YF was not recommended) and disease-free areas (where YF did not occur and vaccination against YF was not recommended) [4], [6]. Zones classified as —transition, and —potential risk, have no records of virus circulation and no indication for YF vaccination. However, these areas possess some environmental parameters that are compatible with the establishment and maintenance of the disease; thus, there was a need for increased YF surveillance activities in those regions. Nevertheless, these parameters were subjectively defined and the non-vaccination of supposedly at-risk people generated ethical problems for Brazilian health authorities.
The transition and potential risk zones were eliminated in 2008. Therefore, only two area types are currently acknowledged: endemic (where vaccination against YF is recommended) and disease-free (where vaccination against YF is not recommended).
In public health emergency situations, the municipalities where vaccination should be recommended are defined by classification methods based on affected or expanded areas. Thus, municipalities are considered to be affected when the virus circulation can be detected, which occurs when YF epizooties are confirmed in nonhuman primates, when there are confirmed human cases, or when the virus is isolated in mosquitoes [6], [7]. Municipalities within 30 km of a municipality where virus circulation has been detected are also considered to be affected areas [6].
The YF vaccine was considered completely safe until 2001, as there had been no reports of serious adverse reactions associated with its administration. However, 12 serious cases were reported in 2001 [10]–[12], and 39 additional cases were identified worldwide through May of 2009 [13]; to date, over 50 cases have been reported [13]–[15]. Two types of serious adverse reactions are commonly reported: neurotropic disease, which is caused by the invasion of the nervous system by the vaccine virus, and viscerotropic disease, which is a pan-systemic infection that is similar to the infection caused by the wild-type virus [13].
A dilemma is thus created for the public health authorities: what proportion of the at-risk population should be vaccinated to minimize the total number of fatal cases from the natural infection of the yellow fever virus (YFV) or the vaccine virus?
This problem applies to the State of Sao Paulo and to other states located in the southern and southeastern regions of Brazil. Briand et al. (2009) [16] developed a methodology for prioritization of areas for vaccination against YF for countries in Africa, using Multiple Correspondence Analysis. Although, in this study the authors had as limitation: the lack of information available, working with a small number of variables.
Using the current situation of the State of Sao Paulo, Brazil, as an example for definition of priority areas for vaccination against YF, this paper aims to adapt the methodology of risk analysis proposed by Briand et al, (2009) [16] in a context with more availability of information, allowing the use of environmental variables potentially related with the eco-epidemiology of YF.
The study was conducted in the State of Sao Paulo, Brazil. Sao Paulo is composed of 645 municipalities, has an area of approximately 250,000 km2, and has an estimated population of 40 million people. There are currently 429 municipalities in the YF-endemic zone and 216 in the disease-free zone.
Aiming to select variables with the most relevance for the eco-epidemiology of YF, two groups were defined for comparison: municipalities that were affected and municipalities that were unaffected by the disease. The study used a case-control model with an ecological approach.
The Text S1 shows the resume of the steps used in this study. Municipalities with confirmed YFV circulation in their territory and the adjacent municipalities were considered to be affected [6]. There were a total of 12 municipalities with confirmed YFV circulation and 57 adjacent municipalities. The 12 confirmed municipalities and 18 randomly selected adjacent municipalities were included in this study and constituted a sample of 30 cases, which is the minimum necessary for the use of the desired statistical analysis.
The unaffected or eligible control municipalities consisted of all of the municipalities that had no reported cases of YF and that were at least 100 km away from any affected municipality. Figure 1 illustrates the methods used to outline these areas and the municipalities selected for this study.
Each municipality was analyzed relative to the moment before the occurrence of YF in its region or prior to its inclusion as an area of recommended vaccination. Following Briand et al. (2009) [16], the variables were selected to relate to risk allocation based on vulnerability according to three main axes: exposition, susceptibility, and resilience. The authors considered exposition as the capacity for YFV to circulate in a municipality. Thus, data included were related to the environment (land occupation, forest fragmentation, wind direction influences, distance for biodiversity conservation unities, distance for municipalities with YFV circulation and proportion of riparian forest), the vectors (temperature, humidity and pluviosity), and the hosts (human displacements and illegal animal trafficking).
Susceptibility was considered as the number of hosts without immunity for YFV that lived in each municipality. The immune human population was calculated as the proportion of immunized people divided by total population of the municipality. Non-human primates comprised the registered species occurrence in each municipality classified by a score according with the importance of each species as YFV amplifier [17]. Susceptibility also included the risk of urbanization of the disease, based on levels of infestation of Aedes aegypti in the municipality, using the Breateau Index [18].
Resilience was defined as the capacity of each municipality to detect the YFV circulation in its territory (Surveillance for Febrile Ictero-hemorrhagic Syndrome), as well as, its capacity of confrontation an outbreak of YF (Medical care capacity).
The Text S2 shows the variables analyzed in the study. The secondary data were primarily obtained from the Internet. The free software Terraview 3.3.1 was used for distance measurements. Historical series were created for the variables temperature, pluviosity, and humidity using the monthly averages from November to May (months with a greater occurrence of YF). The mean pluviosity divided by the mean real evapotranspiration (RET) in the same period was used as a humidity indicator [19].
Variables that showed statistically significant associations (chi-squared test, p<0.05) were selected for the multiple correspondence analysis (MCA). For the application of MCA, all the variables were categorized and treated like qualitative variables. The MCA is an exploratory and descriptive multivariate statistical technique for categorical data analysis. The technique is appropriate for the analysis of contingency tables with a large number of variables. The method analyses the mass distribution, by the pattern of the frequency for the considered categories, aiming to identify the uniformity of the distribution. This analysis was performed to evaluate the relationships among the selected variables and to obtain factors that best represent all variables, considering the level of significance (weight) of each to explain total sample variability (inertia). Thus, the graphic obtained can be studied like a geographic map, analyzing the relationship of proximity by projections of the factors, in way that each point represents each variable. STATISTICA 7 software was used to perform the MCA.
The bivariate analysis identified seven variables associated with YFV circulation (Table 1). The MCA generated 12 factors to explain the total sample variability (inertia). One of the factors could independently explain 28.1% of the total sample variability. None of the other 11 factors were able to independently explain more than 10% of the sample variability.
The analysis of the graphic (Figure 2) allows visualization of the relationship between the variables used for the construction of F factor. The graph contains only one dimension, and each point's disposition represents the position of each variable.
Three main clusters of variables could be noted. The first cluster shows a collection of variables that represent, in theory, lower risk for the occurrence of YF. These variables are represented by extreme values: greater distances to areas with recommended vaccination against YF (DIST_VAC:1) and biodiversity conservation units (DIST_BCU:1), smaller proportions of riparian forest (RIPA:3), fewer routes of illegal wildlife traffic (TRAF:3), less humidity (HUMI:1), less influence of the direction of dominant wind routes (WIND:3), and no surveillance for SFIHS (SFIHS:2).
The second cluster shows variables of intermediate values and the third shows opposite values of those observed in the first cluster. The variable distance to area with recommended vaccination against YF was the only exception observed, where values for —adjacent or up to 30 km (DIST_VAC:1) and —31 to 100 Km (DIST_VAC:2) were clustered between the first and second clusters.
The weights of each variable (Table 2) were identified by MCA, based on geometric distance between than in the graphic. Thus, these values were used on the equation, and the F factor was calculated for each municipality.
It's known that municipalities without the SFIHS are less resilient. So, the association of this variable with the YFV circulation was considered as protection factor. Thus, the positive sign of the variable was inverted for the calculation of the F factor, in way that, municipalities without SFIHS had its F factor increased, and so, considered more vulnerable.
Analyzing the graph (Figure 3) allows us to observe the difference between cases group 1) and controls (group 2) according to the F factor. All municipalities from control group (non-affected) showed values under zero. So, this was defined as the cut point to differentiate risk and no risk. The scale of risk was divided in two to turn the model able to give priority for municipalities with higher F factor values.
Thus, the priority levels for vaccination against YF in municipalities of the State of Sao Paulo were: F factor<0.0 = low risk; 0.0<F factor<2.0 = some risk; F factor>2.0 = high risk.
The study used a large number of variables. Much of the work focused on the collection and standardization of the information from secondary sources (i.e. the internet). The 60 municipalities evaluated in the case-control step was the minimum necessary to support the statistical analysis. The State is composed of 645 municipalities, thus, the collection of all information for each municipality would be unfeasible. Therefore, effort was made to identify the most important variables. This approach, using a subset of municipalities, simplified and optimized the method for broader application to target municipalities with not currently indication for YF vaccination.
The use of secondary data that are available on the Internet is one limitation of this methodology, especially given that the data were not collected for this purpose. However, the authors sought to incorporate official published data on each subject. Therefore, the limitation is admitted for a better replicability of the method.
The study showed the importance of a large number of variables for the ecoepidemiology of YF.
The distance between the municipality and areas with recommended vaccination against YF can be considered an important criterion for the prioritization of a municipality for YF vaccination. Municipalities from affected regions were, for the most part, close to or even inside of areas with recommended vaccination against YF at the moment of the case occurrence. The occurrence of YF in municipalities with a small proportion of susceptible individuals indicates the importance of vaccination coverage of close to 100% for populations living in areas of risk, as is recommended by Brazil's Ministry of Health [6].
Municipalities located in affected regions were closer to biodiversity conservation units (BCU). Mosquito species that serve as vectors for YF are mainly found in well-conserved forest patches [20]. It is possible that BCUs favor the proliferation of these species and increase the chances of disease maintenance by serving as stepping stones for the geographic expansion of the disease.
The Brazilian Forest Code includes the riparian forests in the category of permanently protected areas. Thus, forest patches are more frequently maintained in these environments and generate more stable ecological corridors. These forests represent one of the few environments that allow the displacement of non-human primate populations.
Affected municipalities were closer to main illegal wildlife traffic routes. Trafficking of illegal wildlife can be an important source for the dissemination of viremic non-human primates from areas of virus circulation. Every year, large numbers of non-human primates that originate from YF-endemic regions, such as the Amazon, are apprehended from illegal trafficking [21]. These animals are often returned to forest environments without adequate ecological and sanitary evaluations, which allows for contact between these viremic hosts and vectors of the disease [22].
Climatic factors, such as humidity and temperature, have a direct influence over the abundance of YF mosquito vector species, as well as virus multiplication in its arthropod reservoirs [23]–[26]. Unlike the humidity calculated as a percentage relative to the availability of water vapor in the air, the RET is calculated in mm3, which allows for the evaluation of its relationship with the pluviosity and hydrologic balance of the region. Given that the RET takes into consideration several factors, it is a more complete indicator of climatic conditions than the isolated values of pluviosity and temperature; therefore, it better represents the context of topography and land occupation in the municipalities [19], [27].
Another climatic factor that presented a statistically significant association between the groups was the influence of dominant wind routes that arrive at each municipality. The biological plausibility of this hypothesis is related to the possibility of dispersion of mosquito vectors by dominant winds [28]–[31]. Causey et al. (1950) [29] evaluated the dispersion patterns of mosquitoes of the genus Haemagogus spp. And Sabethes spp. in the State of Minas Gerais, Brazil. Dispersion capabilities of up to 11 km were observed. The authors concluded that environments composed of forest patches, agriculture, and pasture favor the expansion of YF by increasing the wind dispersion of mosquitoes.
The importance of active Surveillance for Febrile Ictero-hemorrhagic Syndrome (SFIHS) was also demonstrated by the present study. The affected municipalities mostly coincided with regions of the state where this surveillance system had been implemented. Therefore, municipalities without SFIHS had less resilience, meaning a lower capacity for disease detection to address a possible virus circulation in its territory.
Due to the large number of important variables for YF eco-epidemiology in this study, it is possible to visualize the complexity of the disease. Several factors probably act simultaneously and in different combinations to determine virus establishment and maintenance in a region. Therefore, multivariate analysis techniques are important for the evaluation of the influence of each variable on the disease's eco-epidemiology.
Variables that showed greater contributions to the variability of the municipalities in relation to the F factor observed in this study were influence of the direction of dominant wind routes, number of illegal wildlife traffic routes, proportion of riparian forest, and the implantation of surveillance of FIHS.
The grouping of the variable distance to area with recommended YF vaccination into groups of —up to 30 km, and —31 to 100 km, suggests a possible need to increase the current 30-km radius for the areas considered to be at risk (expanded areas) during outbreaks of YF.
The legislation that establishes the YF surveillance system in Brazil [6] defines that this System must be based on confirmed cases rather than predictions about the occurrence of the disease in areas of potential risk. The main purpose of the system is focused on the rapid detection of suspect cases and the adoption of emergency measures that will prevent an epidemic outbreak. The previous approach for risk classification of YF in Brazil, using —transition, and —potential risk, areas for guiding control measures, allowed for the intensification of surveillance in areas of known environmental potential for disease establishment. However, this approach was highly subjective because the criterions for defining areas were not described in a systematic way. The difficulty in replicating courses of action led to the simplification that is the current method [32].
However, it is extremely important that a surveillance system, such as that for YF—a fatal disease with great potential for outbreaks—works with models for supporting an evidence-based public-health decision-making process to guide actions in outbreak emergency situations. In the long run, the goal is to interrupt the expansion of the disease to large populated areas or known vulnerable populations. The present study has proposed a methodology for the definition of vulnerable regions for YF using environmental variables and a systematic design that focuses on a regional scale.
The difference between the current method and the method proposed can be noted by the fact that all control municipalities, which were located in area with YF vaccination indicated, but without registration of YFV circulation, were classified as without risk in this study [6], [32].
In this sense, it is recommended that, within the vaccination, municipalities classified as in —risk, pass through an analysis of its structural capacity for confrontation of a YF outbreak, considering: the number of technicians trained and sensitized for: detection of YF suspected cases, treatment and laboratorial diagnosis resources, viability for detection of epizootic events in non-human primates, capacity for conduction of entomological studies, and viability for timely conduction of campaigns of vaccination for target populations when required.
In the case of municipalities classified as “high risk”, it is recommended that, in addition to the measures cited above, the organization of surveillance system for SIHFS be conducted, once this system increases the sensitivity of the Yellow Fever Surveillance System in other regions of State affected by the disease [32].
It is also recommended that professionals using this methodology visualize the geographic distribution of municipalities according with the risk classification. This type of approach can be useful for organization of the action measures for disease control. Moreno & Barata (2011) [32] showed that, in Sao Paulo State, the municipalities with higher risk are the most populated. In cases like these, the increase of surveillance measures can be an option more feasible both financially as well as operationally.
The increased geographic expansion of emergent diseases, such as YF, exposes the health surveillance systems to the need to seek methodologies with multidisciplinary approaches that are able to adapt to different regional realities. Using locally relevant environmental variables and a systematic design, the methodology proposed in this study was able to differentiate municipalities according to their vulnerability for the occurrence of YF. This methodology can be replicated in other regions, standardized, and adapted to each context.
|
10.1371/journal.pgen.1007520 | Hypodermal responses to protein synthesis inhibition induce systemic developmental arrest and AMPK-dependent survival in Caenorhabditis elegans | Across organisms, manipulation of biosynthetic capacity arrests development early in life, but can increase health- and lifespan post-developmentally. Here we demonstrate that this developmental arrest is not sickness but rather a regulated survival program responding to reduced cellular performance. We inhibited protein synthesis by reducing ribosome biogenesis (rps-11/RPS11 RNAi), translation initiation (ifg-1/EIF3G mutation and egl-45/EIF3A RNAi), or ribosome progression (cycloheximide treatment), all of which result in a specific arrest at larval stage 2 of C. elegans development. This quiescent state can last for weeks—beyond the normal C. elegans adult lifespan—and is reversible, as animals can resume reproduction and live a normal lifespan once released from the source of protein synthesis inhibition. The arrest state affords resistance to thermal, oxidative, and heavy metal stress exposure. In addition to cell-autonomous responses, reducing biosynthetic capacity only in the hypodermis was sufficient to drive organism-level developmental arrest and stress resistance phenotypes. Among the cell non-autonomous responses to protein synthesis inhibition is reduced pharyngeal pumping that is dependent upon AMPK-mediated signaling. The reduced pharyngeal pumping in response to protein synthesis inhibition is recapitulated by exposure to microbes that generate protein synthesis-inhibiting xenobiotics, which may mechanistically reduce ingestion of pathogen and toxin. These data define the existence of a transient arrest-survival state in response to protein synthesis inhibition and provide an evolutionary foundation for the conserved enhancement of healthy aging observed in post-developmental animals with reduced biosynthetic capacity.
| Protein synthesis is an essential cellular process, but post-developmental reduction of protein synthesis across multiple species leads to improved health- and lifespan. To better understand the physiological responses to impaired protein synthesis, we characterize a novel developmental arrest state that occurs when reducing protein synthesis during C. elegans development. Arrested animals have multiple survival-promoting phenotypes that are all dependent on the cellular energy sensor, AMP kinase. This survival response acts through the hypodermis and causes a reduction in pharyngeal pumping, indicating that the animal is responding to a perceived external threat, even in adults. Furthermore, exposing animals to pathogens, or xenobiotics they produce, can recapitulate these phenotypes, providing a potential evolutionary explanation for how a beneficial response in adults could evolve through the inhibition of an essential biological process such as protein synthesis.
| The differing phenotypes stemming from the loss of essential cellular functions, such as protein synthesis, are specific to the time in life (development or adulthood) when the deficit occurs. Under such deficits, arresting development is an established strategy at the disposal of animals to ensure future reproductive success.
During its four larval stages, the nematode C. elegans has several possible arrested states that trigger in response to different stressors, including dauer [1, 2], starvation-induced arrest [3], and adult reproductive diapause [4, 5], among others. Dauer diapause occurs under lack of food, high temperature, or high population density, inducing an alternative larval stage 3 [2]; this dauer state carries both metabolic and behavioral changes, including increased stress resistance [6, 7]. This stress resistant and pre-reproductive arrest state is thought to have evolved to allow the worm to conserve its resources, and it affords protection from the environment until a more favorable environment is encountered. Starvation-induced arrest can occur at larval stage 1 (L1), induced from starvation occurring immediately after hatching, and this state similarly results in stress resistance[3]. Two other arrest states are adult reproductive diapause, which is induced by L4 starvation and results in an early-adult arrest state capable of surviving long periods of nutrient deprivation with the ability to later resume reproduction, and impaired mitochondria arrest, induced by deficiency in mitochondrial respiration and resulting in L3 arrest [4, 8]; however, these two states have not yet been directly shown to have stress resistance phenotypes. These examples suggest the existence of cellular programs that function as checkpoints throughout development that stall reproduction to promote fitness [9].
Intriguingly, the same triggers that induce these genetically regulated arrest states during development, when initiated post-developmentally, lead to increased life and healthspan (e.g. daf-2/Insulin IGFI signaling mutants [10–12], mitochondrial deficiency [13, 14]). Moreover, the loss of essential cellular functions was shown to alter animal behavior [15], presumably to avoid further exposure to the environment causal for the perceived loss of cellular homeostasis [16–18].
Protein synthesis inhibition is another trigger of developmental arrest early in life and increased lifespan in adults [16, 19–21], although the underlying mechanisms are not well understood. Similar to inhibiting the insulin-signaling pathway in adults, inhibiting protein synthesis provides several resistances from stress—starvation, thermal, and oxidative [20, 22]. Activation of the energy sensor AMP-activated protein kinase (AMPK) is linked to a reduction in protein synthesis [23–25], and AMPK can be activated by reducing growth via starvation in C. elegans [26] or via inhibiting S6 kinase in isolated mouse cells [27, 28]; this activation includes increased lifespan that is dependent on activation of AMPK in C. elegans [28].
Here we provide new characterization of a C. elegans survival arrest state brought on by reducing protein synthesis, which confers stress resistance and is reversible. Enacting protein synthesis inhibition in the hypodermis alone was partially sufficient for both the arrest and stress resistance phenotypes. Arrested animals had very high expression of a metallothionein and were found to have higher levels of calcium, which may be linked to an observed reduction in pharyngeal pumping. All of these survival phenotypes, save the arrest, were dependent on functional AMPK. Finally, these phenotypes could be recapitulated from exposure to xenobiotics, implying a potential evolutionary context for this fitness-promoting arrest state.
To elucidate the possible connection between the developmental arrest and longevity-promoting effects of protein synthesis inhibition [16, 19–21, 29], we first defined the nature of the developmental arrest in C. elegans. We analyzed the effects of protein synthesis inhibition by targeting distinct and conserved aspects of the protein biosynthesis machinery (S1A Fig). We measured the synthesis of two GFP reporters; a heat shock inducible promoter (S1B Fig) and a mlt-10p driven construct (S1C Fig) that is only expressed between developmental molts as a surrogate assessment for general protein biosynthesis [30]. Because GFP from these reporters is limited to temporally distinct periods, we can robustly measure differences in GFP levels between protein synthesis inhibition conditions. We targeted the translation initiation factor, egl-45/EIF3A, or the small ribosomal protein, rps-11/RPS11, by RNA interference (RNAi), so that we could control the strength and duration of inhibition, thereby avoiding the constitutive arrest that can occur when protein synthesis is inhibited by genetic mutation [31]. While there are many genes involved in protein synthesis that can induce arrest when inhibited [16, 19, 20], egl-45 and rps-11 were selected as RNAi of these genes results in a fully penetrant larval arrest phenotype (S1D Fig). There is a threshold effect to this arrest, as diluting the RNAi to 10% of total food allowed more escaping animals (S1E Fig), while still impairing development. In all RNAi conditions tested at 100% of total food, we observed a potent developmental arrest that could persist beyond 10 days (S1D Fig). To define the developmental arrest state more precisely, we made use of the molting reporter (mlt-10p::gfp-pest) that marks each of the four developmental molts in C. elegans [32]. This revealed a potent arrest after the first molt at larval stage 2 (L2) (Fig 1A–1C). In addition, these animals are morphologically different than other arrest states like dauer and L1 arrested animals (S1F Fig) and are smaller in length than wild type L2s; unlike arrested L2d animals [33] (S1G Fig). Together, these data support the existence of a potent developmental arrest point in response to diminished biosynthetic capacity.
To address the hypothesis that the induced developmental arrest in response to protein synthesis inhibition is beneficial, we challenged L2 arrested animals and non-arrested L2 control animals to oxidative (20mM H2O2, Fig 1D and S2A Fig) or thermal (36°C, Fig 1E and S2B Fig) stress and found the arrested animals were more resistant to all tested environmental insults. Animals that remained in the arrested state for longer periods of time (2 or 10 days) were markedly more protected against oxidative stress and extended exposures to thermal stress (S2C–S2F Fig). Thus, the durability of the response and the capacity to further enhance resistance to perceived deficiencies is enhanced so long as it is needed. Collectively, these data show that loss of protein biosynthetic capacity during development does not induce a decrepit state, but rather a beneficial health-promoting state of impeded development.
The amplification of stress resistance that correlated with time in the arrested state predicted that arrested animals could persist in the L2 stage for much longer than wild type animals. Given this, we examined the lifespan of animals in the arrested state and discovered that egl-45 RNAi and rps-11 RNAi animals had a mean survival in the arrested state of 24 and 12 days, respectively (Fig 1F), compared to a normal eight hour L2 stage (Fig 1A). As such, the developmental arrest resulting from reduction of protein biosynthetic capacity results in health-promoting state of extended diapause.
One hypothesis is that pausing development in the L2 stage alone confers survival benefits. To test this, we screened all annotated RNAi clones that induce early and fully penetrant L2 arrest (S2G and S2H Fig) and measured their ability to resist the same exposure to stress. Despite sharing an L2 arrest phenotype, none of these RNAi treatments resulted in the same decrease in protein synthesis (S2I Fig) or afforded increased survival during stress (S2J and S2K Fig). As such, arrest at the L2 stage does not require a loss in biosynthetic capacity and is not inherently stress resistance-promoting. In addition, the phenotypes observed are not tied to RNAi responses, as ifg-1(ok1211) mutant animals that arrest at the L2 state [31] are more resistant to oxidative stress as compared to wild type controls (S2L Fig). We also tested the long-term survival of acn-1, let-767, and pan-1; while only acn-1 maintained long-term L2 arrest (S2H Fig), the survival of acn-1 RNAi treated animals was significantly shorter than rps-11 and egl-45 RNAi treated animals (S2M Fig).
Finally, we tested the necessity of daf-16/FOXO, a transcription factor that is required for dauer arrest [9], in these survival phenotypes. Reducing protein synthesis in daf-16(mgDf47) mutants still causes developmental arrest (S3A Fig) and results in increased resistance to oxidative (S3B Fig) and thermal (S3C Fig) stress. We further note that these animals are not dauers, morphologically (S1F Fig) and are not resistant to treatment with 1% SDS—a phenotype of animals that successfully enter dauer diapause. Moreover, reducing protein synthesis in daf-2(e1368) mutants, which form constitutive dauers at the restrictive temperature of 25C, enter this L2 arrest stage instead of developing into dauers. These findings support the protein synthesis inhibition arrest state at the L2 larval stage and prior to dauer formation, which is an alternative L3 stage (S3D Fig).
Considering the need for every cell to sense and respond to changes in biosynthetic capacity, but also the benefit of coordinating a systemic physiological response to a perceived organism-level deficit in any tissue, we hypothesized that the response to protein synthesis inhibition would be both cell autonomous and non-autonomous. The germline is a facile model for cell division in early larval development in C. elegans [34]. Similar to the developmental arrest observed at the organism level, tissue-general protein synthesis inhibition resulted in the clear arrest of the reproductive tissue at a stage typical for L2 animal development (Fig 2A). We next sought to determine which tissues were capable of initiating the L2 arrest. Using tissue-specific RNAi, we systematically reduced the expression of egl-45/EIF3 or rps-11/RPS11 in the intestine, germline, or hypodermis (S4A Fig). Similar to tissue-general RNAi, hypodermal-specific protein synthesis inhibition induced potent developmental arrest (Fig 2B–2D) and halted germline proliferation (Fig 2E). In contrast, while still slowing development, intestinal or germline-specific RNAi was unable to induce developmental arrest (S4B–S4J Fig). Germline-specific protein synthesis inhibition results in sterility (S4H–S4K Fig), which differentiates the cell autonomous effects of protein synthesis inhibition from the cell non-autonomous impact on the entire organism when diminished biosynthetic capacity is restricted to the hypodermis. Hypodermal-specific protein synthesis inhibition was the most effective at enhancing resistance to oxidative (Fig 2F) and thermal (Fig 2G) stress, as compared to germline- and intestine-specific RNAi (S4M–S4P Fig), which had modest or no effect on stress resistance. Moreover, hypodermal-specific protein synthesis inhibition initiated post-developmentally was capable of increasing lifespan and, in the case of egl-45 RNAi, was at least equally potent as tissue-general protein synthesis inhibition (S4Q Fig). As predicted by their essential roles in protein synthesis, egl-45/EIF3 and rps-11/RPS11 expression is detectable in several tissues (S4R–S4U Fig), but the differences in the expression level and location could explain the variance in the strengths of phenotypes observed in egl-45 RNAi versus rps-11 RNAi. Nevertheless, these data identify the hypodermis as an important mediator of organismal regulation of growth and development in response to diminished biosynthetic capacity.
We examined the transcript levels of a panel of genes with established roles in stress adaptation (see Methods) under both 24 hours and 120 hours exposure to protein synthesis inhibition (collected after 24 and 120 hour exposure to RNAi) [35]. Despite the enhanced stress resistance observed in protein synthesis inhibition-induced L2 arrested animals, the expression of most genes tested—including several heat shock proteins, redox homeostasis pathway components, and isoforms of superoxide dismutase—was significantly repressed (S5A–S5J Fig).
The notable exception in this panel was the expression of mtl-1, a metallothionine involved in metal homeostasis, which after 24 hours of either egl-45/EIF3 or rps-11/RPS11 RNAi was increased >10-fold (Fig 3A and S5E Fig); in animals arrested for 5 days, mtl-1 was increased >100-fold (Fig 3B and S5F Fig). This temporal enhancement was not observed for other genes involved in stress adaptation (S5G–S5J Fig). Moreover, hypodermal-specific protein synthesis inhibition also induced mtl-1 expression (Fig 3A and S5E Fig), consistent with the notion that the hypodermis is a potent sensor for organismal biosynthetic capacity. As mtl-1 is activated in response to heavy metals, we challenged protein synthesis inhibition-arrested animals to toxic levels of Cd2+ (50mM) and discovered this arrest state also enhanced resistance to heavy metal stress (Fig 3C). Because heavy metal resistance was not previously annotated in adults with protein synthesis inhibition [19–21], we initiated protein synthesis inhibition post-developmentally by egl-45/EIF3A or rps-11/RPS11 RNAi, which also resulted in resistance to Cd2+ exposure (S5K Fig). Similar to oxidative and thermal stress, hypodermal-specific RNAi of egl-45/EIF3 or rps-11/RPS11 could recapitulate the whole animal RNAi phenotype (S5L Fig). We next tested whether the increase in mtl-1 was causative for the resistance, so we created a double mutant of mtl-1(tm1770) and mtl-2(gk125) (mtl-2 is a related metallothionine also activated in response to heavy metals), which greatly attenuated the ability to survive Cd2+ exposure when protein synthesis is inhibited (S5M Fig). Based on these heavy metal responses, we wanted to further test if hypodermal RNAi could increase mtl-1 to the same degree as observed in wild type animals exposed to protein synthesis inhibition for extended periods. Correlating with the rate of developmental arrest, mtl-1 levels increase out to 48 and 120hrs of exposure to hypodermal specific RNAi of egl-45 or rps-11 (S5N Fig). However, animals with longer exposure to rps-11 RNAi have mtl-1 transcript levels that return to near wild type levels, which correlates with the escape from developmental arrest under hypodermal specific rps-11 RNAi (Fig 2D).
Although heavy metals are not abundant in standard growth media, these findings led us to examine the total metal content of animals in protein synthesis inhibition arrest by inductively coupled plasma-atomic emission spectroscopy (ICP-AES). The metal profiles revealed a significant reduction in Mg2+ and Mn2+ and a marked increase in Ca2+ (Fig 3D and S6A Fig). These steady-state concentrations of metals were maintained in animals trapped in the arrested state for 5 days (Fig 3D and S6A Fig). mtl-1;mtl-2 double mutant animals reduced multiple metal species by 10–20%, but did not affect Ca2+ levels (S6B–S6D Fig); protein synthesis inhibition treatment in this mutant was still able to induce many of the same Mg2+, Mn2+, and Ca2+ changes as seen in wild type, consistent with the transcriptional induction of mtl-1 acting as a stress response rather than as the upstream effector. Moreover, animals acutely exposed to CaCl2 treatment as larvae have an mtl-1 transcriptional profile that mirrors animals with protein synthesis inhibition (Fig 3E), suggesting that the increase in Ca2+ could be physiologically significant and promote the increased mtl-1 expression.
Animals have adopted several strategies, ranging from molecular adaptation to changes in behavior, in order to cope with less than ideal growth conditions [36], and calcium plays several critical functions in these physiological responses. As such, we examined the behaviors of animals arrested from protein synthesis inhibition and noted a marked decrease in pharyngeal pumping (Fig 4A and S7A Fig), a rhythmic behavior influenced by calcium transients [37, 38]. The reduction in pharyngeal pumping was significant after 24-hours of protein synthesis inhibition and was more pronounced the more time animals were in the arrested state (S7B Fig); despite this reduction, a basal level of pumping continues even after 15 days in the arrested state (S7B Fig). Similar to the developmental arrest and enhanced stress resistance observed in daf-16(mgDf47) animals, daf-16 is not required for the reduction in pharyngeal pumping rates when protein synthesis is inhibited (S7C Fig). In line with previous cell non-autonomous effects, hypodermal-specific protein synthesis inhibition effectively reduced pharyngeal pumping (Fig 4B), while protein synthesis inhibition in other somatic tissues could not evoke the same magnitude of responses (S7D and S7E Fig). This reduction of pharyngeal pumping is intriguing as this behavior is correlated with food intake [39], and caloric-restriction (CR) is an established means of enhancing organismal health- and lifespan [40, 41]. With this in mind, we measured pharyngeal pumping in adult worms fed egl-45 or rps-11 RNAi to induce protein synthesis inhibition, which are long-lived [16], and also discovered a significant reduction in pharyngeal pumping (S7F Fig). Taken together, these data define reduced pharyngeal pumping as a physiological response of protein synthesis inhibition during development and adulthood.
Protein synthesis is energetically expensive, and it is possible that protein synthesis inhibition leads to a state of excess ATP, which could be redirected to other cytoprotective pathways that drive stress resistance [42]. However, we found that animals exposed to protein synthesis inhibition during development have 50% less cellular ATP (Fig 4C). AAK-2/AMPK is a conserved sensor of energy homeostasis that responds to changes in cellular AMP/ATP levels [43]. Indeed, animals with protein synthesis inhibition have significantly higher AMP/ATP and ADP/ATP ratios (Fig 4D). As such, we tested aak-2 mutants for the protein synthesis inhibition survival and arrest phenotypes. aak-2(ok524) mutants exposed to protein synthesis inhibition were still arrested as L2 animals with reduced germ cell counts (S8A–S8C Fig), but failed to dampen pharyngeal pumping rates (Fig 4E and S8D Fig), which importantly uncouples these two protein synthesis inhibition responses and suggests that the developmental phenotypes are not a result of diminished food intake. Additionally, aak-2 mutant animals failed to evoke protein synthesis inhibition responses observed in wild type animals (Fig 4F). Specifically, aak-2 mutants have minimal, often undetectable, changes in the expression of mtl-1 during protein synthesis inhibition (S8F Fig)—a phenotype similar to daf-16 mutant animals (S8G Fig), which is a known regulator of the mtl-1 locus (S8H Fig). aak-2 mutants are also as sensitive to Cd2+ as wild type animals (S8I Fig), which further supports the connection between mtl-1 expression with resistance to environmental metal exposure. Furthermore, aak-2 mutants with protein synthesis inhibition are as sensitive to oxidative and thermal stress as wild type animals (S8J, S8K, S8M and S8N Fig), indicating the essentiality of AMPK signaling in protein synthesis inhibition-induced stress resistance.
We then tested mutant animals harboring a truncated and constitutively active (CA) form of AAK-2 [44], which slowed development [44] and afforded resistance to oxidative stress while restoring thermal stress resistance under reduced protein synthesis, relative to aak-2 mutants (S8A, S8C, S8J, S8L, S8M and S8O Fig). Intriguingly, expression of a constitutively activated version of AMPK (CA-AMPK [44]) restored the reduction of pharyngeal pumping phenotype when protein synthesis was reduced (S8D and S8E Fig). Taken together with the AMP/ATP and ADP/ATP levels (Fig 4G), these data define an AAK-2/AMPK molecular pathway that initiates organismal-level physiological responses to cellular deficiencies in protein synthesis. Importantly, our studies reveal a clear role for AMPK signaling in mediating the survival responses to protein synthesis inhibition beyond developmental arrest.
In the context of a worm’s natural environment, we postulated that the ability to pause development in response to a perceived cellular deficiency would be advantageous—and perhaps evolved—as a response mechanism to deal with environmental hazards. In the wild, C. elegans consume diets that are far more complex than the simple and homogenous E.coli lawn provided to them in the laboratory [1]. These wild diets include heterogeneous populations of microorganisms, some of which can produce xenobiotic compounds that can target and disable essential biological pathways. Recently, the soil and intestinal microbiome of C. elegans has been characterized [45–47]. While only appearing at rates ranging from 0.001–0.1% in soil samples found in these studies, we chose to focus on the genus Streptomyces, as it is soil-dwelling, readily accessible with the lowest biosafety level, and has several members that produce commonly utilized molecules that can potently inhibit eukaryotic protein synthesis [48]. If wild C. elegans came upon a microcosm of Streptomyces species, or any other organism capable of producing xenobiotics that reduce protein synthesis, it would be important to have defenses available against these molecules. We exposed worms to S. griseus, S. griseolus, or S. alboniger, that produce cycloheximide (CHX), anisomycin, and puromycin, respectively (S9A Fig). Exposure to these Streptomyces species grown under stationary conditions for five days, in order to initiate secondary metabolism and the creation of these protein synthesis inhibition molecules [49], resulted in delayed reproduction (S9B Fig) and significant reduction of their pharyngeal pumping in two species (Fig 5A). This is in contrast to exposure with microbes in exponential phase growth which attenuates secondary metabolism [49] (Fig 5A).
Exposure to pathogens can alter several physiological parameters in the host, and of all the pathogens tested, exposure to S. griseus exerted the strongest influence on pharyngeal pumping. The remarkably similar impact that exposure to S. griseus had on C. elegans development and physiology, as compared to RNAi-induced protein synthesis inhibition, drove a further examination of how exposure to cycloheximide (CHX), the bioactive secondary metabolite produced by S. griseus, affected C. elegans survival during development. CHX is a potent inhibitor of ribosome processivity and has recently been shown to exert health-promoting effects in adult C. elegans by an unknown mechanism [50].
Satisfyingly, CHX exposure upon hatching, which inhibits new protein synthesis (S9C Fig), also resulted in arrested animal development (S9D Fig and Fig 5B), and can arrest in a dose-dependent manner (S9D Fig). Although animals arrested by RNAi-mediated protein synthesis inhibition can continue development upon removal from the RNAi state, not all animals in the population mature into fertile adults (S9E–S9G Fig)—likely a result of the persistence of RNAi [51–54]. However, initiating protein synthesis inhibition via exposure to 0.05mg/ml CHX rather than RNAi of essential protein synthesis factors (S1A Fig) enabled studies of recovery from the arrest state without the complications of RNAi. Once removed from the xenobiotic, developmentally arrested animals resume development—indicating the arrest state is truly transient (Fig 5B, S2 Table). The CHX-induced arrest state caused reduced pharyngeal pumping (Fig 5C), arrested germ cell proliferation (Fig 5D), increased organismal [AMP]/[ATP] ratio (Fig 5E). Importantly, this arrest state phenocopied all RNAi-based protein synthesis inhibition survival responses (Fig 5F) including: enhanced resistance to oxidative (S9H Fig) and thermal stress (S9I Fig), induced the expression of mtl-1 (S9J Fig) decreased cellular ATP (S9K Fig), and resulted in metal profiles similar to animals fed RNAi targeting egl-45/EIF3 and rps-11/RPS11 (S9L and S9M Fig). 0.05mg/ml CHX exposure may not fully arrest all animals, as some daf-2(e1368) animals at the restrictive temperature did become dauers (S2Q Fig). Animals that are released from CHX arrest have minimal (if any) changes in reproductive output (S9N Fig), have a small but significant increase in resistance of oxidative stress (S9O Fig), are delayed ~16-20hrs to reproduction (S9P Fig), and have normal pumping rates at physiological day 3 of adulthood (S9Q Fig). Thus, this transient arrest state is survival promoting when the deficiency in protein synthesis is present and is not afforded once homeostasis is reestablished, similar to animals released from dauer [36]. Intriguingly, the ability of Streptomyces griseus to reduce pharyngeal muscle pumping required the presence of live bacteria co-culture (S9R Fig). In addition, increasing doses of CHX, similar to the threshold effects seen with RNAi targeting genes involved in protein synthesis (S1E Fig), could further reduce the pumping rate of the arrested animal (S9S Fig). Thus, the complexity of the environment and drug dosage are important for balancing the induction of this survival state.
In response to impaired organismal protein synthesis, animals are capable of entering an arrest state, reaping survival benefits, and exiting to become reproductive adults (Fig 5B). In our studies, we are forcing continual exposure of animals to protein synthesis inhibiting RNAi or xenobiotics, which is likely "unnatural", as previous studies of lethal RNAi treatment and xenobiotic treatments leads to aversion behaviors [15, 17]. With this in mind, we predict that in the wild the perceived loss of translation would evoke a similar aversion response—allowing animals to escape to new pathogen-free environments. This model is supported by our studies with cycloheximide exposure, which drives a rapid induction of arrest and stress resistance, from which animals can quickly recover. In this regard, we believe that the use of cycloheximide as a transient inducer of protein synthesis inhibition in the worm will be of great use in studying protein synthesis inhibition going forward in order to circumvent the complications of RNAi expansion over the worm lifespan and subsequent generations. Given that there is a dose response to CHX exposure, higher doses can be utilized to prolong the arrest state and enhance arrest phenotypes although prolonged exposure to higher concentration reduces the rate of escape (S2 Table).
The lack of necessity of DAF-16 for the developmental arrest in response to protein synthesis inhibition indicates that the reduced protein synthesis pathway functions independently from the dauer development pathway. Yet, while most dauer constitutive daf-2 mutants that are arrested from CHX do not form dauers, intriguingly ~20–25% will develop into dauers instead of undergoing protein synthesis arrest (S3D Fig). This finding suggests that animals can either alternatively arrest in the L2d stage [33, 55, 56], or that the CHX dose requires a higher threshold for complete arrest of animals (especially given the 100% non-dauer RNAi-treated animals). Of note, reduced protein synthesis arrested animals are distinct from the L2d stage as they are of smaller length than wild type L2s (S1G Fig) (unlike 50% longer L2d animals [33]), functional AMPK is not necessary for the reduction of germ cell numbers (S8B Fig) as it is in L2d/dauer animals [57], and we have never observed them becoming dauers after exiting the arrest state. Future characterization of any phenotypic parallels between L2d and reduced protein synthesis arrest, especially in the context of the differing role of AMPK in controlling germ cell proliferation, will be of interest for future studies.
A persistent question in biology asks how cellular status is communicated across the organism and, more importantly, how an appropriate homeostatic response is engaged. Protein synthesis inhibition in the hypodermis alone was sufficient for all arrest and healthspan phenotypes. In addition to its important role in the molting process during larval development, the hypodermis has recently been implicated as being important in dietary checkpoints in larval arrest [9, 58]. Although it is known that C. elegans tissues have differential capacity for RNAi, our work bolsters the hypodermis as a key tissue in larval development, and identifies a new cell non-autonomous communication pathway to initiate systemic responses. Given that the hypodermis is the first barrier to its external environment that covers the entire organism, it is reasonable that C. elegans might evolve sensing mechanisms for hypodermal cellular changes to influence whole-body cellular signaling. It is also possible that the high demand for protein synthesis during growth of the developing hypodermis amplifies the tissue-general effects of protein synthesis inhibition in this tissue, with or without specifically evolved signaling pathways. However, proliferation alone is not the only factor that influences responses to protein synthesis inhibition. The germline is a highly proliferative tissue in C. elegans, and while protein synthesis inhibition in the germline did not result in the same L2 arrest state as tissue-general or hypodermis-specific reduction, it did result in pre-reproductive adult animals with mild stress resistance (S4 Fig). It remains to be seen if this germline arrest is also reversible, similar to starvation-induced adult reproductive diapause [4].
It is important to note the differences in stress resistance when protein synthesis is reduced in specific tissues. While hypodermis-specific RNAi of protein synthesis components results in increased stress resistance that is consistent when RNAi is initiated in all tissues, intestine-specific RNAi resulted in no change to stress resistance capacity except for a few instances of increased resistance only observed for rps-11 RNAi. The more tissue-general expression of rps-11/RPS11 (S4 Fig), may explain these minor phenotypic differences as compared to egl-45/EIF3 RNAi. Taken together, these data support the idea that the systemic stress responses that stem from the loss of rps-11 are mediated by effects across multiple tissues. In contrast to the hypodermis and intestine, germline-specific loss of protein synthesis resulted in modest or no changes in oxidative stress resistance and surprisingly lead to reduced thermal tolerance. This suggests that the oxidative and thermal stress resistance responses, at least in the germline, may be uncoupled or, alternatively, that reducing protein synthesis in the germline activates a separate pathway that negatively affects thermal stress resistance. Finally, it is also worth noting that there is considerable variation in stress resistance among these tissue-specific RNAi strains. We attribute much of this both to the use of RNAi variance, as well as the ever-present "leakiness" of these tissue specific strains that can sometimes spread RNAi effects to other tissues [59, 60].
The metallothionein, mtl-1, is highly (>100-fold) upregulated under reduced protein synthesis. The increased expression of mtl-1 was required for heavy metal resistance in animals with protein synthesis inhibition, which is notable since hypersensitivity to cadmium has not been reported in adult C. elegans lacking MTL-1 or MTL-2 [61]. This finding further advocates for the importance of uncoupling developmental and adult specific responses. Transcription of MT1, the mammalian homolog of mtl-1, is also upregulated by oxidative stress agents in cell lines and mice [62, 63], so it is possible that protein synthesis inhibition causes an increase in ROS that triggers mtl-1 transcription; however, then we would also expect to see increased transcription of SKN-1 target genes (e.g. gst-4), which we do not observe. Moreover, mtl-1 expression was not necessary for the arrest, oxidative or thermal stress resistance, or reduced pumping, as daf-16 mutants (which lack mtl-1 expression, S8 Fig) still display both phenotypes. Thus, given the very specific transcription of mtl-1, the changes in expression are likely due to the presence of its most well-defined binding partners, metal cations. Traditional targets of MTL-1 are Zn2+, Cd2+, and Cu2+, but mammalian homologs can bind to Mg2+, Mn2+, and Ca2+ [64–66]. The increase in Ca2+ ions could be the cause of this high transcriptional response, especially given that Ca2+ treatment could induce mtl-1 in worms (Fig 3E). However, it is also possible that higher levels of other heavy metals, such as Cd2+, which never reached our detection limits, are responsible. Given that mtl-1 expression was disposable for the arrest, stress resistance, and reduced pumping rate, the increased expression change is a "biomarker" for the reduction of protein synthesis, rather than a central player in this developmental state. Given the ability for calcium to upregulate this mtl-1 response (Fig 3), we expect the protein synthesis loss triggers calcium abundance and daf-16 activation [16], that both go on to increase mtl-1 levels.
It is possible that the reduction of cellular ATP we observe reflects the use of ATP to “power” survival processes [42]. However, a ~50% reduction in ATP after 24hrs of protein synthesis inhibition is a remarkable loss, and it would not explain how this energy usage would be sustained to continue stress resistance over extended time periods, especially when accompanied by a reduction in pharyngeal pumping (thereby reducing food/energy intake even further). Our data support an alternative model where increases in the [AMP]/[ATP] and [ADP]/[ATP] ratios activate AMPK pathways that signal for downstream survival pathways (Fig 4C and 4F). The underlying mechanism driving the imbalance to cellular adenylate pools will be of future interest.
We found that AMPK was necessary for all of our protein synthesis inhibition survival phenotypes, except for arrest. AMPK activation has been implicated in survival phenotypes before, including glucose restriction pathways [67] and oxidative stress resistance [68] in C. elegans. Juxtaposed to our work, activating AMPK (such as via AICA ribonucleotide) causes a decrease in protein synthesis [23–25]. While our work focuses directly on protein synthesis alone, AMPK is also increased in rsks-1/S6K mutants [27, 28] and under starvation conditions [26]. This suggests that AMPK and protein synthesis may work together in a circular pathway or that they affect each other by cell non-autonomous signaling. In addition, an upstream activator of AMPK, ARGK-1, is both important for rsks-1/S6K mutant longevity, and its overexpression caused reduced pumping rates in worms [69]; further study into the role of ARGK-1 in this protein synthesis inhibition survival state will be of interest in future studies. As a final note, C. elegans lacking the elongation factor efk-1, which is activated by AMPK, fare worse under nutrient starvation conditions [70]; thus, there are multiple connections between starvation, protein synthesis, and energy homeostasis, and understanding them in context of survival states is important to consider.
Previous studies suggest that the effects of protein synthesis inhibition on adult lifespan are distinct from caloric restriction (CR) [19] and that the CR state can drive a reduction in protein synthesis[20]. Our data suggest that during development the opposite is also true: that protein synthesis inhibition can reduce pharyngeal pumping leading to a CR-like state. CR across most organisms has both life- and healthspan promoting effects; however, the evolutionary basis of the CR response is unknown. One hypothesis generated from this study is that the physiological response to CR might stem from an ancient program to promote stress resistance when the presence of diminished biosynthetic capacity is perceived. Microorganisms such as Streptomyces provide a potential evolutionary explanation to a mechanism of a pathogen-derived CR pathway by engaging behavioral avoidance phenotypes toward toxin-producing pathogens [15]. It is important to note that Streptomyces was found at very low levels in recent studies looking at C. elegans soil samples [45–47]. Our xenobiotic experiments are not meant to emulate the wild environment, but to capture the interaction between the worm and a harmful species in the environment. It is altogether possible that there are areas (or times in history) where Streptomyces, or other species capable of inhibiting host protein synthesis, are a more common occurrence, demanding the need for such an arrest survival response documented here.
There are connections between immune function and the regulation of protein synthesis—both to exposure to protein synthesis-impairing xenobiotics (ExoA, Hygromycin) as well as potential surveillance mechanisms for reduced protein translation as a surrogate for infection [71–73]. Pathogen response pathways can also be closely linked to promoting proteostasis [74]. In addition, a recent study found that C. elegans can enter a diapause to avoid pathogens (unlike our study, this is reliant on the formation of dauers [75]). Nevertheless, our findings support the idea that the loss of protein synthesis might be perceived as "an attack" by a pathogen, which initiates a reduction in pharyngeal pumping, that could minimize ingestion of toxin-producing microbes. Given the remarkable overlap in phenotypes resulting from protein synthesis inhibition by pathogen-derived xenobiotics and our genetic and RNAi-mediated protein synthesis inhibition, it is suggestive that this survival-arrest state could have evolved as a stress response to the presence of pathogens (Fig 6). This idea parallels models of adult longevity pathways, which may have connections to xenobiotics targeting other essential pathways besides protein synthesis [35].
Unlike previous models that suggest the developmental arrest resulting from early loss of protein synthesis is a detrimental state [42], these studies provide an alternative way of thinking about these developmental responses. The induction of protective responses to reduced protein synthesis is survival-promoting, and we predict that the capacity to engage these pathways would enable future opportunities for reproduction once the inhibition is alleviated. Lastly, our results provide an example of how the evolution and selection of developmental pro-fitness pathways may be utilized effectively later in life under the right conditions. Just as dauer diapause from reduced insulin/IGF-1 signaling (IIS) has mechanistic similarities with adult longevity responses when IIS is reduced post-developmentally, our studies establish a similar fitness-driven developmental program as the underlying mechanism of the enhanced healthy aging observed in adults with compromised protein biosynthetic capacity. The exceptional degree of conservation of these cellular pathways across organisms is suggestive that the pre- and post-developmental responses to protein synthesis inhibition observed in C. elegans could be similarly shared, even among humans.
Worm strains were grown at 20°C for all experiments except dauer studies that were conducted at 25°C. All strains were unstarved for at least 3 generations (except for L1 synchronization) before being used in any experiments. List of strains used: N2 Bristol (wild type), DR1572 daf-2(e1368), GR1329 daf-16(mgDf47), MGH171 (sid-1(qt9); Is[vha-6::sid-1::SL2::gfp], JM43 (rde-1(ne219); Is[wrt-2p::rde-1], myo-2p::rfp]), NL2098 (rrf-1(pk1417)), GR1395 (mgIs49[mlt-10p::gfp-pest, ttx-3::gfp]IV]), SPC365 mtl-1(tm1770); mtl-2 (gk125), RB754 (aak-2 (ok524)), SPC366 (aak-2(ok524); uthIs248[aak-2p::aak-2(genomic aa1-321)::GFP::unc-54 3'UTR (gain of function allele); myo-2p::tdTOMATO]), SPC363 (Ex[egl-45p::rfp; rol-6(su1006)]), SPC364 (Ex[rps-11p::gfp; rol-6(su1006)]), CL2070 (dvIs70[hsp-16.2p::GFP; rol-6(su1006)]), KX38 (ifg-1(ok1211)/mIn1 [mIs14 dpy-10(e128)]). Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440).
E. coli strain HT115 (DE3) containing empty vector L4440 (hereafter referred to as Control RNAi), or plasmid against a gene of interest, was grown overnight (16-18hrs) at 37°C and seeded on NGM plates containing 5mM isopropyl-β-D-thiogalactoside (IPTG) and 50ug/ml carbenicillin. The bacteria were allowed to generated dsRNA overnight before being used within the next 1–3 days (stored at 20°C for this period if not used immediately). Dose response curves were established by feeding HT115 bacteria expressing the indicated RNAi clone diluted with HT115 bacteria harboring the control RNAi plasmid L4440. 0.05mg/ml Cycloheximide (CHX) or water (vehicle control) was added on top of bacteria and allowed to dry and rest for at least 1 hour before placing worms on treated bacterial lawns; this was the concentration of CHX throughout this paper, unless otherwise noted.
Loss of protein synthesis was determined via measurements of de novo synthesis of GFP through both an internal (via natural development) and external (via high temperature) induction method. External: plated animals expressing hsp-16.2p::GFP were maintained at 20°C and fed RNAi since hatching. After 24hrs, one set of worms was shifted to 36°C for 3hr, while the other was mounted for the baseline 0hr time point. The baseline plate was also checked after 3 hours at room temperature as a control for any room temperature-induced GFP expression. Internal: plated animals expressing mlt-10p::gfp-pest, treated with RNAi or drug since hatching, were imaged via the same methods for GFP expression at 12, 14, and 16 hours post-feeding. Worms were imaged at 20x magnification with bright field and GFP filter (Zeiss Axio Imager).
Plated animals, treated with drug or RNAi since hatching, were counted in 24 hours intervals via a compound microscope as larval stage 1–3 (size), larval stage 4 (vulval invagination), adult (size), or reproductive (presence of internal eggs). In food switching assays, worms were moved to rde-1 RNAi after 24hrs on the listed RNAi. rde-1 RNAi was used to inhibit the RNAi machinery because RNAi effects can persist even after moving animals off of food containing double stranded RNA for multiple generations.
Plated animals, treated for 24 or 48 hours on drug or RNAi since hatching, were placed at 36°C for up to 12 hours. Every 3 hours, one set of plates was removed to room temperature. Worms were allowed to recover for at least 10 minutes, and then counted for survival immediately by checking for touch response to prodding with a platinum wire.
Plated animals, treated for 24 or 48 hours on drug or RNAi since hatching, were washed with M9 buffer twice in microcentrifuge tubes, then treated with 20mM H2O2 for up to 1 hour while rocking at room temperature. Every 20 minutes, one set of worms was removed from rocking, washed 3 times in M9 buffer, and plated back onto new plates containing their previous treatment (drug or RNAi). Worms were checked 1 hour after plating to count any acute deaths ("straight line" bodies or ruptured vulvas) only by eye, and 24 hours after plating to count final survival as done in thermotolerance assay.
Plated animals, treated for 24 hours on RNAi since hatching or at L4/YA stage, were washed with K-medium (32mM KCl, 51mM NaCl in dH2O) twice in microcentrifuge tubes, then treated with 5 or 50mM CdCl2 in K-medium (hatched or YAs, respectively) for 30 minutes while rocking at room temperature. After 30 minutes, worms were washed 3 times in K-medium, and plated back onto new plates containing their previous treatment (RNAi). Worms were checked 1 hour after plating to count any acute deaths ("straight line" bodies or ruptured vulvas) only by eye, and 24 hours after plating to count final survival as done in thermotolerance assay.
Wild type and daf-2(e1368) were placed as synchronized L1s onto the listed RNAi clone or drug at 25°C for 48hrs. Worms were then washed in M9, pelleted, and treated with 1% for 30min while rocked at room temperature. Treated animals were then plated onto plates with HT115 bacteria and counted for survival.
Drug- or RNAi-treated animals were washed with M9 buffer twice in microcentrifuge tubes, then frozen at -80°C in TRI-Reagent® (Zymo Research, R2050-1-200). After at least 24 hours at -80°C, RNA was extracted from samples using the Direct-zol™ RNA MiniPrep kit (R2052). Quantitative reverse transcription PCR (qRT-PCR) was performed on the RNA samples with gene specific primers (Table 1).
For evaluation of mtl-1 induced by calcium, wild type animals, grown for 24 hours on Control RNAi, were washed with K-medium twice in microcentrifuge tubes and then treated with 500mM CaCl2 (in K-medium) for 30 minutes at room temperature. Animals were then washed three times with K-medium, frozen at -80°C in TRI-Reagent® as above, and the same protocol as above was utilized.
Two 24-well plates, each containing a single GR1395 worm on RNAi or Control RNAi, were visualized by fluorescence microscopy every hour for 72 hours. Worms were marked as green or non-green to indicate molting or non-molting, respectively. Worms that crawled off the side of the plate or burst were censored.
Plated animals, treated for 24 hours on RNAi or drug since hatching, were imaged at 20x magnification (Zeiss Axio Imager), and individual germ cells were counted with the Cell Counter plugin on Fiji software [76].
Plated animals, treated for the indicated time on drug or RNAi since hatching, were imaged via the Movie Recorder at 8ms exposure using the ZEN 2 software at 10x magnification (Zeiss Axio Imager). Animals with zero pumping were excluded.
1000 or 500 plated animals, treated for 24 or 48 hours on drug or RNAi since hatching respectively, were washed 3 times in M9 buffer (keeping ~100μl of supernatant at final wash), snap frozen in a dry ice/ethanol bath, and placed at -80°C until use. Frozen pellets were boiled for 15 minutes and spun down at 14,800g at 4°C. The supernatant was then diluted in dH2O (1/10) (Adapted from[77]). Samples were tested for protein content via Bradford analysis (Amresco M173-KIT), and ATP was assessed via the ENLITEN® ATP Assay System (Promega).
To determine relative levels of ATP/ADP/AMP, we followed the same method as above, but did not dilute the supernatant. Protein supernatant was directly assayed via the ATP/ADP/AMP Assay Kit (University at Buffalo, Cat. # A-125) to determine total ATP/ADP/AMP in each sample; these values were then directly compared to determine relative ratios.
8,000–10,000 (L4 stage) or 20,000–25,000 (L2 stage) animals, treated for the listed time on the listed RNAi clone, were collected into microcentrifuge tubes (tubes weighed beforehand) using isotonic buffer (150mM Choline Chloride, 1mM HEPES, pH 7.4 with NaOH, filter sterilized). Worms were washed 3 times over 30 minutes (pelleting at 1,000g/30s each time) to clear gut content and then finally pelleted at 12,000g/2min. Worm pellets were then dried at 60°C for 48 hours using a heat block. Worm pellets were weighed after drying, and ICP analysis of the samples was conducted by Dr. David Kililea, Children's Hospital Oakland Research Institute. Before ICP analysis, dried pellets were acid digested with Omnitrace 70% HNO3 at 60°C overnight. Samples were diluted with Omnitrace water for a final concentration of 5% HNO3. Derived metal content was normalized to dried worm pellet weights. Each animal is compared back to 24hr Control RNAi treated animals. 48hr Control RNAi animals are given as a reference for what the metal content of a chronologically matched animal would be; albeit animals that are L4-YA stage and thus 2–3 larger with higher food intake.
Streptomyces Alboniger (ATCC 12461), Griseus (ATCC 23345), or Griseolus (ATCC 3325) were grown at 26°C, shaking, in Tryptone-Yeast Extract Broth (5g Tryptone, 3g Yeast Extract in 1L dH2O, pH 7; taken from ATCC® Medium 1877: ISP Medium 1) for 5 days before plating unless otherwise noted. Strains were plated on Yeast Malt Agar plates (HiMedia Laboratories, M424), and mixed 1 part to 3 parts 25x HT115 when used with worms. For the egg laying comparisons, 100ul Saccharomyces cerevisiae was also added to induce competition; to compare total number of eggs, worms were mounted at ~52hrs after dropping to food source, and imaged at 20x magnification with DIC (Zeiss Axio Imager). For testing dead HT115, 75ml/L of 2.5% Streptomycin was added to 25x HT115 and the mix was rocked for 24hrs at room temperature. This mixture was then used in place of the 25x HT115 above.
For survival in the arrested state, worms were dropped on the listed RNAi and counted each day (for the majority) for survival. Survival was assessed by touch response to prodding with a platinum wire. The Control RNAi wild type control strain used in this experiment was moved each day starting at adult day 1 as necessary until reproduction ceased.
For tissue-specific lifespan analysis, worms were grown on Control RNAi until L4/young adult age, and then transferred to the listed RNAi plates treated with 50μM FUdR. Survival was assessed every other day as above.
For all assays, animals were only censored (bursting, vulval protrusion, etc.) after the first counted death.
Worm morphological comparisons were imaged at 20x magnification with DIC filter (Zeiss Axio Imager). Worm length comparisons were made in ImageJ using the segmented line tool down the midline of each animal from head to tail.
For GFP and RFP reporter strains, worms were mounted in M9 with 10mM Sodium Azide, and imaged at 40x magnification with DIC and GFP/RFP filters (Zeiss Axio Imager). Fluorescence is measured via corrected total cell fluorescence (CTCF) via ImageJ and Microsoft Excel. CTCF = Integrated Density–(Area of selected cell X Mean fluorescence of background readings).
For imaging of heat-induced GFP expression via strain CL2070, plated animals were maintained at 20°C and fed RNAi since hatching. After 24hrs, one set of worms was shifted to 36°C for 3hr, while the other was mounted (as above) for the baseline 0hr time point. The baseline plate was also checked after 3 hours at room temperature as a control for any room temperature-induced GFP expression. Worms were imaged at 20x magnification with bright field and GFP filter (Zeiss Axio Imager).
Thermotolerance, oxidative stress, and heavy metal stress were all compared using Fisher's Exact Test using the statistical software R [78]; specifically, the bars in each graph represent a unique set of biological replicates (2–6 independent biological replicates, see S1 Table) relative to its own independent control cohort (and the significance level relative to this control is indicated by the # of stars above each bar); this test is employed as we are comparing the categorical variables of Alive vs Dead, and data is presented as changes in survival. Comparison of all RNAi clones and CHX for protein synthesis rates under the mlt-10p::GFP promoter was performed using one-way ANOVA. Lifespan curves were compared and analyzed via Log-Rank using JMP Pro 12. qPCR, worm fluorescence, metal content, ATP/ADP/AMP levels, and pharyngeal pumping comparisons were made with Student's t test using Microsoft Excel. When comparing groups of three or more, Bonferroni multiple comparison post-correction was employed on Fisher's test, ANOVA, and t tests.
|
10.1371/journal.pntd.0004126 | Animal Reservoirs of Zoonotic Tungiasis in Endemic Rural Villages of Uganda | Animal tungiasis is believed to increase the prevalence and parasite burden in humans. Animal reservoirs of Tunga penetrans differ among endemic areas and their role in the epidemiology of tungiasis had never been investigated in Uganda.
To identify the major animal reservoirs of Tunga penetrans and their relative importance in the transmission of tungiasis in Uganda, a cross sectional study was conducted in animal rearing households in 10 endemic villages in Bugiri District. T. penetrans infections were detected in pigs, dogs, goats and a cat. The prevalences of households with tungiasis ranged from 0% to 71.4% (median 22.2) for animals and from 5 to 71.4% (median 27.8%) for humans. The prevalence of human tungiasis also varied among the population of the villages (median 7%, range 1.3–37.3%). Pig infections had the widest distribution (nine out of 10 villages) and highest prevalence (median 16.2%, range 0–64.1%). Pigs also had a higher number of embedded sand fleas than all other species combined (p<0.0001). Dog tungiasis occurred in five out of 10 villages with low prevalences (median of 2%, range 0–26.9%). Only two goats and a single cat had tungiasis. Prevalences of animal and human tungiasis correlated at both village (rho = 0.89, p = 0.0005) and household (rho = 0.4, p<0.0001) levels. The median number of lesions in household animals correlated with the median intensity of infection in children three to eight years of age (rho = 0.47, p<0.0001). Animal tungiasis increased the odds of occurrence of human cases in households six fold (OR = 6.1, 95% CI 3.3–11.4, p<0.0001).
Animal and human tungiasis were closely associated and pigs were identified as the most important animal hosts of T. penetrans. Effective tungiasis control should follow One Health principles and integrate ectoparasites control in animals.
| Tungiasis is a neglected skin disease, frequent in resource-poor communities in the tropics. It is caused by sand fleas (also called jigger fleas) which burrow in the skin of humans and animals. Tungiasis can cause physical disabilities and the associated wounds facilitate entry of pathogens including those causing tetanus. In Brazil, presence of tungiasis in animals increased the risk and severity of the human disease. Until now, no systematic studies on animal tungiasis in Africa have been conducted. Therefore, an epidemiological study was performed in Busoga sub-region, where tungiasis in humans is very common. Tungiasis was detected in pigs, dogs, goats and a cat, respectively, in their order of significance. Animal tungiasis was strongly associated with human tungiasis and the presence of the disease in animals increased the risk of human infection by a factor of six. Our findings confirmed, for the first time, a strong correlation between the presence of tungiasis in animal reservoirs and the human population in Africa. Therefore, control of tungiasis in animals should be integrated in all interventions geared at controlling tungiasis in endemic communities.
| Tungiasis is an ectoparasitosis that accrues from the penetration of female sand fleas into the skin. Tunga penetrans [1] and Tunga trimamillata [2] are the only species known to cause tungiasis in both humans and animals. Currently, zoonotic tungiasis is endemic in southern America, the Caribbean and sub-Saharan Africa. While T. penetrans occurs in all endemic areas, T. trimamillata has only been reported in a few countries in South America [3,4]. In the endemic areas, tungiasis in humans is heterogeneously distributed [5–7]. In Uganda, human tungiasis occurs in all regions but the prevalence appears to be particularly high in the Busoga sub-region, South Eastern, and Karamoja in North Eastern, Uganda [8]. These regions are among the poorest in the country.
Epidemiological studies carried out in resource-poor communities in Africa and South America have reported point prevalence of up to 60% among humans [6,7,9]. Hitherto, no systemic investigations have been carried out in Uganda but an impromptu outbreak investigation in some parishes of Busoga sub-region in 2010 reported prevalences of up to 73% in the general population [8]. For Karamoja, a study conducted in Napak District in different seasons, reported prevalences ranging from 18.4% at the end of the rain season to 56.6% in the dry season [10]. In poor communities, tungiasis is associated with severe morbidity [11] leading to physical disability and immobility [12]. In addition, in non-vaccinated individuals, tungiasis predisposes to tetanus and may contribute to transmission of blood borne pathogens such as Hepatitis B Virus (HBV) and HIV if non-sterile instruments are used to remove embedded sand fleas and are subsequently shared between household members [13]. Deaths from tungiasis-related complications are commonly reported in Uganda [14].
T. penetrans infects a wide range of domestic, peri-domestic and wild mammals such as pigs, dogs, cats, goats, cattle, rodents, elephants, jaguars, monkeys and even armadillos. The relevance of each of the animal species in the epidemiology of human tungiasis varies from one endemic area to another. While in urban Brazil, dogs, rodents and cats are the species most frequently infected by T. penetrans, in West Africa pigs appear to be the important animal reservoirs [3,15–17]. In Brazil, infected animals seem to increase the risk of infection in humans and are associated with a high prevalence of tungiasis at a community level [15]. Although the economic significance of T. penetrans infections in animal production has not been systematically studied, existing literature points out a significant effect on growth rate, secondary bacterial infections and defects of limbs [3]. Tungiasis may also lower product quality and hence, marketability of animals. In sows, it has been reported to cause agalactia with subsequent starvation of piglets if it affects their mammary glands [18]. Obviously, poor production and decreased marketability perpetuate community impoverishment.
To date no systematic studies have been conducted to describe the epidemiology of tungiasis in animals in East Africa. In order to identify the major animal reservoirs of T. penetrans in rural Uganda and to investigate the association between animal and human disease, a cross sectional study was carried out in ten endemic villages located in Bugiri District, Busoga sub-region. The study revealed that the prevalence of animal tungiasis was high and that the disease prevalence and parasite loads in humans and animals correlated. Pigs were identified as the most important domestic animal hosts for T. penetrans.
The study was carried out in ten villages situated in Bulidha sub-county, Bugiri district, Busoga sub-region in South Eastern Uganda. Bugiri district was purposively selected amongst the ten districts of Busoga because of the high prevalence of human tungiasis reported and confirmed during a preliminary survey. Bugiri district is about 178 km in the South Eastern direction away from the capital Kampala and lies between longitude 33°10’ and 34°00’ East and latitudes 0°6’ and 1°12’ North [19].
The ten villages were; Masolya, Makoma 1, Busakira, Busano, Isakabisolo, Namungodi, Matyama and Busindha situated in Makoma parish; Kibuye and Nagongera in Wakawaka and Bulidha parishes, respectively. These were purposively selected because human tungiasis was reported to be highly prevalent by the local health personnel, a fact which was also verified during a preliminary visit. The study area and study sites with infected hosts are illustrated in Fig 1. Since there were no estimates regarding the size of the animal populations and prevalence of T. penetrans infections in the various animal species, a relatively large number of villages was included in the study. All households in the villages with at least a pig, a dog or a cat were selected for the study. All mammals accessible in the selected households were examined. Poultry were examined whenever available.
The communities were constituted by seven tribes belonging to three ethnic groups, namely Luo (Japadhola), Bantu and Nilo-hamites (Itesot). People depend on rain-fed subsistence crop and livestock agriculture. Other economic activities such as fishing in Lake Victoria and temporary work in sugar and tree plantations also contribute to households’ incomes. The major crops grown in the area were maize, cassava, rice, coffee and bananas while goats, pigs and some cattle were the major livestock reared. Dogs, cats, sheep and rabbits were other domestic animal species found in the villages. In addition, a variety of poultry especially chicken and ducks, and to some extent pigeons, turkeys as well as guinea fowls are raised. Dogs, cats and poultry roam freely on compounds throughout the year. With the exception of unweaned young animals, which roam on compounds unrestricted, other livestock are tethered on or near compounds during the crop production seasons. However, they are released intermittently after harvest with minimum food supplementation.
Homesteads are located on relatively large compounds that are close to gardens or bushes where household waste is dumped. All roads and paths are made of murram. The area experiences two rainy seasons; one between April and June and the other from August to November with an average annual rain fall of 1200 mm. Average daily temperatures range from 16.7°C to 25.1°C. Water is mainly obtained from communal boreholes, springs and shallow wells. Electricity is limited to major social facilities and less than 1% of the households have electricity.
A cross sectional study was conducted between January 22 and March 28, 2014 which coincided with the middle and end of the dry season when the attack rate of T. penetrans usually peaks [20]. Initially, a mission was undertaken to explain the study objectives to the local medical and veterinary health personnel as well as to the local leaders. Since the study aimed to compare the significance of different animal reservoirs, only animal rearing households were included. All households with at least one pig, dog or cat were included as these species have been reported to be the most important animal hosts of T. penetrans in sub-Saharan Africa [15,17]. Households meeting the selection criteria were located with the guidance of local leaders.
In each consenting household, first a census of the animals and humans was performed. Then all mammals (dogs, pigs, cats, cattle, sheep, goats and rabbits) as well as all humans present on the compound during the investigator’s visit were examined for tungiasis. Only the poultry that were accessible were examined. If the household leader was not around, the household was visited again. Households were also revisited (up to three times) when any household member was not present; when some pigs, dogs or cats identified in the census could not be traced or when they ran away during the first investigation. In Masolya and Makoma 1, which were randomly selected among the ten study villages, all remaining households with at least one goat were also sampled to obtain an unbiased sample of goats. Wherever residents claimed to have seen rats with T. penetrans, it was attempted to trap rats in cages placed in household premises and close to the entrances of termite moulds.
The study objectives were explained to the household heads and informed written consent was obtained. Thereafter, data was collected on social, environmental, behavioral and animal management practices through interviewing the household head and observations. Then humans and animals were examined for T. penetrans-associated lesions. Diagnosis was made clinically and humans were examined by means of a rapid assessment method [21]. To estimate the intensity of T. penetrans in humans, lesions of a randomly selected foot of up to three randomly selected children between three and eight years of age per household were counted since this is the age group with the highest intensity of infection [6,7].
To perform a thorough clinical examination, mammalian and avian species were restrained physically. However, most dogs and cats could only be examined after sedation with ketamine (Umedica Laboratories PVT. LTD, India) and xylazine (Xyla, Interchemie Werken, Netherlands). Vomiting during sedation was prevented using atropine (Gland Pharma, M. L. 103/AP/RR/97/F/R). Examination of animals was systematically performed from the head, along the trunk to the tail including the lower abdomen and the limbs through observation, hair parting and palpation. Particular emphasis was given to the paws and digits for canines and ungulates, respectively. To increase the visibility of T. penetrans lesions, the distal body parts were scrubbed with a brush and water. Sex, age and breed together with the findings of a complete clinical examination of infected animals were recorded on a standardized form. Detailed information regarding infected animals such as age was obtained by asking the owners since none of the households kept written animal records. All trapped rats were euthanized by wrapping the cages in a piece of cloth immersed in diethyl-ether.
Staging of lesions was performed according to the Fortaleza classification [22]. Viable stages were characterized by: presence of a dark brown to black spot surrounded by a reddened or swollen area (stage II) and a raised yellow to white nodule of 2–13 mm in diameter with a dark center in the skin (stage III). A brown to black, circular, raised patch in the middle of a necrotic area with or without erosions or ulcers (stage IV) and an epidermal circular shallow crater with necrotic edges (stage V) were the features considered to reflect dying or dead sand fleas [22]. In humans and animals, sores indicating that an embedded parasite had been manipulated were also documented. Photographs were taken to document the findings.
A total of 16 (two, four and ten) embedded sand fleas were carefully extracted from some goats, dogs and pigs respectively. Animals were chosen from different villages and extraction was performed by enlarging the flea pore with tweezers. Sand fleas from humans were obtained from consenting humans who were transported by car, for treatment to Bulidha Health Center III, which was the nearest Health Unit. All extracted sand fleas were preserved in 70% ethanol and examined at the College of Veterinary Medicine, Animal Resources and Biosecurity (COVAB) using a light stereo-microscope by looking for characteristic features as described before [3].
Some alcohol-fixed sand fleas were exported to Germany, Freie Universität Berlin for scanning electron microscopy. Before electron microscopy, sand fleas were cleaned as described previously [23]. Briefly, sand fleas were dehydrated in ethanol for two hours and then kept overnight in acetone. The following day they were transferred to individual glass containers containing xylene before sonication for 30 minutes. Then, the sand fleas were washed in acetone for two hours before they were mounted on stubs, sputtered with gold and examined with the aid of a Zeiss Supra 40 VP scanning electron microscope at the Institute of Geological Sciences, Freie Universität Berlin.
Studies involving humans were conducted according to the “National Guidelines for Research involving Humans as Research Subjects” were approved by the Ministry of Health, Vector Control Division (reference no.: VCD-IRC/054). The ethical committee of the College of Veterinary Medicine, Animal Resources and Biosecurity (reference no.: VAB/REC/14/101) approved the studies involving animals. In addition, approval for both animal and human studies was obtained from the National Council of Science and Technology Uganda (reference no.: HS1621). Animal studies adhered to the “Animals (Prevention of Cruelty) act”, chapter 39, constitution of Uganda. Participation of humans was optional and written consent was obtained from household heads who also consented on behalf of their children as parents or guardians. All other adult household members (from 18 years and above) orally consented to the study, always obtained in the presence of the health workers from the nearest public health units. Only oral consent was obtained for these persons since the vast majority of participants were below 18 years and management of too many data forms was difficult under field conditions. Oral consent of the adult participants was documented on the evaluation sheets. This procedure was approved by the Ministry of Health and the National Council of Science and Technology, Uganda in the above stated documents. All humans with tungiasis were given a basic health kit consisting of a basin, a bar of soap, a sachet of detergent, individual towel, safety pins, cotton wool and an antiseptic (Dettol, Reckitt Benchiser, Dubai). Such a health kit is routinely supplied by the Ministry of Health of Uganda to affected households. Severely affected humans were transported by car to the nearest health unit for medical attention. Wounds on infected dogs, goats and pigs were cleaned with clean water and soap while antiseptic treatment was conducted using iodine tincture (SEV Pharmaceuticals Ltd, Kampala, Uganda) or a wound spray (Supona aerosol, Pfizer Laboratories (Pty) Ltd, South Africa).
Data was entered into an Excel database (Microsoft Office, 2007) and validated by checking all entries again using the data collection tools before exportation to Stata Software package, Version 13 (Stata corporation, College Station, Texas 77845 USA) or R 3.1.2 in R Studio 0.98.1103 via a csv text file. Either Chi-square or Fisher’s exact tests were used to determine the significance of differences between proportions. In case of multiple testing, p values were corrected with the Bonferroni-Holm method as implemented in the p.adjust function of R. The Spearman’s rank correlations coefficient was calculated to establish the relationship between pairs of continuous variables. The Wilcoxon rank sum test was used to compare differences in the number of lesions between animals and/or humans groups.
Household prevalence was calculated as the proportion of households with at least one infected household member or animal to that of the total households sampled. Tungiasis prevalence among animals and humans was computed as the proportion of the number of infected animals/humans to the respective number examined. For risk factor analysis, initially, bivariate logistic regression was performed to calculate odds ratios (ORs) to assess the association between the occurrence of the infection in animals and exposure variables. Confidence intervals (95%) with no continuity correction for the prevalence [24] were computed as Wilson score intervals using www.vassarstats.net/prop1.html.
Multivariate logistic regression analyses to identify factors with effect on the chance of occurrence of tungiasis in animals in general, pigs or dogs were conducted using the”glm” function in the R software. For identification of risk factors determining the occurrence of animal tungiasis irrespective of species, the variables used included sex of household head, ethnic group of household head, education level of household head, household size, homestead size, estimated annual income, human tungiasis, manure disposal distance from compound and method of manure disposal. Others included; number of animal species in households, presence of pigs, dogs, goats, cattle, cats, chicken and other poultry respectively as well as the period of rearing animals.
For analysis of risks factors of pig tungiasis in households with pigs, the variables; “presence of tungiasis in humans”, “number of animal species”, “number of pigs”, “distance of pigs from human housing”, “presence of other ectoparasites in pigs”, “number of dogs”, “presence of tungiasis in dogs”, “number of cats”, “number of goats”, “number of cattle”, “number of chicken”, “presence of other poultry” and “sanitation of the pig dwellings (clean vs. dirty)” were initially considered. The analysis of risk factors for presence of dogs with tungiasis in households used only households with dogs and started with the variables “presence of tungiasis in humans”, “number of animal species”, “number of pigs”, “presence of tungiasis in pigs”, “number of dogs”, “presence of other ectoparasites in dogs”, “number of cats”, “number of goats”, “number of cattle”, “number of chicken” and “presence of other poultry”.
Some variable such as “presence of tungiasis in cats or goats”, “ectoparasite control in pigs”, pig management system, type of floor of pig residence or “presence of rats” were not included since the numbers of households with these states were either very low or close to 100%. Significance of individual factors was determined using the t test statistic implemented in “glm”. The Akaike information criterion (AIC) was used to compare models and the “drop1” function in R was used to progressively identify variables that could be excluded from models. Finally, pseudo-R2 according to McFadden was determined.
In the 10 villages together, 236 households were selected using the criterion of having at least one pig, dog or cat. In addition, 26 and 31 households were selected in Masolya and Makoma 1, respectively, solely on the criterion of having at least one goat. The other goat owning households had been selected due to the presence of pigs, dogs or cats. Out of 158 households with pigs in the 10 villages, three (two in Busano and one in Busindha) declined to participate. Hence, 155 pig owning households were sampled. Only one household out of 121 with at least one dog declined to participate. All the 19 households with cats in the 10 study villages were sampled. All households with at least one goat in Makoma 1 were sampled but in Masolya one household out of 48 goat owning households was excluded due to the absence of the household head during three successive visits.
Overall, there were seven species of domestic mammals (pigs, dogs, cats, goats, cattle, sheep and rabbits) and five avian species (chicken, ducks, pigeons, turkeys and guinea fowls) being reared in the area. The total number of sampled households that had the various animal species and rat trapping sites in each of the 10 villages are shown in S1 Table.
The total number of animals of the different species varied greatly in the target villages (S2 Table). Pigs (median = 40.5, range = 10–156) and goats (median = 29.5, range = 25–222) were the predominant domestic mammalian species. While chicken (median = 235, range = 140–430) and ducks (median = 22, range = 7–69) were the most abundant avian species. The median number of cattle and dogs in villages among sampled households were 12.5 (range = 3–31) and 29.5 (range = 12–53) respectively. Cats (median = 2.4, range = 0–4), sheep (median = 0, range = 0–7), rabbits (median = 0, range = 0–4), turkeys (median = 0, range = 0–10), pigeons (median = 4, range = 0–45) and guinea fowl (median = 0, range = 0–6) were rare.
There was also a considerable variation in the number of animals of each species reared per household. Pig rearing households had a median number of two pigs (range = 1–20) and those with goats had a median of 4 goats (range = 1–18). Households with dogs had a median of two dogs (range = 1–10) and those with cats had a median number of one cat (range = 1–3) per household. The highest variation occurred among chicken rearing households which had a median of eight chicken (range = 1–60). Other species were found in very few households. Although an attempt was made to trap rats from all the villages at 34 sites, only 65 rats were trapped in cages from 22 sites across five villages (S1 and S2 Tables).
With the exception of one pig rearing household (which was raising pigs intensively in a concrete floored house), the majority (154 out of 155; 99.4%) confined pigs on earthen floors during the crop growing season and released them to scavenge after harvest. Dogs, cats and chicken roamed freely with no restriction on compounds. Dogs were not housed at all and cats lived inside the human houses. Chicken and other avian species were mainly housed in earthen kitchens or inside the human house in most households (79.8%, 162 out of 203) or in provisional structures or cages on compounds (20.2%, 41out of 203). In all households goats were tethered during the day in bushes and kept at night in the kitchen, house or verandas (48.6%, n = 103); provisional structures (31.6%, n = 67) or in open spaces on pegs on the compound 19.8% (n = 42). Ectoparasite control was practiced (but with no defined regular schedule) in 12.9% (20 out of 155), 10.4% (22 out of 212) and 5% (6 out of 120) of pig, goat and dog owning households, respectively. Owners used ectoparasiticides by either washing or spraying the animals.
All gravid females extracted from infected animals and humans for identification exhibited a clover-like exoskeleton structure of the anterior extremity of the hypertrophic abdomen and hence were characterized as T. penetrans (S1A Fig). Scanning electron microscopy also confirmed the presence of this structure (S1B Fig)
Proportions of households with tungiasis in animals and humans in the study area are provided in Table 1. Animal tungiasis was detected in nine of the ten villages with an overall prevalence of 26.3% (n = 62, 95% CI 21.1–32.2%) out of the 236 households. Nagongera was the only village without any case of animal tungiasis. No animal cases were detected among the 57 additional households from Makoma 1 and Masolya which were selected on the criteria of having at least one goat to achieve a representative sample of goats in these villages. The overall prevalence as well as species-specific prevalence varied widely. The proportions of households with at least one infected animal were also highly variable among villages with a median proportion of 22.2% (range = 0–71.4%). Busindha village had the highest prevalence of tungiasis among households (71.4%, 95% CI 45.4–88.3%).
Pigs, dogs, goats and a single cat were the only infected domestic mammalian species out of the seven examined (Table 2). Only two cases of tungiasis were detected in goats (one in Masolya and one in Busindha village). Goat tungiasis was detected in only one goat rearing household out of 97 (1%, 95% CI 0.2%-5.6%) in Masolya and Makoma 1 combined, where unbiased sampling was undertaken. The only infected cat was found in Matyama village. Rats (n = 65) and all poultry species examined were not infected even in households where other animal species and humans were heavily infected. However, many chickens were infested with the flea Echdinophaga gallinacea. Pig tungiasis occurred in nine of the 10 villages, while in dogs, tungiasis occurred in only five villages. In both species, the prevalence varied considerably (median = 16.2%, range = 0–64.1%; median = 2%, range = 0–26.9%, respectively). Pigs were significantly more affected than other species (pigs vs. dogs, p<0.0001; pigs vs. cats, p = 0.02; pigs vs. goats, p<0.0001. There was no significant difference in prevalence between the dogs and cats (p = 0.54), though dogs were significantly more affected than goats, p<0.0001.
There was no correlation between prevalence of pig tungiasis and the size of the pig population at both household and village levels (rho = 0.09, p = 0.28 and rho = 0.3, p = 0.44 respectively). The same trend was evident for dogs (household level rho = 0.08, p = 0.4; village level rho = 0.05, p = 0.9). Among pig rearing households, there was no significant difference in the proportion of those with infected pigs between those that practiced ectoparasite control for pigs and those that did not (6 out of 20 vs. 48 out of 135, p = 0.42). This was also true for dog owning households (2 out of 6 vs. 12 out of 114, p = 0.145). The two cases of goat tungiasis occurred in households with infected pigs and humans. The only infected cat had a single lesion and was detected in a household with neither human nor other animal species tungiasis. The infected goats were three week old kids while the cat was two years old.
Families had a median size of eight (range 1–24) members with a median of two households (range 1–8) on the same compound. Most households (89%, n = 210) had earthen floored houses which occupants occasionally smeared with cow dung to minimize dust accumulation in the houses. Of the 1766 examined humans from the 236 households, 856 (48.5%) were females while 910 (51.5%) were males. Human tungiasis was detected in all the ten villages (Tables 1 and 3). In 80 (33.9%, 95% CI 28.2–40.2%) of the 236 households (which had at least one dog, cat or a pig), at least one human was affected (Table 3). The prevalence of households with human tungiasis also varied greatly (median 27.8%, range 5–71.4%). In the additional 57 households (selected on the criterion of at least one goat after covering those selected on the criterion of possessing a dog, cat or pig) in Masolya and Makoma, 21 (36.8%) had at least one human case.
Among humans examined for tungiasis in the 236 animal rearing households sampled, 254 (14.4%, 95% CI 12.8–16.1%) were infected but in the 57 additional households, 48 out of 382 humans (12.6%, 95% CI 9.6–16.3%) were infected (p = 0.20). Prevalences of human infections in the villages (Table 3) were significantly variable (median 7%, range 1.3–37.3%; p<0.0001). The prevalence of tungiasis was significantly higher in males (n = 154, 16.9%, 95% CI 14.6–19.5%) than females (n = 100, 11.7%, 95% CI 9.7–14.0%; p = 0.001). Children (0–15 years) had a significantly higher prevalence of tungiasis than other humans above 15 years (20.3% vs. 5.5%; p<0.0001). The prevalence of human tungiasis was highest in children of 6–15 years (Fig 2). Overall, the variations in prevalence of tungiasis among age groups were statistically significant (p<0.0001). There was also a significantly higher prevalence in the two youngest age groups compared with the middle age groups and a clear peak in the age group of 6–15 years (Fig 2).
In 40 (17%, 95% CI 12.7–22.3) of the sampled households, animal and human tungiasis coexisted. These constituted 50% and 64.5% of the households with human and animal tungiasis respectively. Animal tungiasis increased the odds of the occurrence of human tungiasis in households by six times (OR = 6.1, 95% CI 3.3–11.4%; p<0.001) and vice versa. In Busindha, all animal rearing households that had infected animals also had infected humans. In Isakabisolo, the proportions of households with infected animals were similar to that of humans but infections were detected in different households, i.e. there were households where only human or only animal tungiasis was detected. In other villages the proportions differed (Table 1).
Overall the proportions of households with human tungiasis did not differ significantly from those with animal tungiasis in the 10 villages (p = 0.07). A strong correlation existed between the prevalence of households with human tungiasis and those with animal cases within the 10 villages as illustrated in Fig 3A. (rho = 0.85, p = 0.002). At household level, the prevalence of animal tungiasis correlated with human tungiasis prevalence (rho = 0.4, p<0.001) as shown in Fig 3B. Also, at household level, the prevalence of tungiasis in the mostly affected animal species strongly correlated with human prevalence (dogs rho = 0.34, p = 0.0002; pigs rho = 0.5, p<0.001).
Pigs had the highest parasite load followed by goats and dogs: median = 8 lesions (inter-quartile range (IQR) = 3–30; range of 1–246 lesions per pig); goats median = 20 (6 and 34 lesions in 2 goats); dogs median = 2 (inter quartile range = 2–3; range = 1–8). The only affected cat had a single non-viable lesion. Of the 3357 lesions in pigs, 2243 (66.8%) were viable (Fortaleza stage IIa-IIIb) and 1114 (33.3%) were non-viable (Fortaleza stage IV). The number of lesions per pig was highly variable (Fig 4). Among the infected pigs, those which had the highest infection intensity (>30 lesions, n = 30, 24.8%) presented 79% of the total number of embedded sand fleas.
The 20 infected dogs had a total of 53 lesions of which 32 (60.4%) were viable while 21 (39.6%) were non-viable. Occasionally, it was observed that dogs bite at flea lesions and exteriorized the fleas with their teeth. Out of the 20 dogs only 2 (10%) had ≥5 lesions while the other 18 had light infections (1–4 lesions per dog). Of the total 40 lesions found on the two goat kids, 27 (67.5%) were viable while the other 13 (32.5%) were non-viable. In pigs, no correlation was observed between age and the total number of lesions (rho = 0.014, p = 0.88) but in dogs the number of lesions per dog significantly decreased with age as shown in Fig 5. (rho = -0.47, p = 0.039). The number of lesions did not differ between sexes: female median 10 (IQR 3–39) vs. male median 6 (IQR 3–30) in pigs (p = 0.37); female median 2 (IQR 2–3) vs. male median 2 (IQR 2–3) in dogs (p = 0.88).
In the 236 households, sand flea lesions were counted in 111 infected children aged three to eight years, the age group known to have the highest intensity of infection. These included 62 boys and 49 girls. A total of 340 lesions were documented from one randomly selected foot of these children. The median number of lesions per infected child was 2 (range 1–18). The number of lesions per foot never differed significantly between boys (median 2, range 1–18) and girls (median 2, range 1–8; p = 0.42).
Pigs had a significantly higher number of lesions than other species combined (median 8, range 1–246 vs. median 2, range 1–34; p = 0.0002). Accordingly, the number of lesions was also significantly higher for pigs than dogs (p < 0.0001). The median number of lesions in infected animal species strongly correlated with the median number of lesions in children three to eight years of age at household level as illustrated in Fig 6 (rho = 0.47, p<0.0001). The prevalence of human tungiasis at household and village levels correlated strongly with the number of lesions in pigs at the respective levels (rho = 0.5, p<0.0001; rho = 0.8, p = 0.002). At household level, an increase in human tungiasis intensities corresponded with an increase in the odds of occurrence of animal infections (OR = 1.8 CI 1.4–2.2, p<0.0001) and vice versa (OR = 1.3 CI 1.1–1.4, p < 0.0001).
Of the infected 121 pigs, only 18 (15%) had ever received ectoparasite treatment and none of these had a well-defined interval of treatment. The period between the last time of pig treatment and the examination date ranged from one week to 10 weeks. Nine pigs had received ectoparasitic treatment between one and two weeks before the examination date while the rest (9 pigs) had received treatment between five and 10 weeks prior to the examination date. Ectoparasiticides which had been used on infected pigs included pyrethroids (6), amitraz (6) and a traditional concoction of molasses and a local gin (waragi) which was used on one pig. Three pig owners could not recall the type of ectoparasiticide they had used to treat five of the infected pigs.
Although, the total number of sand flea lesions per infected pig was lower in pigs treated with ectoparasiticides than the untreated (treated median 6.5, IQR = 3–13 vs. untreated median 8, IQR = 3–39), the difference was not significant (p = 0.34). For treated pigs, there was no correlation between the number of embedded sand fleas and the time span since when the pigs received the last ectoparasiticide treatment (rho = -0.16, p = 0.53).
Pig dwellings were located at a median distance of 11 meters (range 0–50 m) from the edge of human compounds. The distance of pig dwellings (places of pig confinement) from human compounds had a weak positive correlation with the total number of lesions per pig (rho = 0.18, p = 0.043).
Ectoparasite control had been attempted for only two of the infected dogs (10%) with no definite control interval. While one dog had received the ectoparasiticidal treatment one week before the examination date, the other had received the same treatment four weeks ago and in all cases α-cypermethrin 10% was used. Incidentally, the more recently treated dog had more lesions than the dog treated three weeks before (three lesions vs. one lesion). Neither the two infected goat kids nor the infected cat had ever received any ectoparasiticidal treatment.
A bivariate analysis of risk factors was undertaken (S3, S4 and S5 Tables). Factors strongly associated with occurrence of tungiasis in all animals irrespective of species; pigs and dogs within households are summarized in Table 4. Occurrence of infected animals in households was strongly associated with human infections (OR = 6.0, 95% CI 2.4–9.1; p < 0.0001) and presence of pigs in a household (OR = 5.8, 95% CI 2.5–13.5; p < 0.0001) among other factors. For households with infected humans, the risk of animal tungiasis increased with the number of infected humans. One to four infected humans in households compared to none increased the risk by five times (OR = 5.0, 95% CI 2.6–9.5; p < 0.0001) but presence of 5–12 infected individuals raised the odds to 12 times (OR = 12.2, 95% CI 4.1–35.8; p < 0.0001). Pig tungiasis occurred in strong association with dog tungiasis (OR = 7.4, 95% CI 1.5–36.9; p = 0.02) and human tungiasis (OR = 7.0, 95% CI 3.4–14.7; p<0.0001) while presence of infections in dogs was also strongly influenced by human tungiasis (OR = 12.2, 95% CI 2.6–57.4; p = 0.002) and pig infections (OR = 5.6, 95% CI 1.7–18.2; p = 0.004).
In multivariate analysis, presence of animal tungiasis in households was strongly influenced by presence of human tungiasis (OR = 6.5, p < 0.0001) and presence of pigs (OR = 5.9, p = 0.0002) as illustrated in Fig 7A. In addition, the number of animal species reared in the household (OR = 1.6, p = 0.02) and the size of the homestead (OR = 1.4, p = 0.02) significantly increased the odds detecting animal tungiasis among households. Slightly but non-significant protective effect was observed in association with the presence of chicken and goats in the households. However, the overall model had a poor to moderate fit (McFadden pseudo R2 = 0.25).
Due to the overall low fit of the animal tungiasis model, separate models were estimated for pig and dog tungiasis. The odds of households to have pigs infected with T. penetrans were significantly increased if there was also human tungiasis as illustrated in Fig 7B (OR = 2.1, p = 0.001). The final model showed an excellent fit (McFadden pseudo R2 = 0.82). It also included the variables; “presence of other poultry”, “dog tungiasis” and “pig herd size” which all increased the odds of pig tungiasis although their effects were all not significant. In addition, housing of pigs had a very low OR of 1.6×10−9 suggesting that it has strong protective effect against pig tungiasis. However, this effect was not significant due to the very low number of households with housing for pigs hence a very wide 95% CI.
The same analysis for the presence of dog tungiasis in households (Fig 7C) suffered from the low number of households with dog tungiasis identified in the study (n = 14). Nevertheless, the overall fit of the model was very good (McFadden pseudo R2 = 0.71). The variable with the strongest influence on the odds of households to have dog tungiasis (human tungiasis, OR = 4.0 × 109 had a very wide 95% CI. The number of goats significantly increased the odds (OR = 2.0, p = 0.04) while pig tungiasis had a non-significant influence. Presence of poultry other than chicken appeared to be slightly protective but this effect was also not significant.
Various domestic, peri-domestic and sylvatic mammals have been reported as suitable hosts of zoonotic sand fleas. Domestic and peri-domestic animals which are reportedly more important in the epidemiology of tungiasis than sylvatic reservoirs include cats, dogs [16], rats, mice [15,17], pigs [17,25,26], goats [27], sheep and cattle [3]. The significance of the different animal reservoirs for T. penetrans differs between endemic areas. While cats, dogs and peri-domestic rodents have been reported to be the most important animal hosts for T. penetrans in Northeast Brazil [15,16,25], pigs have been identified as the major reservoirs in a single study in West Africa [17].
Studies looking for an association between animal and human tungiasis are scarce. Previous studies performed in Northeast Brazil [15,16,25] and Nigeria [17] were limited to one or two small communities, covered a few animal species and were less systematically conducted compared to this study. For the first time, it was attempted to identify the risk factors that determine the occurrence of animal tungiasis. Since the study area is typical of many rural settings in Uganda, the results can presumably be extrapolated to other rural communities in the country and probably East African countries.
Pigs were the animal species predominantly affected and they also had the highest intensity of infection. These findings corroborate observations made in Nigeria [17]. Since both, the prevalence of tungiasis in pigs at village and household level as well as the intensity of infection correlated strongly with the respective measurements in humans, it can be proposed that pigs are the reservoir hosts that contribute most to the occurrence of tungiasis in humans in the study area. Dogs, especially puppies, cats and goats had very low prevalences of tungiasis. Hence they are alternative hosts of T. penetrans in the study area. To what extent these species contribute to amplification and propagation of T. penetrans in humans is unknown.
Cattle and other mammalian species with the exception of pigs, dogs, cats and goats were not found to carry T. penetrans. However, since only a few of them were sampled, it cannot be excluded that they also act as reservoirs of T. penetrans in the study area and other areas of Uganda. Reports from South America indicate that cattle seem to be particularly susceptible to T. trimamillata infection and T. penetrans infection is encountered usually as a co-infection [3]. In the study area, the absence of tungiasis among cattle could also be attributable to the higher rate of ectoparasite control in cattle (46% of the cattle owning households) than for other species.
Chicken and other avian species have been suggested as hosts for tungiasis [25]. However, whether chicken and other avian species are actually appropriate T. penetrans hosts remains to be demonstrated. In this study, many chicken examined had E. gallinacea infestations which people mistook for T. penetrans. Since chicken stay close to human dwellings (in the study area all chicken are strayed on the compound and the majority spent nights either in human houses or in small separate kitchen buildings), one would expect that if chicken were susceptible, they would have the highest burden of T. penetrans infections in infected households compared to other susceptible species. It seems that thick feathers on the chicken body and the scaly legs are barriers to sand flea penetration.
The current study did not detect any case of tungiasis among the 65 rats trapped from five villages, a finding which contrasts with reports from Brazil and Nigeria where rodents were identified as important reservoirs of T. penetrans [15,17]. In marked contrast to this observation, 28% (n = 82) of the household heads interviewed claimed to have seen rats with tungiasis either in their houses or homesteads. Since the residents’ abilities to discern tungiasis from other parasitic diseases of rodent feet, such as other flea species or even myiasis, may be low, their assertions have to be considered with great caution and further studies are recommended.
Both cases of tungiasis in three week old goat kids occurred in association with heavy human and pig infections in the respective households. In contrast, adult goats were not infected even in premises with many infected humans and pigs with high intensity of T. penetrans. Although goats were the most abundant domestic mammalian species in the area, they appear to be of no epidemiological significance in the transmission of T. penetrans. A soft hoof wall and the skin around the coronary band of the digits together with the practice of leaving kids to roam freely on compounds coupled with sheltering kids in human or close to human dwellings could explain the infections in kids. In contrast, adults have hard hooves and are tethered in bushes most of the time during the day.
Tungiasis in goats has been reported in a few countries in South America. As is the case with cattle, it was mostly due to T. trimamillata with some reports of co-infections with T. penetrans [3]. A study on animal reservoirs of T. penetrans in rural communities in Ethiopia reported a prevalence of 3.2% in goats in one out of four study communities. In contrast, tungiasis was detected in sheep in all communities with a prevalence of up to 29.5% [28]. The few sheep encountered and examined in this study were not infected. It remains unclear why prevalences greatly differed between the two small ruminant species in Ethiopia despite their anatomical similarities.
All households involved in the study had outside resting places for residents, either under a tree or below an erected temporary shade. Since pigs are kept away from human houses most months of the year, the high prevalence of tungiasis in pigs strongly suggests that transmission of T. penetrans also occurs distant to the compounds. Pigs are mostly reared on earthen floors with minimal environmental sanitation management. For this reason, dirty pig dwellings were identified as a risk factor for pig tungiasis at least in the bivariate analysis. The organic material may favor the off-host development of T. penetrans thus predisposing pigs to heavy infections [20]. This was in fact indicated by the high proportion of viable lesions particularly in pigs as compared to dogs. Of course, the practice of allowing them to roam on human compounds also increases the risk of their infection and also facilitates the shedding of sand flea eggs in human dwellings when pigs are infected.
The prevalence of tungiasis and the infection intensity among pigs reported in this study in some villages were comparable to those reported in Nigeria [17] but were much higher than those in Brazil [25]. Pigs may be very susceptible to sand flea infections because of the highly vascularized coronary band and large bulb at the sole of the hoof with a soft skin cushion [3]. Unlike dogs, which are never confined, pig movements are restricted during the crop growing rainy season and intermittently during the dry season. Hence, minimal movement-related thickening of the bulb epidermis takes place. This may explain the lack of variation in the infection intensities between piglets and adult pigs.
Although the prevalence of tungiasis in dogs was the second highest after that of pigs, dogs were considerably less frequently infected than in previous studies in Nigeria and Brazil [15–17]. Therefore, dogs appear to be less important as animal reservoirs in Uganda than in Nigeria and Brazil. Moreover, the number of dogs in Ugandan rural villages is comparatively small. The significance of dogs in the epidemiology of T. penetrans infections in the Ugandan setting generally decreased with dog age probably because of the increasing thickness of the keratin layer of the foot pads as the dogs mature. Since dogs were not restrained, they moved a lot between households and even between villages. This would incite hyperkeratosis of the foot pads. Additionally, mature dogs tend to bite and exteriorize the embedded sand fleas. Despite the low prevalence of dog tungiasis and decreasing infection intensity in older dogs, dogs might be epidemiologically important because they might distribute flea eggs in a much wider range than pigs. Detection of tungiasis in a single cat underlines previous observations that cats may act as reservoirs of T. penetrans [15]. However, the number of cats is low in most rural communities of Uganda.
The patterns of human and animal infections were similar in the study area among animal rearing households i.e. the prevalence and infection intensity strongly correlated. This, together with their geographical coexistence (Fig 1), suggests an inter-linkage between animal and human T. penetrans infections in Uganda. This however requires validation by systematically sampling humans. Since the focus of the present study was on identification of important reservoir hosts of zoonotic tungiasis, only animal rearing households were included. Also co-existence of infections in both humans and animals may indicate a common source of infection which was not directly examined in this study. However, occurrence of tungiasis in animals increases environmental contamination which in turn may increase the prevalence and intensity of infections in humans and vice-versa resulting in the strong correlation. The sharing of T. penetrans infections between humans and animals is most likely facilitated by the practice of allowing animals to roam freely with minimal confinement.
Households with pigs, cats and dogs had higher odds of having at least one infected animal, a finding corroborating earlier studies which identified these animal species as risk factors for human infections [15–17]. Animals from households with young household heads (15–35 years) were more at risk than those from households with older heads in the bivariate analysis. This relationship may be indirectly related to the presence of children, who are particularly at high risk in such households; lack of adequate knowledge to control sand fleas and probably lack of adequate financial resources to ensure adequate hygiene and environmental sanitation. The strong association of animal infections with large household sizes is probably also indirect and may be due to the fact that such households have many children below 15 years who constitute the most vulnerable group of humans to T. penetrans.
The presence of many animal species in households may contribute to poor environmental sanitation and overall poor animal management particularly regarding ectoparasite control. Poor environmental sanitation confers favorable conditions for off-host sand flea development and propagation [29]. The reasons for the association between T. penetrans and other pig ectoparasites such as lice in this study are not known as the ecological interactions between these parasites have not been studied yet. However, the factors that favor the occurrence of ectoparasites in general such as poor environmental sanitation and lack of ectoparasite control probably contribute to high prevalences of both; sand fleas and other ectoparasites.
In the multi-variate analyses, only a few variables remained statistically significant suggesting that the interaction of many factors predispose to animal on the household level. Nevertheless, multivariate logistic regression also revealed a very strong association of human and animal tungiasis in households and in particular the close connection between human and pig tungiasis. The role of dogs in the epidemiology of T. penetrans in Uganda could not be unequivocally described. On the one hand, the OR associated with dog tungiasis were always very high but on the other hand, effects were not significant due to the low number of infected dogs. Larger data sets, probably involving several transmission seasons, would be required to statistically confirm the role of dogs in the epidemiology of animal and human tungiasis.
The study was cross sectional in nature; hence seasonal patterns could not be demonstrated. Also the study only established the human tungiasis burden in animal rearing households with the most important animal hosts of T. penetrans. The findings therefore may not reflect the situation in the general population. Additionally, it was only possible to demonstrate the association between human and animal tungiasis but causal relationships are difficult to determine. While pigs could be confirmed as epidemiologically important reservoirs, the role of other animal species remained unresolved either due to low number of cases (dogs, cats) or because they were not systematically sampled (e.g. cattle).
This study demonstrated a strong correlation between animal and human tungiasis in animal rearing households, which both occurred with high prevalence in rural endemic villages of Uganda. Pigs were identified as the major hosts of T. penetrans. An effective tungiasis control strategy; therefore, calls for an integrated One Health approach. In addition to treatment of humans and environmental sanitation, ectoparasite control should be encouraged among animal owners to eliminate animal infections. However, there is still need to evaluate the therapeutic and prophylactic effects of commercial pesticides against T. penetrans
|
10.1371/journal.pntd.0004120 | Carbohydrate Recognition Specificity of Trans-sialidase Lectin Domain from Trypanosoma congolense | Fourteen different active Trypanosoma congolense trans-sialidases (TconTS), 11 variants of TconTS1 besides TconTS2, TconTS3 and TconTS4, have been described. Notably, the specific transfer and sialidase activities of these TconTS differ by orders of magnitude. Surprisingly, phylogenetic analysis of the catalytic domains (CD) grouped each of the highly active TconTS together with the less active enzymes. In contrast, when aligning lectin-like domains (LD), the highly active TconTS grouped together, leading to the hypothesis that the LD of TconTS modulates its enzymatic activity. So far, little is known about the function and ligand specificity of these LDs. To explore their carbohydrate-binding potential, glycan array analysis was performed on the LD of TconTS1, TconTS2, TconTS3 and TconTS4. In addition, Saturation Transfer Difference (STD) NMR experiments were done on TconTS2-LD for a more detailed analysis of its lectin activity. Several mannose-containing oligosaccharides, such as mannobiose, mannotriose and higher mannosylated glycans, as well as Gal, GalNAc and LacNAc containing oligosaccharides were confirmed as binding partners of TconTS1-LD and TconTS2-LD. Interestingly, terminal mannose residues are not acceptor substrates for TconTS activity. This indicates a different, yet unknown biological function for TconTS-LD, including specific interactions with oligomannose-containing glycans on glycoproteins and GPI anchors found on the surface of the parasite, including the TconTS itself. Experimental evidence for such a scenario is presented.
| In this study we demonstrated the binding of TconTS lectin domains (TconTS-LD) to high-mannose N-glycans and provide evidence for a biological function for this interaction. TconTS1 and TconTS2 lectin domain bind to galactosyl as well as mannosyl glycans with different affinities as shown by glycan array analysis. Along this line, we have also demonstrated binding of TconTS-LD to high-mannose N-glycans on glycoproteins, which is competitively inhibited by the corresponding free high-mannose N-glycans, underlining the TconTS-LD substrate specificities. TconTS1 dimerisation is mediated by TconTS-LD binding to its N-glycans and enzymatic N-deglycosylation of TconTS leads to an increase of monomers. STD NMR results obtained indicate that oligo-mannosylated glycans bind to TconTS-LD to a different site than lactose. In summary, this is good evidence that the lectin domains of TconTS1 and TconTS2 play relevant roles modulating the biological functions of these and possibly other trans-sialidases. Further detailed analysis on TconTS-LD and its role in enzyme activity will lead to a better understanding and possibly to new strategies against Nagana in livestock.
| The protozoan parasite Trypanosoma congolense is the most prevalent cause of animal African Trypanosomiasis (AAT), also called Nagana in cattle and other livestock, causing death to millions of animals resulting in huge economic losses [1–3]. During the parasite’s life cycle in the mammalian host and the tsetse fly vector T. congolense undergoes different developmental stages utilising various strategies to escape the defence systems of both host and vector. For instance, trypanosomes are unable to synthesise sialic acid (Sia) [4], instead T. congolense, like several other trypanosomatids, expresses an unusual glycosyl-transferase called trans-sialidase (TS) that transfers Sia from host cell glycoconjugates to its own surface structures [5,6].
TS are found in both the African and South American trypanosomes [7–10]. However, their roles in parasite development and pathogenesis appear to be species dependent, as the relevance of TS or sialidase activities has been shown for nagana caused by T. congolense [9], but not for sleeping sickness caused by T. brucei ssp. The TS from Trypanosoma cruzi (TcTS), the causative agent of Chagas' disease in humans [11], is the best characterised [12–18] with the mechanism of sialic acid transfer and catalytic activity being described in detail. It has been suggested that the catalytic domain (CD) of TconTS is located at the N-terminus and folds into a β-propeller structure, similar to that of known bacterial and viral sialidases [19–21]. The CD of trypanosomal TS is presumed to be linked via a well-conserved, relatively long α-helix (22 to 25 amino acids) to a C-terminal domain, whose function has remained unclear. The crystal structure of TcTS [14] revealed that the C-terminal domain folds into a β-barrel topology similar to that of known plant lectins such as GS4 (Griffonia simplicifolia lectin 4) [22], GNA (Galantus nivalis agglutinin, Snowdrop lectin) [23], LOL (Lathyrus ochrus lectin) [24] and WGA (Wheat germ agglutinin) [25]. This structural similarity suggests that the C-terminal domain may be a potential carbohydrate-binding site or “lectin-like” domain (LD). In contrast to TcTS, only a few studies have investigated the enzymatic activities of T. brucei TS (TbTS) [26–29] and TconTS [6,30,31]. However, given the overall high primary sequence similarity of all TS at the catalytic site it can be assumed that the molecular mechanisms of the African TS are similar to those described for TcTS.
In addition to several other TS-like genes, T. congolense possesses eleven TS1 (TconTS1) genes encoding variants with 96.3% overall amino acid identity [30]. Recombinant TconTS1 variants are able to desialylate fetuin in the presence of lactose to generate α2,3-sialyllactose (3‘SL), demonstrating not only their TS activity but also their sialidase activity, reflecting their ability to hydrolyse terminal sialic acids [30]. Three additional T. congolense TS family members, TconTS2, TconTS3 and TconTS4, have recently been described, sharing more then 40% amino acid identity [31]. Furthermore, kinetic data show that all TconTS investigated so far have different affinities for glycoprotein and oligosaccharide substrates [30,31].
Previous research on trypanosomal TS has been focussed on investigating the TS CD with respect to substrate specificities, mechanisms of sialic acid transfer and sialidase activities [6,12,14,27,30–34]. However, apart from sequence and structural data, limited information regarding the LD of the trypanosomal TS is available, with the actual function of the LD in TS remaining unknown. Due to structural similarities with known lectins, it had been proposed [13,35] that TS LD binds carbohydrates and may play a role in mediating cell adhesion. Recently, Ammar et al. suggested that TconTS lectin like domain binds sialic acids and is involved in endothelial cell activation [36]. However, to the best of our knowledge no direct evidence for carbohydrate-binding specificities of TS LD has been described.
Interestingly, a detailed phylogenetic analysis comparing TS domains revealed that the TconTS-CDs grouped the highly active TconTS together with the less active enzymes. In contrast, when aligning TconTS-LDs, the highly active TconTS grouped together [31], indicating a potential role of TconTS-LD in modulating enzyme activities.
Here we report on the biochemical characterisation of four recombinant TconTS-LD (TconTS1-4) employing glycan array, STD NMR and binding/inhibition assays that identified TconTS-LD as a carbohydrate recognition domain (CRD), with several oligosaccharides identified as TconTS-LD binding partners. Interestingly, prominent ligands involved were oligosaccharides with terminal mannose, which are not substrates for trypanosomal TS activity, suggesting previously not reported biological functions. In addition, we provide strong evidence for a second binding site for oligosaccharides with terminal galactose moieties.
To characterise the LDs of TconTS the gene sequences encoding the TconTS1-LD, TconTS2-LD, TconTS3-LD and TconTS4-LD were subcloned into a modified pET28a bacterial expression vector as described under Methods. All proteins comprise a N-terminal poly histidine tag (His-tag) directly attached to maltose binding protein (MBP) as well as C-terminal SNAP- and Strep-tags. In these proteins, the tags flanking TconTS-LD can be enzymatically cleaved using the tobacco etch virus (TEV) and human rhinovirus 3C (HRV 3C) proteases (Fig 1A).
Recombinant protein comprising only His-MBP-SNAP-Strep, but no TconTS-LD was used in control experiments. For TconTS-LD interaction studies, based on homology models (Fig 1B and 1C) two sets of constructs were generated with or without the α-helix connecting CD to the LD (Table 1), to investigate its potential influence on binding activity. Expression conditions were optimised for efficient production of soluble TconTS-LD in amounts ranging from 0.5–2 mg/L bacterial culture as described under Methods. After tandem affinity chromatography employing Ni-NTA and Strep-tag consecutively, protein purity was confirmed by gel electrophoresis and Western blot analysis (Fig 1D). All TconTS-LD were obtained as pure proteins clearly showing a migration shift due to the presence or absence of the N-terminal α-helix.
Glycan array analysis was performed to identify potential TconTS-LD oligosaccharides binding partners. Recombinant TconTS-LD containing His and MBP fusion tags (Table 1 and Fig 1A) were pre-complexed with anti His mouse polyclonal antibody, anti-mouse-IgG-TexasRed conjugated rabbit polyclonal antibody and anti-rabbit-IgG-TexasRed conjugated donkey polyclonal antibody. These were then applied to glycan arrays printed onto SuperEpoxy2 glass slides comprising 367 diverse biologically relevant glycan structures (S2 Fig). The major subset of glycans bound by TconTS-LD are summarised in Fig 2 (full binding data provided in S2 Fig and S1 Table). As expected, initial glycan array experiments revealed signals associated with maltose, maltotriose, isomaltotriose, maltotretraose, isomaltotetraose and related glycans due to the binding of MBP (S1A Fig). Therefore, 10 mM maltose was added as a competitor during binding and washing steps to inhibit the MBP interaction with maltose and related structures present on the arrays. Under these conditions the majority of maltose related signals disappeared. Only some signals for maltotriose, maltotetraose and other maltodextrins remained. Given that maltotriose has a more than 6-fold higher affinity for MBP (Kd: 0.16 μM) compared to maltose (Kd: 1 μM) [37], 10 mM maltotriose instead of maltose was used during binding and 1 mM in all wash steps. Under these conditions, binding of MBP to all remaining maltose related structures was successfully inhibited (S1B Fig). Another option that could have been used to prevent MBP associated binding to our glycan array would have been a proteolytic cleavage using the TEV protease cleavage site of the recombinant TconTS-LD protein (Fig 1A). However, the removal of the MBP-tag and subsequent purification of TconTS-LD leads to low yield of pure TconTS-LD, since often the protease digest is not complete. Therefore, we choose to inhibit MBP binding to maltose-related structures on the glycan arrays with maltotriose in the analyses of all eight TconTS-LD constructs. Glycan array analysis of TconTS2-αHel-LD and TconTS2-LD showed clear binding to several different galactobiose and lactose containing oligosaccharides, as well as to some of their N-actetylamine derivatives listed in Fig 2. Also several fucosylated, and two sialylated glycans were bound, although the binding to these structures was less pronounced compared to unsubstituted N-acetyllactosamine. Whereas binding to potential TS substrates containing galactose was not unexpected, surprisingly, we also observed binding to α1-6-mannobiose and α1–3,α1-6-mannotriose, which was similar for TconTS2-LD with and without the α-helix. No obvious preference of TconTS2-LD for any of the oligomannose isomers present on the array was identified. The number of glycan structures bound by TconTS1-LD was lower than that observed for TconTS2-LD, and no binding to any glycan structures was observed for either TconTS3-LD or TconTS4-LD under the conditions used.
TconTS2-LD showed the highest lectin activity on glycan arrays. Therefore, in further experiments we focused on TconTS2-LD to more fully characterise and confirm the binding of TconTS-LD to both galactose and mannose containing oligosaccharides observed on the glycan array. Several NMR-based methods have been employed to investigate protein carbohydrate interactions on a structural level. For example, line broadening and peak shifts of 1H-NMR signals from amino acid side chains provide information on the type of amino acids involved as well as the occupation of the binding site and thus equilibrium kinetic data, as has been shown for Siglec–1 [38]. Saturation transfer difference (STD) NMR experiments provide important information on the binding epitope of the complexed carbohydrate ligand, since the relative signal intensities of the difference spectra provide direct information on the proximity of the affected protons to the protein [39]. Protein signals are selectively saturated at -1.00 ppm (on-resonance) and subtracted from an off-resonance spectrum (30 ppm) resulting in the final STD NMR spectrum revealing only protons and functional groups of a binding ligand that are in close proximity to the protein surface. Therefore, STD NMR has been widely used to analyse the binding of lectins to their specific carbohydrate ligands. Lactose and α1–3,α1-6-mannotriose were used as ligands for TconTS2-LD as described under Methods. Fig 3A shows the 1H NMR (off resonance) and STD NMR spectra of α1–3,α1-6-mannotriose. The relative signal intensities of the STD spectrum (red line) are almost identical to those of the oligosaccharide 1H NMR spectrum (black line). Binding of lactose to TconTS2-LD was also clearly observed (Fig 3B). It is important to note that relatively strong STD NMR signals at 3.36 ppm (β-GlcH2) and at 3.92 ppm (GalH4) provide good evidence that both monosaccharide units of lactose are in close contact with the protein.
Taken together, the STD NMR data confirmed binding of both lactose and α1–3,α1-6-mannotriose to TconTS2-LD, which were initially identified by glycan array analysis, and raises the question as to whether both oligosaccharides bind to the same or distinct sites on TconTS2-LD. To try and address this question, an additional STD NMR competition experiment was performed, where an equal quantity of lactose was added to the TconTS2-LD/α1–3,α1-6-mannotriose complex. If lactose was able to bind to the same site as α1–3,α1-6-mannotriose, the two oligosaccharide ligands would compete, which then would lead to a reduction in the STD NMR signals for one or both ligands, depending on their relative affinities for this site [40]. However, no such reduction was observed for either lactose or α1–3,α1-6-mannotriose (Fig 3C and 3D) suggesting that both ligands are likely to bind simultaneously to different binding sites on TconTS2-LD.
We established microtitre plate based binding and inhibition assays to further characterise TconTS-LD binding affinity and specificity. Our glycan array and STD NMR experiments revealed TconTS-LD binding to oligo-mannose oligosaccharides. To further investigate how this specificity mediates interactions of TconTS-LD with glycoproteins, recombinant human Siglec 2 (huS2-Fc, described under Methods) expressed in Chinese hamster ovary Lec1 (CHO-Lec1) cells was used as a model glycoprotein. Due to the lack of N-acetylglucosaminyltransferase 1 (GnT1) CHO-Lec1 cells are unable to synthesise complex and hybrid N-glycan structures. Therefore, these proteins contain only high-mannose glycans of the type Man5GlcNAc2-Asn [41]. Purified huS2-Fc immobilised in microtitre plate wells was incubated with different concentrations of TconTS2-LD and binding was detected as described under Methods. TconTS1-catalytic domain (TconTS1-CD) was used as a control for binding specificity. As shown in Fig 4A, concentration-dependent binding of TconTS2-LD to immobilised huS2-Fc was clearly observed, reaching a maximum intensity at approximately 2 μg/mL TconTS2LD due to saturation of the binding sites. Wells without immobilised huS2-Fc were used as a control. No detectable binding to immobilised huS2-Fc was observed for TconTS1-CD at 4 μg/mL (Fig 4A), confirming the specificity of this assay.
To investigate whether the binding of TconTS2-LD to huS2-Fc was mediated by its high-mannose N-glycans, huS2-Fc was treated with Endoglycosidase H (EndoHf), a recombinant glycosidase, which specifically cleaves high-mannose and some hybrid oligosaccharides from N-linked glycoproteins [42]. Released N-glycans were then isolated and used in a 1:2 serial dilution as potential competitive inhibitors of TconTS2-LD binding. Fig 4B shows the concentration-dependent inhibition of TconTS2-LD by the EndoHf released N-glycans. Importantly the undiluted purified N-glycans inhibited binding completely, demonstrating that the interaction of TconTS2-LD with huS2-Fc is exclusively mediated by binding to N-glycans.
The enzymatic activities of TconTSs were previously characterised [30,31] using recombinant proteins expressed in CHO-Lec1 cells that therefore contain N-glycans of the high-mannose-type, similar to the recombinant huS2-Fc used here in binding/inhibition assays. N-glycosylation site prediction analysis revealed 8–9 potential sites in TconTSs, and none for the attached SNAP-tag (S3 Fig). In view of the interaction of TconTS-LD with huS2-Fc, we addressed the question of whether TconTS could oligomerise through binding of N-glycans. First, the presence of mannosylated glycans on TconTS1 and TconTS2 expressed in CHO-Lec1 cells was confirmed by lectin blot analysis using concanavalin A (ConA) (Fig 5). To analyse TconTS oligomerisation we used gel permeation chromatography of TconTS expressed in fibroblasts. These proteins contain CD and LD followed by SNAP- and Strep-tags, but no His-MBP at the N-terminus. Fig 6A shows the chromatograms of recombinant TconTS1 and TconTS2 under identical conditions. A double peak of similar intensities was observed for TconTS1, whereas TconTS2 showed a clear single peak with a small shoulder in front of it. The molecular weight (MW) of the TS1 peak eluting at 13 mL (peak 2) was 293 kDa and for that at 16.9 mL (peak 3) was 119 kDa (Fig 6A). This is consistent with peak 3 representing TconTS1 monomers and peak 2 dimers. Furthermore, a molecular mass of 603 kDa calculated for peak 1 eluting at 10 ml is consistent with tetramers of TconTS1. The deviation from the expected monomer (101 kDa without glycosylation) from the calculated MW (119 kDa) can be explained by increased hydrodynamic volumes (Stokes radii) of the oligomeric TconTS1, which is well-known to influence the elution behaviour of a molecule in size-exclusion chromatography [43]. Additionally, glycosylation also influences the protein elution behaviour. For the prominent peak of TconTS2 in Fig 6A eluting at 17 mL (peak 2) a MW of about 113 kDa was calculated, which is consistent with the TconTS2 monomer (100 kDa without glycosylation). The small shoulder at 13.3 mL (peak 1) that represents a MW of about 260 kDa is consistent with the dimeric form of TconTS2 (200 kDa without glycosylation). These findings strongly suggest that both TconTS1 and TconTS2 exist as monomers as well as oligomers in solution, however at different ratios. TconTS1 showed an approximate 1:1 ratio of monomer to dimer, whereas TconTS2 mainly migrates as monomer under the conditions used. To address whether oligomerisation is mediated by N-linked glycans, TconTS1 was enzymatically deglycosylated using EndoHf, and the resulting oligomeric state assessed by size-exclusion chromatography. As shown in Fig 5a clear molecular weight shift as well as a reduction in signal intensity for ConA binding was observed, strongly indicating the release of mannose containing glycans from TconTS1. Subsequent gel permeation chromatography of the deglycosylated TconTS1 resulted in a changed elution profile (Fig 6B, dashed line) compared to the untreated protein (Fig 6B, solid line). Calculating the molecular weight, peak 2 of the deglycosylated protein in Fig 6B is determined as the dimeric form with 260 kDa (elution volume: 13.3 mL) and peak 3 as the monomer with 109 kDa (elution volume: 17.2 mL). The small differences in MW can again be explained by the reduced glycosylation effect on the Stokes radius after EndoHf treatment. Comparing the MW of monomer and dimer from the EndoHf treated sample (peak 2: 260 kDa; 3: 109 kDa, dashed line) to those from the untreated (2: 293 kDa; 3: 119 kDa, solid line), it can be seen that both are decreased due to the loss of high-mannose N-glycans released by EndoHf. Interestingly, it was also observed that EndoHf treatment of TconTS1 reduced the abundance of dimers indicated by the smaller peak 1 (dashed line) in chromatogram B compared to untreated TconTS1 (Fig 6B, solid line). In summary, these results provide strong evidence for oligomerisation of TconTS1 by binding to its N-glycans. Similar, but less pronounced is the oligomerisation of TconTS2.
Lectin domains (LD) of the four TconTS1-4 were expressed and characterised with respect to their ability and specificity to bind carbohydrate structures. Structural comparison to other known bacterial (Salmonella typhimurium LT2) [44] and viral (Vibrio cholerae neuraminidase) [45] sialidases, as well as to plant lectins (Griffonia simplicifolia lectin 4, GS4; Lathyrus ochrus lectin, LOL) [22,46] provided structural evidence for potential carbohydrate-binding of TconTS-LD. Besides typical structural elements seen for several lectins, such as the β-barrel topology, each TconTS-LD also comprises a cluster of histidine, phenylalanine and arginine residues in its potential binding site, a rather shallow indentations (Fig 1C). These could presumably be involved in carbohydrate recognition via aromatic side-chain and sugar ring interaction as well as hydrogen bonding, as described for other lectins [47]. Furthermore, it is noticeable that this potential TconTS-LD carbohydrate-binding site is oriented in the same direction as the TconTS-CD catalytic site, similar to T. cruzi TS, T. rangeli SA and leech IT-sialidase [14,35,48]. This structural organisation of TconTS-CD and LD appears to be stabilised by a relatively extended close contact site between both domains comprising a network of hydrogen bonds and complementary hydrophobic patches (Fig 7).
We have identified TconTS-LD as a carbohydrate-binding domain. Employing glycan arrays, TconTS2-LD showed binding to a variety of oligosaccharides (Fig 2), whereas only a few of the glycans presented on the arrays were bound by TconTS1-LD, and none by TconTS3-LD or TconTS4-LD. It remains unclear, why only TconTS1-LD containing the α-helix at its N-terminus showed binding activity in the glycan array experiments. One explanation could be that folding or the accessibility of the binding site is compromised for TconTS1-LD without the α-helix serving as a spacer to the N-terminal MBP tag. It appears unlikely, that the α-helix is directly involved in binding activity, since the binding pattern and signal intensities for TconTS2-LD were not influenced by the presence of the α-helix. Furthermore, our binding assay data (Fig 3A) have provided evidence that TconTS1-LD also binds high-mannose glycans, which were not available on the glycan arrays, besides 1N and 4D. In addition, EndoH treatment reduced oligomerisation of TconTS1. Finally, it should also be pointed out that most of the differences in amino acid sequence among the 11 TconTS1 gene variants occur in the LD and are clustered close to the postulated binding site [30]. Therefore, it may well be possible that the other TconTS1-LD variants, besides the TconTS1a-LD used in this study, have different carbohydrate-binding specificities. Interestingly, variants that differ only in this cluster (TconTS1b and TconTS1e) have different kinetic properties for the trans-sialylation reaction, which further supports the hypothesis that TconTS-LD modulates also the enzymatic activities. This may also explain the low enzymatic activity of TconTS3 and TconTS4, since no carbohydrate-binding activity was observed for TconTS3-LD and TconTS4-LD. However, as discussed above for TconTS1-LD, TconTS3-LD and TconTS4-LD may bind to carbohydrate structures not present on the glycan arrays. In this context it should be kept in mind that the amino acid sequence diversity of TconTS-LDs (S2 Table) is high enough to allow different binding activities. Furthermore, the structures of spacers used to immobilise the glycans may have prevented binding of TconTS3-LD and TconTS4-LD, since these affect lectin binding to glycan arrays [49,50]. However, given the diverse spacer structures and lengths utilised here, it appears less likely that binding would have escaped detection. Finally, we cannot exclude the possibility that insufficient folding of TconTS-LDs in the bacteria during expression is the reason for undetected lectin activity.
According to a phylogenetic analysis comparing separately TconTS-LDs and TconTS-CDs [31] TconTS1-LD and TconTS2-LD are more closely related with each other than TconTS3-LD and TconTS4-LD. However, when comparing CDs, TconTS2-CD is more closely related to TconTS3-CD and TconTS4-CD, but least to TconTS1-CD. Interestingly, the highest enzyme activities of TconTS were found with TconTS1 and TconTS2, whereas TconTS3 and TconTS4 were 100- or 1000-fold less active [31]. Together with the carbohydrate-binding activities of TconTS1-LD and TconTS2-LD described in this study, it can be postulated that LDs may directly influence enzyme activities. In agreement with this hypothesis is the observation that disrupting a salt bridge between R642 and E648 in the LD of T. cruzi TS enhanced trans-sialidase relative to sialidase activity [35,51]. Along this line it could be postulated that binding of oligosaccharide substrates to TconTS via oligomannose clusters may lead to an improved presentation of the terminal acceptor galactose residues towards the active centre of the CD and therefore lead to an enhanced catalytic efficiency.
Ammar et al. recently demonstrated that TconTS can induce activation of endothelial cells [36] and assumed that the TconTS-LD may be involved in this process. In this context, they prepared a mutant (Y438H) of TconTS1 (TcoTS-A1 according to their nomenclature) to prevent transfer activity. Since this mutant competed with Mal (Maackia amurensis lectin) binding to α2,3-linked Sia of the surface of endothelial cells, they concluded that lectin binding to cell surface carbohydrates play the key role in endothelial cell activation. Based on our data regarding the carbohydrate-binding specificities of TconTS-LD, it appears unlikely that the LD mediates the activation of endothelial cells. However, their findings are in agreement with an involvement of the CD.
TconTS2-LD binds structures containing mannose, such as α1-6-mannobiose and α1–3,α1-6-mannotriose (Fig 2) and several galactose and lactose containing oligosaccharides. Given that the few fucosylated and sialylated glycans recognised by TconTS2-LD all possess core galactose, lactose or N-acetyl-lactoseamine units, it can be assumed that fucosylation and/or sialylation at least in some positions does not interfere with binding. For example, when fucose is linked to positions of the glycan, which is more solvent exposed, it will not disturb ligand binding.
Binding of TconTS-LD to galactosyl, lactosyl and potentially also sialyl glycans was not completely unexpected, since they also serve as substrates for TconTS [6,30,31] and lectin-like binding to sialidase substrates is not uncommon. For example, the lectin-like domain of Vibrio cholerae sialidase (VCS) binds N-acetylneuraminic acid in a similar manner compared to the catalytic domain but without hydrolytic activity [52].
Engstler et al. 1995 investigated procyclic TconTS substrate specificity using a selection of sialylated donor substrates and galactosylated acceptor substrate oligosaccharides, including the monosaccharide mannose, as substrates for TS [5]. Whereas transfer of Sia to terminal galactose oligosaccharides and even the monosaccharide galactose was observed, mannose did not appear to be a suitable acceptor for sialic acid transfer. Thus, our discovery that TconTS-LD binds to mannosylated oligosaccharides suggests a yet unknown function of the LD, distinct from that of the CD. The overall ligand specificities of TconTS-LD, binding to both mannosylated as well as galactosylated glycans, but not to glucose containing oligosaccharides (Figs 2 and S2), differs from that of other mannose-specific lectins, such as concanavalin A (ConA, Canavalia ensiformis) [53,54], LOL (Lathyrus ochrus lectin) [24] or GNA (Galantus nivalis agglutinin, Snowdrop lectin) [23]. In contrast to mannose and glucose, which both show equatorial orientation of the C4-OH group, in Gal the orientation of C4-OH is axial, which does not support binding to ConA or LOL, since C4-OH is involved in carbohydrate recognition by these lectins. Similarly, GNA specifically binds Man via an essential hydrogen bond to its C4-OH group. Considering these ligand selection mechanisms, it appears more likely that two structurally independent binding sites provide TconTS2-LD binding to both mannose- and galactose-containing oligosaccharides. This hypothesis is supported by our STD NMR data that clearly indicate that lactose and α1–3,α1-6-mannotriose do not compete for the same site on TconTS2-LD. According to published STD NMR data for lactose [55] and α1–3,α1-6-mannotriose [56], it could be concluded that both moieties of lactose interact with TconTS2-LD, at least partially. For example, this is indicated by the βGlc-H2 and βGal-H4 protons of lactose (Fig 3B). All signals observed in the off-resonance spectrum of α1–3,α1-6-mannotriose could also be identified in the STD NMR spectra, showing no clear preference for any proton. This indicates specific binding and that all three mannose units appear to be in close contact with the protein. A similar binding epitope of α1–3,α1-6-mannotriose has recently been described for the antibiotic Pradimicin S [56]. The binding epitope for lactose is less uniform, suggesting that not all protons of the disaccharide are in the same close contact with the protein. The exception constitutes the βGlc-H2-proton, for which a two-fold higher STD signal intensity was observed compared to several other protons with similar STD effects than most of the protons of α1–3,α1-6-mannotriose (Fig 3D). This finding might be explained by selective interaction of the βGlc configuration of lactose with TconTS2-LD. An important result was that the STD NMR effects of both oligosaccharides were independent of the presence of the other ligand, since they were identical, if TconTS2-LD was incubated with an equimolar mixture, to those obtained for the individual compounds. If they did compete for the same site, reduced STD NMR signals would have occurred for either both ligands, if they bind with similar affinities, or at least for that ligand, which binds with much lower affinity, if they bind with different affinities [40]. It can be excluded that the STD NMR signals of lactose or α1–3,α1-6-mannotriose reflect interactions with the MBP tag, since no binding of MBP to these structures have been observed in our glycan array experiments, which is in agreement with previous studies on MBP specificity applying diverse methods [57,58] including STD NMR [59]. In conclusion, our findings suggest that two distinct binding sites exist on TconTS2-LD, similar to the lectins WGA [60] and GNA [23].
Interestingly, in a previous STD NMR study TcTS binding to lactose was only observed in the presence of Neu5Ac [16]. Apparently, TcTS and TconTS2-LD have different carbohydrate-binding activities, since TconTS2-LD clearly binds to lactose and α1–3,α1-6-mannotriose in the absence of Sia. Furthermore, the binding epitope for lactose in complex with TconTS2-LD is distinct from that observed for TcTS [16], most pronounced is this difference for the STD NMR signal of βGlc-H2, which was not observed for TcTS. This further underlines that the binding site for lactose on TconTS2-LD is distinct from the acceptor-binding site of the CD described for TcTS.
The crystal structure of TcTS [14] revealed that the binding pocket of the catalytic domain is located at the same side as carved β-barrel groove of the lectin domain, in which conserved histidine and tyrosine residues were identified, known to be involved in carbohydrate recognition of other lectins [47]. Therefore, we assume this position to be the potential carbohydrate-binding site on TconTS-LDs. Furthermore this hypothesis is in agreement with our data assuming TconTS-LD interacts simultaneously with more then just one monosaccharide, suggesting an extended binding site, also termed sub-site multivalency [61]. However, detailed structural studies, such as X-ray crystallography of TconTS2-LD with these oligosaccharide ligands are required to locate and investigate these binding sites precisely.
For several lectins it has been reported that oligomerisation enhances binding due to interactions of multivalent glycoconjugate ligands to multiple binding sites of oligomeric lectins [47,62,63]. Also for TconTS1-LD and TconTS2-LD multivalent interactions in larger complexes strengthen binding to the target glycoprotein, since in our binding/inhibition assays pre-complexing of TconTS-LD with the anti His-tag mAb and the corresponding secondary antibodies used for detection lead to stronger signals compared to applying every component in consecutive steps (S4A and S4B Fig).
By large, this binding is mediated by the high-mannose N-glycans of huS2-Fc, since it could be inhibited with N-linked oligosaccharides, enzymatically released from recombinant huS2-Fc, as competitive inhibitors. Interestingly, for the monosaccharide α-methyl-mannopyranoside, which is a well-known inhibitor for ConA [64], only slight inhibition of TconTS-LD binding could be observed at 50 mM (S4C Fig). This evidence for binding towards complex oligosaccharide ligands was already reported earlier for several other lectins, including the lectin GS4 (Griffonia simplicifolia lectin IV), which shows high affinity to poly- but not to monosaccharides [47,65]. This mechanism, providing poor affinities for lectins to monosaccharides, prevents unspecific interference from competing, structurally similar molecules and enhances ligand selectivity. In addition, recombinant TconTS enzymes expressed in CHO-Lec1 cell lines contain these N-linked high-mannose-type oligosaccharides. Eight to nine potential N-linked glycosylation sites are found in recombinant TconTS, distributed over CD and LD. As expected, lectin blot analysis using ConA, for detection of mannose oligosaccharides, clearly confirmed the presence of N-linked mannosylated glycans on recombinant TconTS expressed in CHO-Lec1 cells (Fig 5). Recombinant TconTS exhibit much higher enzyme activity, if expressed in these fibroblasts compared to those expressed in bacteria (S4D Fig). This is possibly related to the absence of N-glycosylation in bacteria, which may have an indirect effect on enzyme activity by influencing proper enzyme folding or even directly by glycan mediated TconTS oligomerisation. In agreement with the latter scenario, binding to N-linked mannosyl oligosaccharides of recombinant TconTS expressed in CHO-Lec1 cells leads to oligomerisation. This conclusion is supported by (1) the observation that TconTS1 elutes in about equal amounts as monomer and as dimer in size exclusion chromatography (Fig 6A) and (2) that the removal of N-linked glycans with EndoHf glycosidase leads to a clear shift from the dimer towards the monomeric form of TconTS1 (Fig 6B). The remaining dimers are likely to be due to inefficient deglycosylation, as suggested by ConA lectin blot analysis (Fig 5) or due to other carbohydrate-independent protein-protein interactions. It remains unclear as to why TconTS1 oligomerises to a larger extent than TconTS2. One possible explanation may be differences in the glycosylation pattern of the two recombinant enzymes. That is, potential N-glycosylation sites are distributed differently in both enzymes with nine potential sites for TconTS1, eight for TconTS2, only one site being conserved. Finally, it should be mentioned that it is also unclear, which of the predicted N-glycosylation sites of recombinant TconTS utilised in this work are actually glycosylated. In addition, it should be noted that the glycosylation pattern of the native TconTS is unknown as well.
Based on the finding that TconTS-LD binds to high-mannose-type N-glycans we assumed that glycoconjugates containing high mannosylated structures might be preferred natural acceptor substrates for TconTS. Interestingly, such high-mannose-type glycans have been identified on the surface of T. congolense in both, bloodstream (mammalian host) and procyclic (tsetse vector) forms, either linked to amino acids or as part of the GPI anchors [66–69]. The African parasites express two major stage specific, glycosylphosphatidylinositol (GPI)-anchored glycoproteins on their surfaces, the variant surface glycoprotein (VSG) of the bloodstream form and procyclins of the procyclic form. During development of the bloodstream form (BSF) in the mammalian host, the parasites express a surface coat composed of hundreds of immunologically distinct VSG molecules (antigenic variation) to evade host immune response. These VSGs share relatively little primary sequence homology [70] but are structurally related to each other [71]. It has been demonstrated that VSGs from T. congolense and T. brucei BSF are highly glycosylated, exhibiting glycan structures similar to those of higher eukaryotic N-linked oligosaccharides. Interestingly, they are composed of N-linked high-mannose-type oligosaccharides (Man5-9) and N-acetyllactosamine oligosaccharides, as well as branched poly-N-acetyllactosamine oligosaccharides with a Man3-4 core (GalGlcNAc)3Man3GlcNAc, which were also found to be sialylated in T. congolense [72–74]. The fact that terminal Sia were found on VSGs indicate that these are substrates for TS present on the cell surface. In this context, it is plausible, that TconTS-LD contributes to the binding of T. congolense VSG to TconTS via oligomannose oligosaccharides present on these glycoproteins.
When parasites are taken up by tsetse fly through a blood meal, VSGs are replaced by procyclic stage specific, membrane bound, major surface proteins known as procyclins or procyclic acidic repetitive proteins (PARP) [75,76] in T. brucei and glutamic acid/alanine-rich protein (GARP) in T. congolense [66–68,77]. Interestingly, compared to the highly N-glycosylated T. brucei BSF VSGs, procyclins only contain a single N-glycosylation site, substituted with oligomannose oligosaccharide Man5GlcNAc2, which is unusual and rare, but not unique [78]. The primary sequence of T. congolense procyclic GARP does not contain a single potential N-glycosylation site, which was also experimentally confirmed, as well as the absence of conventional O-linked glycans [77]. However, two large Man and Gal-rich oligosaccharides of the type Man11Gal6-7 linked via phosphodiester bonds to two threonine residues were found [77]. Considering these findings, GARP may also be a potential binding partner for TconTS-LD. Indeed, TconTS-mediated sialylation of GARP has been demonstrated and even procyclin was equally efficiently sialylated by the same enzyme, indicating their functional similarities at least as substrates for TS [5]. However, sialylation of procyclin occurs at the glycan moiety of its GPI anchor [26], which has been structurally characterised [79–81]. They share the common core structure of GPI anchor EtNH2-HPO4-6Man(α1–2)Man(α1–6)Man(α1–4)GlcNH2(α1–6)-PI, but with an additional glycosylation at the Man3-core, which is unique for African trypanosomes [80,81]. It comprises oligolactosamine oligosaccharides (Gal-GlcNAc)9 for T. brucei procyclin GPI anchors and oligogalactosyl oligosaccharides Gal5-7 for T. congolense, which both represent substrates for trans-sialylation [77,80]. In this context it is important to note that GARP was co-purified and co-immunoprecipitated with TS-form 1 from procyclic cultures of T. congolense [6], indicating a relatively strong interaction between these two surface proteins. It is in complete agreement with the data of this study that this interaction is mediated by TconTS-LD binding to the GPI anchor Man3-core of GARP, since binding to similar oligomannose oligosaccharides like Man(α1–3)Man and Man(α1–6)Man was observed by glycan array and STD NMR analysis. Homology models of TconTS1-4 revealed, that the distance between the catalytic tyrosine residue in the active centre of TconTS-CD and a conserved phenylalanine residue in the predicted TconTS-LD carbohydrate-binding site, ranges from 40.5 to 42.6 Å. With an average diameter of 7 Å for a single hexose unit, an oligosaccharide of minimum 6 monosaccharide units from the Man3-core is needed to reach the catalytic centre of TconTS-CD, depending on the glycosidic linkage of the oligosaccharide. Consistent with this, the oligosaccharide Gal5-7 of the GARP GPI-anchor has the appropriate size to reach the catalytic centre when TconTS-LD is bound to the Man3-core. In this case both TconTS domains could interact simultaneously with different sections of the same oligosaccharide, whereby the binding affinities of each domain contribute to the overall TconTS binding affinity, which is then clearly enhanced, consistent with the observed co-purification of TS-form 1 with GARP from T. congolense cultures [6]. This would be somewhat similar to the situation reported for the Vibrio cholerae sialidase, where Neu5Ac binds to a lectin domain without hydrolytic activity, leading to an increased affinity of the enzyme for highly sialylated regions [52]. In addition, the apparent organisation of the two native TS-forms isolated by Tiralongo et al. [6], who observed TS-form 2 as dimers and TS-form 1 as oligomers is in very good agreement with the oligomannosyl oligosaccharide-mediated interaction of recombinant, high-mannosylated TconTS1 and TconTS2. It should be noted that these are not identical to purified TS-forms 1 and 2 described by Tiralongo et al. [6]. Furthermore, T. brucei TS has been previously purified by ConA affinity chromatography from procyclic trypanosomes, suggesting that also this TS is highly mannosylated in its native state on the parasite [27].
In summary, we identified TconTS1-LD and TconTS2-LD as a carbohydrate recognition domain (CRD), exhibiting different binding affinities to several oligogalactosyl, oligolactosamine and oligomannosyl glycans, via two independent binding sites. Functionally, the interaction with specific oligomannosyl structures appears to be required to facilitate TconTS oligomerisation, and binding to oligogalactosyl and oligolactosamine oligosaccharides may represent the recognition event associated with TS acceptor substrate binding. Since the LD from T. congolense, T. brucei, T. cruzi and potentially other trypanosomes are structurally related, this may be a general function of TS-LD and may open new avenues for the design of novel inhibitors for therapeutic applications controlling trypanosomiasis in Africa and Latin America.
All chemicals and reagents used in this study were analytical grade. Recombinant EndoHf glycosidase (EndoHf) was from New England Biolabs, UK. Pfu and Taq DNA polymerase, HindIII, NcoI, NotI, SalI Fast Digest restriction enzymes, T4-DNA ligase, isopropyl-β-D-1-thiogalactopyranoside (IPTG), Dithiothreitol (DTT), Coomassie Brilliant Blue (Page Blue), protein molecular weight marker (PageRuler), GeneJET DNA Gel Extraction Kit, BCA Protein Assay Kit, enhanced chemiluminescence system (ECL-Kit), fluorescein diphosphate tetraammonium salt (FDP), Luria Broth (LB) microbial growth medium, anti His-tag mouse polyclonal antibody, anti-mouse-IgG-alkaline phosphatase-conjugated donkey polyclonal antibody (serum purified) were from Thermo Scientific, Germany. Biozym LE Agarose was from Biozyme Scientific, Germany. StrepTactin Sepharose, purification buffers and anti Strep-tag rabbit polyclonal antibody were from IBA, Germany. Anti-mouse-IgG-TexasRed conjugated rabbit polyclonal antibody, anti-rabbit-IgG-TexasRed conjugated donkey polyclonal antibody were purchased from Life Technologies. β-D-galactosyl-(1–4)-α-D-glucose (4α-lactose), β-D-galactosyl-(1–4)-α-D-N-acetylglucosamine (4α-N-acetyllactosamine), 4-α-D-maltose, α-D-glucopyranosyl-(1–4)-α-D-glucose (4α-maltose), α-D-glucopyranosyl-(1–4)-α-D-glucopyranosyl-(1–4)-α-D-glucose (4α-maltotriose), α-methyl-D-mannose, α-D-mannosyl-(1–2)-D-mannose (2α-mannobiose), α-D-mannosyl-(1–3)-D-mannose (3α-mannobiose), α-D-mannosyl-(1–4)-D-mannose (4α-mannobiose), α-D-mannosyl-(1–6)-D-mannose (6α-mannobiose), α-D-mannosyl-(1–3)-[α-D-mannosyl-(1–6)]-D-mannose (3α,6α-mannotriose), polyethylene glycol sorbitan monolaurate (TWEEN 20), Gel Filtration Markers Kit for protein molecular weights between 29,000–700,000 Da were from Sigma-Aldrich, Germany. Concanavalin A (ConA), Sepharose and biotinylated recombinant ConA were purchased from Galab, Germany. VECTASTAIN ABC detection system was from Vector laboratories, UK. Ultrafiltration units Vivacell and Vivaspin6 were from Sartorius, Germany. X-ray film was purchased from GE Healthcare, Sweden. Protino Ni-NTA Agarose and NucleoBond Midi Plasmid DNA Purification Kit were from Macherey-Nagel, Germany. Polyvinylidenedifluoride (PVDF) membranes were from Millipore, Germany. 96-well transparent microtitre plates were from Sarstedt, Germany. High binding 384-well black microtitre plate were purchased from Corning, USA. 6 mL gravity flow columns were from Biorad, Germany.
To obtain non-glycosylated TconTS-LD as recombinant proteins in sufficient amounts, a bacterial expression system based on a modified pET28a+ expression vector was established. Modifications made comprised an N-terminal poly-histidine-tag (His-tag) followed by maltose binding protein (MBP) and a tobacco etch virus (TEV) protease cleavage site [82]. MBP was used to enhance expression and solubility of TconTS-LD in E. coli. SNAP- Strep- and His-tags were employed for affinity purification, detection and immobilisation of recombinant protein as previously described [30,31] The DNA sequence encoding the His-MBP part was amplified form pETM–41 (EMBL, Germany) using Pfu DNA polymerase and the appropriate primers (S3 Table). The purified PCR products were ligated into the NcoI and SalI digested pET28a (Novagene, USA) bacterial expression vector and transformed into chemical competent E. coli BL21 (DE3)(BD Bioscience, Clonetech, USA). Sequence identity was confirmed by commercial sequencing at the Max Planck Institute for Marine Biology (MPI) Bremen and results were evaluated using the Geneious Software.
The modified eukaryotic expression pDEF-based vectors coding for TconTS1, TconTS2, TconTS3 and TconTS4 [30,31] were used as template to amplify TconTS1-LD, TconTS2-LD, TconTS3-LD and TconTS4-LD containing C-terminal SNAP and Strep tags. Two sets of sense primers were designed for each TconTS variant (S3 Table). The same reverse primer including a NotI restriction site (underlined) was used for all TconTS-LD constructs, since all TconTS-LD constructs contain C-terminal Strep-tag. Purified PCR-products were ligated in frame into the HindIII and NotI digested, modified pET28aMBP vector and transformed into E. coli Rosetta (DE3) pLacI (BD Bioscience, Clonetech, USA). Plasmid preparations of pET28aMBP/TconTS-LD were prepared and characterised as described above.
E. coli Rosetta (DE3) pLacI colonies freshly transformed with pET28aMBP/TconTS-LD were inoculated in 20 mL of 50 μg/mL kanamycin containing Luria Broth (LB) medium and incubated overnight at 37°C and 240 rpm shaking. 2 mL of these overnight cultures were transferred into 1 L of 50 μg/mL kanamycin containing LB medium and grown at 37°C and 240 rpm until an optical density at 600 nm of 0.5 was reached. Recombinant protein expression was then induced by the addition of isopropyl-β-D-thiogalactopyranoside (IPTG), with a final concentration of 0.1 mM and cells were incubated for additional 120 min at 37°C and 240 rpm. Cells were harvested by centrifugation for 15 min at 1500 x g, 4°C and the pellet was resuspended in 20 mL lysis buffer 50 mM NaH2PO4, pH 8.0, 300 mM NaCl. Lysis was done by ultrasonication on ice applying 9 pulses of 20 sec each (50 Watts) with 10 sec pauses between pulses. The bacterial lysates were centrifuged for 30 min at 15000 x g, 4°C. Clear supernatants were transferred to 4 mL of equilibrated Ni-NTA beads and incubated on a rotation wheel (6 rpm) at 4°C for 120 min. The suspensions were transferred to 6 mL gravity flow columns in portions, until all beads were settled in the column. Beads were washed with 40 mL wash buffer containing 50 mM NaH2PO4, pH 8.0, 150 mM NaCl, 20 mM imidazole. Recombinant TconTS-LD was eluted using 250 mM imidazole in 50 mM NaH2PO4, pH 8.0, 150 mM NaCl and directly applied to a new gravity flow column containing 1.6 mL StrepTactin beads equilibrated with wash buffer (100 mM Tris-Cl, pH 8.0, 150 mM NaCl and 1 mM EDTA) and beads were washed with 5 column volumes of wash buffer. Recombinant proteins were eluted with wash buffer containing 2.5 mM desthiobiotin and dialysed against 10 mM phosphate buffer, pH 7.4 using a Vivaspin6 filtration unit with a 100 kDa cut off. Purified TconTS-LD was characterised by SDS-PAGE and Western blot analysis and quantified by BCA assay using bovine serum albumin (BSA) as standard.
T. congolense recombinant TconTS1 and TconTS2 containing catalytic (CD) and lectin domain (LD) expressed by CHO-Lec1 cells were prepared from culture supernatants and characterised using SDS-PAGE, Western Blot and BCA assay analysis as described [30].
Glycan arrays consisting of 367 diverse glycans with and without the presence of one of three spacers (sp2, sp3 or sp4 [49]) were prepared from two previously described glycan libraries [83,84]. Amine containing glycans with spacer’s sp2, sp3 or sp4 were synthesised as previously described [49] and glycans without spacers were amine functionalised as previously published [85]. All glycans were suspended in 1:1 DMF:DMSO at a concentration of 500 mM and were printed onto SuperEpoxy 2 glass slides (ArrayIt, Sunnyvale, CA) using a ArrayIt SpotBot Extreme array spotter in a six pin subarray print per glass slide format. All glycans were printed in replicates of four, including four FITC control spots as well as additional position controls (S1 Fig), per subarray using SMP4 pins and a contact time of 1 second at 60% relative humidity, with pins being reloaded after every 12 spots.
Prior to performing glycan array experiments, slides were scanned using a ProScanArray Microarray 4-laser scanner (Perkin Elmer, Waltham, MA) using the blue argon 488 laser set to the FITC settings (492 nm excitation and 517 nm emission). Array slides were blocked with 0.1% BSA in 50 mM phosphate buffered saline (PBS), pH 7.4 for 5 min at 22°C. After washing with PBS, each slide was dried by placing them in an empty 50 mL tube and centrifuging for 5 min at 200 x g (900 rpm). Recombinant TconTS-LD (2 μg) was incubated at a molar ratio of 1:2:3 with anti His-tag mouse polyclonal antibody (10 mg/mL, Cell Signalling Technology), anti mouse-IgG-Alexa555 conjugated rabbit polyclonal antibody (2 mg/mL, Life Technologies) and anti rabbit-IgG-Alexa555 conjugated goat polyclonal antibody (2 mg/mL, Life Technologies) in 50 mM PBS, pH 7.4 containing 0.1% BSA and 10 mM maltotriose for 15 min on ice protected from light. All subarrays on the slide were isolated using a Gene Frame (1.5 x 1.6 cm, 65 μL, Abgene, Epsom, UK) prior to the addition of the TconTS-LD-antibody mix to the array. A coverslip was applied to the GeneFrame and array slides incubated for 30 min at 22°C in the dark. The GeneFrame and coverslip were subsequently removed and the slide gently washed twice with 50 mM PBS, pH 7.4 containing 0.01% TWEEN 20 and 10 mM maltose, and once with 50 mM PBS, pH 7.4 containing 10 mM maltose. Slides were dried by centrifugation for 5 min at 200 x g (900 rpm), allowed to air dried for a further 5 min, and the fluorescence associated with the array spots detected using the microarray scanner settings outlined above. Image analysis and spot visualisation was performed using the ProScanArray software, ScanArray Express (Perkin Elmer). The resulting images were visually examined. Fluorescence signals were judged as being positive, if all four replicates for a glycan were clearly detectable (S1 Fig).
TconTS2-LD was several times buffer exchanged to 10 mM deuterated phosphate buffer, pD 7.4 using a Microcon centrifugal ultrafiltration device (cut off 10 kDa). 200 μL of a solution containing 5.5 μM TconTS2-LD was prepared for each experiment. 1024 Scans per STD NMR experiment were acquired as described before [40]. In separate experiments, lactose and 3α,6α-mannotriose were added to the TconTS2-LD solution resulting in 3.45 mM (for lactose) or 1.73 mM (for 3α,6α-mannotriose or in the mixture of both oligosaccharides) final concentrations. The STD NMR spectra were obtained by subtracting the on- from the off-resonance spectra. As controls, STD NMR spectra for only TconTS2-LD or ligand were recorded under identical conditions used. Data acquisition and evaluation was performed using NMR software TopSpin 3.2 (Bruker Daltonics, Germany).
Microtitre plate based binding and inhibition assays, used for characterising protein binding to sialylated glycoproteins, has been described for siglecs [86]. In this study a modified version of these assays was established to investigate TconTS-LD binding to mannosylated glycoproteins. Recombinant huS2-Fc expressed and purified as previously described [87] was used as binding partner for TconTS-LD, since it contains high-mannose N-glycans due to expression in CHO-Lec1 cells. 5 μL of 5 μg/ mL huS2-Fc in 50 mM NaHCO3, pH 9.6, were immobilised on a high binding 384-well microtitre plate (Corning, USA) overnight at 4°C. The plate was washed five times with 20 μL 10 mM Tris-Cl, pH 7.5, 150 mM NaCl containing 0.05% Tween20 (TBS-T) per well. A 1:2 serial dilution of TconTS2-LD ranging from 4–0.125 μg/ mL in 10 mM TBS-T was prepared. 0.2 μg/ mL anti His-tag mouse polyclonal antibody and 0.2 μg/ mL anti mouse-IgG alkaline phosphatase(AP)-conjugated donkey polyclonal antibody was added to each dilution step and incubated on ice for 30 min. 5 μL of each sample were transferred in triplicates onto the washed microtitre plate and centrifuged for 1 min at 600 x g. The plate was covered with parafilm and incubated at 4°C for additional 3.5 hours. After washing the plate 4 times with 10 mM TBS-T and twice with 10 mM TBS, 20 μL of freshly prepared fluorescein diphosphate (FDP) substrate solution (50 mM Tris-Cl, pH 8.5, 10 mM MgCl2, 20 μM FDP) was added to each well and the kinetic fluorescence measurement was immediately started employing a Tecan Infinite F200 Pro microtitre plate reader (Tecan, Germany). As controls, wells containing and lacking immobilised huS2-Fc were incubated with both antibodies but in the absence of TconTS2-LD. For comparison, 4 μg/ mL recombinant catalytic domain (CD) of TconTS2 was used instead of TconTS2-LD under the same conditions used.
Inhibiton assays were performed following the same procedure as for the binding assay, but free oligomannose N-glycans were added as potential competitive inhibitors during the incubation with TconTS2-LD. Oligomannose N-glycans were released by endoglycosidase H (EndoHf) treatment of 10 μg huS2-Fc in 50 mM sodium citrate, pH 5.5 at 37°C for 4 hours. Proteins were acetone precipitated at -20°C overnight [30]. Following centrifugation the supernatant was transferred into a fresh reaction tube and solvent was removed using SpedVac evaporator for 1 hour at 30°C, 100 mbar. Glycans were resuspended in 10 mM TBS and used in inhibition assay as 1:2 dilution series. Data acquisition was done using the software Magellan 7.2. Binding and inhibition curves were generated using SigmaPlot 11.
TconTS1 and TconTS2 were enzymatically deglycosylated using recombinant EndoHf cleaving the chitobiose core (GlcNAc(1–4)-β-GlcNAc) of high-mannose N-glycans from glycoconjugates [88]. In brief, 500 μL 10 mM phosphate buffer, pH 7.4 containing 100 μg TconTS and 4000 units EndoHf were incubated for 4 hours at 37°C. For deglycosylation of TconTS under denaturing conditions, 100 μg TconTS were incubated in 20 μL denaturing buffer containing 40 mM dithiothreitol (DTT) and 0.5% sodium dodecyl sulfate (SDS) for 10 min at 95°C. After the addition of sodium citrate, 50 mM final concentration, pH 5.5 and 4000 units EndoHf, reaction mix was incubated for additional 60 min at 37°C. N-deglycosylation efficiency was determined by ConA lectin blot analysis.
Oligomerisation of TconTS was analysed employing a fast protein liquid chromatography (FPLC) system (Amersham Pharmacia, USA) using Superdex 200 10/300 GL (GE Healthcare, Sweden) size exclusion column and photometric detection at 280 nm. Chromatographic analysis were done at 4°C. In brief, column was equilibrated with 10 mM phosphate buffer pH 7.4 and calibrated using a gel filtration marker kit for protein molecular weights between 29,000–700,000 Da (Sigma-Aldrich, Germany) according to manufactures instructions. 100 to 300 μg TconTS in 500 μL sample volume were injected and separated at a flow rate of 0.5 mL/min. Absorbance at 280 nm was continuously recorded through an analog writer and subsequently transformed to digital chromatograms using the software SigmaPlot 11. EndoHf treated samples were analysed in the same manner.
Protein samples were separated employing SDS-PAGE as described previously [89] using a MiniProtean III electrophorese Unit (Bio-Rad, Germany) and stained with Coomassie Brilliant Blue.
Western blot analysis were performed as previously described [30], using primary anti Strep-tag rabbit antibody (1:1000) and secondary anti rabbit-IgG donkey horseradish peroxidase(HRP)-conjugated antibody (1:40000). Membranes were developed using enhanced chemiluminescence system (ECL-Kit, Thermo Scientific, Germany) and X-ray film (GE Healthcare, Sweden).
ConA lectin blots were performed similar to the procedure for Western blotting. Instead of applying primary antibody, 2 μg/mL solution of biotinylated recombinant ConA in 10 mM phosphate buffer, pH 7.4 was added to the membrane and incubated overnight at 4°C. Avidin-biotin HRP conjugated system (VECTASTAIN ABC-Kit, Vector Labs, UK) was used for detection according to manufactures instructions.
Homology models of TconTS-LD containing or lacking the α-helix were calculated employing the molecular modelling software Yasara 13.3.26 [90–95] as previously described [30]. In brief, crystal structure of Trypanosoma cruzi TS (PDB: 3b69) [14] was used as a template structure for calculating the models. Yasara homology modelling module were modified manually from the default settings of the program as follows: Modelling speed: slow, PsiBLASTs: 6, EValue Max: 0.5, Templates total: 1, Templates SameSeq: 1, OligoState: 4, alignments: 15, LoopSamples: 50, TermExtension:10. The molecular and electrostatic potential surfaces were calculated using the ESPPME (Electrostatic Potential by Particle Mesh Ewald) method of Yasara Structure with the following parameters: Force field: AMBER96, Algorithm used to calculate molecular surface: numeric, Radius of water probe: 1.4 Å, Grid solution: 3, Maximum ESP: 300 kJ/mol. Red colour indicates a positive potential, blue a negative and grey a neutral.
|
10.1371/journal.pntd.0000928 | A DNA Vaccine against Chikungunya Virus Is Protective in Mice and Induces Neutralizing Antibodies in Mice and Nonhuman Primates | Chikungunya virus (CHIKV) is an emerging mosquito-borne alphavirus indigenous to tropical Africa and Asia. Acute illness is characterized by fever, arthralgias, conjunctivitis, rash, and sometimes arthritis. Relatively little is known about the antigenic targets for immunity, and no licensed vaccines or therapeutics are currently available for the pathogen. While the Aedes aegypti mosquito is its primary vector, recent evidence suggests that other carriers can transmit CHIKV thus raising concerns about its spread outside of natural endemic areas to new countries including the U.S. and Europe. Considering the potential for pandemic spread, understanding the development of immunity is paramount to the development of effective counter measures against CHIKV. In this study, we isolated a new CHIKV virus from an acutely infected human patient and developed a defined viral challenge stock in mice that allowed us to study viral pathogenesis and develop a viral neutralization assay. We then constructed a synthetic DNA vaccine delivered by in vivo electroporation (EP) that expresses a component of the CHIKV envelope glycoprotein and used this model to evaluate its efficacy. Vaccination induced robust antigen-specific cellular and humoral immune responses, which individually were capable of providing protection against CHIKV challenge in mice. Furthermore, vaccine studies in rhesus macaques demonstrated induction of nAb responses, which mimicked those induced in convalescent human patient sera. These data suggest a protective role for nAb against CHIKV disease and support further study of envelope-based CHIKV DNA vaccines.
| Chikungunya fever epidemics are sustained by a cycle of human-mosquito-human transmission, with the epidemic cycle being similar to those of dengue and urban yellow fever. While the threat of a pandemic continues to engage the public's attention, the peculiar problems associated with the more immediate and very real seasonal epidemics are also worthy of consideration. Specifically, there are limited viral strains that have been characterized and available for laboratory study as well as limited knowledge of immune responses induced to the virus. In this study, we isolated CHIKV virus from an acutely infected human patient and used this new virus to develop a neutralization assay and a challenge stock, which is effective in a mouse model. Furthermore, we analyzed the ability of an envelope-based synthetic DNA-based vaccine to impact viral disease in the mouse model and to generate protective levels of immune responses in nonhuman primates. We observed that this novel vaccine approach generated protective levels of immune responses in both mouse and non-human primate models. We believe that these studies advance the field of Chikungunya vaccine research as well as the study of immune protection to CHIKV.
| Chikungunya virus (CHIKV) is an enveloped, single-stranded, positive-sense RNA virus that belongs to the family Togaviridae, genus Alphavirus, and is part of the Semliki Forest virus antigenic complex [1], [2], [3], [4]. CHIKV has been responsible for unprecedented, explosive outbreaks during 2004 and 2007 in India and the Indian Ocean islands [2], [4], [5], [6], [7]. These outbreaks represent the largest documented cases associated with the virus [8]. Chikungunya fever, the disease caused by CHIKV, was first recognized in epidemic form in East Africa in 1952-1953 [3], [9] and the viral agent was first isolated at that time from the blood of a febrile patient in Tanzania [9]. In the local Swahili dialect, “Chikungunya” means “stooping” or “bending”, which describes the physical position often assumed by CHIKV-infected patients [3], [9], [10]. Since that time, CHIKV has been identified as the agent responsible for major epidemics in both Africa and Southeast Asia and continues to be a re-emerging agent of great interest to public health [1], [11], [12], [13]. Despite its importance as an emerging virus and potential biological weapon, there are no specific licensed vaccines or antiviral treatments for Chikungunya.
Currently, CHIKV is geographically distributed from Africa through Southeast Asia and South America and is principally transmitted to humans through Aedes mosquitoes [2], [14]. Recently, a mutation in the CHIKV envelope (E1-A226V) was found to be directly responsible for the significant recent increase in CHIKV infectivity, and studies confirmed that this single amino acid substitution can influence vector specificity. This finding provides a plausible explanation of how this variant virus caused an epidemic in a region lacking the normal insect vector Ae. Aegypti [15], [16].
While Chikungunya is not typically associated with human mortality, epidemics often present public health threats due to substantial morbidity, suffering, and loss of economic productivity. The incubation period of the virus ranges between 1–2 weeks and infected individuals usually experience an acute illness with fever, headache, rash, nausea, vomiting, incapacitating polyarthralgia, severe muscle pain, and joint stiffness [17]. The most prominent clinical feature of CHIKV disease is arthralgia, which can be debilitating and prolonged [5], [17], [18], [19], [20]. Though the pathogenesis of the virus in humans is not exactly clear, recent findings of CHIKV infection in muscle tissue and macrophages could explain some features of its clinical manifestations [12], [18], [21], [22], [23]. Due to these characteristic clinical symptoms of infection, outbreaks of CHIKV have devastating public health and economic effects.
The first reported outbreak of Chikungunya occurred on Lamu Island, Kenya, in 2004. Later the virus spread to La Reunion Island, infecting more than two hundred thousand individuals [11], then to other islands in the Indian Ocean [24], and then finally into India in 2006 [25]. Furthermore, the La Reunion isolate outbreaks were associated with unexpected morbidity [26]. Importantly, exposed travelers returning from the affected areas to Europe, the US, Canada, Hong Kong, and numerous other countries have carried the virus into these new niches where the imported cases were subsequently identified [8]. Although these instances of viral importation were effectively controlled, they serve as a reminder of how easily this agent could be introduced and spread into new geographical locations including industrialized nations.
Thus far, while several vaccines have been developed against this disease, such as a formalin inactivated vaccine [27], a virus-like particle vaccine [28] and a live attenuated vaccine [29], none have advanced to clinical development and therefore illustrate an important area of need. Recently, a novel consensus-based DNA vaccine was developed by our laboratory and reported to be immunogenic in mice [30], inducing both measurable cellular and humoral immune responses. However, the neutralizing and hemagglutination-inhibiting antibody responses were not examined. While nAb to CHIKV during natural infection in humans are not well understood, recent sero-surveys during outbreaks suggest a protective role for prevention of replication [23], [28], [30], [31], [32]. Furthermore, prior infection is thought to be protective against subsequent CHIKV infection. Therefore, by examination of nAb titers during natural infection in humans, a benchmark for vaccine development in this study, we aimed to establish a correlation with these responses and protection.
Accordingly, we isolated a new viral isolate from an acutely infected patient, and termed PENN CHIKV-2008 (PC-08), and characterized its biology in mice, and also used it to develop an in vitro neutralization assay. Furthermore, we modified our previous DNA vaccine to optimize for the capacity of neutralization by designing a single consensus envelope DNA vaccine construct expressing all three envelope proteins. We compared its effects in mice with a CTL-only-inducing Capsid vaccine in a challenge model. Finally, we compared in non-human primates vaccine-induced immune responses with human CHIKV convalescent sera as a measure of protective immunity.
Samples used in the study were provided by Regional Medical Research Center (RMRC), Indian Council of Medical Research (ICMR), Port Blair, India and Sri Ramachandra Medical College & Research Institute (SRMC&RI), Chennai, India. These samples were previously obtained with proper informed consent at the respective institutions. The collected samples were coded and stored. The samples do not contain identifying information regarding the patients that donated the samples and under an agreement between the collaborating institutions that determined at no point was the key decoding patient data disclosed to the investigators performing the assays. The study was reviewed by the respective Institutional Human Ethical Committees and approvals were obtained. The de-identified samples were transported to University of Pennsylvania, PA, USA following EHRS guidelines and after obtaining the CDC import permit to Import or Transport Etiologic Agents, Hosts or Vectors of Human Disease (Permit # 2008-03-027). Appropriate practices and procedures as defined in the Biosafety in microbiological and biomedical laboratories (US Dept. of Health and Human Services) were used in sample handling. Samples were stored at −80°C in a bio safety level-3 (BSL-3) facility at the University of Pennsylvania, PA, USA.
Primate studies were conducted by the subcontract at Bioqual Inc, MD. The animal management program of this institution is accredited by the American Association for the Accreditation of laboratory Animal Care and meets NIH standards as outlined the in the Guide and care and use of laboratory animals. This institution also accepts as mandatory PHS policy on Humane Care of Vertebrate Animals used in testing, research and training.
Vero 76 (ATCC CRL-1587) and RD (ATCC CCL-136) cells were cultured in complete medium (Eagle's Minimum Essential Medium) containing 10% fetal bovine serum, 1 mM glutamine, 1 mM sodium pyruvate, 100 U/ml penicillin and 100 µg/ml streptomycin. Cells were incubated in a 5% CO2 humidified incubator at 37°C [33]. 8-week-old female BALB/c mice (Jackson laboratories, Indianapolis, IN) were used in these experiments. All animals were housed in a temperature-controlled, light-cycled facility in accordance with the guidelines of the National Institutes of Health (Bethesda, MD, USA) and the University of Pennsylvania Institutional Animal Care and Use Committee (IACUC). Rhesus macaques (Macaca mulatta), aged 4–8 years, were housed at Bioqual, Inc, Rockville, MD 20850. The experiments reported herein were conducted according to the principles set forth in the Guide for the Care and Use of Laboratory Animals, Institute of Laboratory Animal Resources, National Research Council, Department of Health and Human Services (DHHS) Publication number NIH-86-23 (1985).
The CHIKV DNA constructs pCHIKV-E1, pCHIKV-E2, and pCHIKV-Capsid have been previously described [30]. The combined CHIKV-envelope construct was designed with the structural genes E3, E2 and E1 linked together in a single construct with furin cleavage sites between the individual genes [34]. The consensus gene sequences were constructed using the predicted consensus sequences from the sequences available in the NCBI Genbank database and designated as pMCE321. We note that while the 6K protein is also a structural constituent of envelope, we did not include it in our vaccine construct because we sought to create a minimal vaccine construct capable of inducing broadly protective immune responses. The primary role of 6K is postulated to function in the selection of lipids that interact with the transmembrane domains of the glycoproteins [35]. Consensus sequences were optimized for Env expression, including codon and RNA optimization (GeneArt, Regensburg, Germany), a novel leader sequence was added [36] as were furin cleavage sites to facilitate Env processing as previously published [34] and inserted into the pVax1 expression vector (Invitrogen, CA) and designated as pMCE321. DNA preparations were made at Aldevraon (Forgo, ND), as previously described and formulated at 10 mg/ml in water.
Expression of pMCE321 was verified by immunoblotting and immunofluorescence. Vero and RD cells (1×106cells) were transfected with pMCE321 constructs using the Fugene transfection method (Roche, Indianapolis, IN). Forty-eight hours post-transfection, proteins were isolated using cell lysis buffer, fractionated on SDS-PAGE (12%), and transferred to nitrocellulose using iBlot Dry Blotting System (Invitrogen, CA, USA). Immunodetection was performed using SNAP i.d. Protein Detection System (Millipore, MA, USA) with specific mouse antiserum and the expressed proteins were visualized with horseradish peroxidase-conjugated goat anti-mouse IgG using an ECL detection system (Amersham Pharmacia Biotech, Piscataway, NJ) [33].
For immunofluorescence, Vero cells (2×105 cells) were seeded in 2-chamber tissue culture treated glass slides (BD Falcon, MA, USA) and transfected with pVax1-E1, pVax1-E2, and pMCE321 vaccine constructs or control pVax1 vector. Forty eight hours post-transfection, cells were fixed with 2% paraformaldehyde, blocked with Glycine/BSA, and then incubated overnight at 4°C with mouse anti-Env IgG antibodies (1∶500 dilution). Excess antibodies were washed off and the secondary antibody AlexaFluor 488-anti mouse IgG (Invitrogen, Molecular Probes, USA) was added and incubated for 2 hours at 37°C. The cells were counter stained with DAPI for visualizing the nucleus and fixed with fluoromount G (Electron Microscopy Sciences, PA, USA). The confocal images were acquired with Zeiss LSM510 META NLO Laser Scanning Confocal Microscope with Two Photon Excitation at the Biomedical Imaging Core, University of Pennsylvania, PA, USA.
For histopathology studies, naïve and CHIKV DNA vaccinated mice, challenged with the CHIKV isolate, were bled and sacrificed on day 14 p.i. Tissue samples (brain, heart, lung, liver, and kidney) were collected and fixed in 10% buffered formalin solution for 24 h, and then stored in 70% ethanol prior to embedding, sectioning, and staining using hematoxylin and eosin (H&E) stain [23], [37].
CHIKV-infected patient sera samples were obtained from ICMR, Port Blair, India and SRMC&RI, Chennai, India. The number of days from onset of illness to sampling ranged from 1 to 14 d. All cases had complained of fever with median duration of 3–5 d. Other common symptoms included Chills and Rigors (23%), myalgias (6%), gastrointestinal symptoms of diarrhea, abdominal pain (20%), vomiting (20%), severe joint pain (3%) and headache (13%) (Table 1).
The isolation of virus from patient sera was carried out in Vero cells grown to 90–95% confluence in Eagle's complete medium (MEM) in T-25 tissue culture flasks (BD Falcon, USA). Patient serum (100 µl) was mixed with MEM (400 µl), adsorbed onto the cell culture (after removing the growth medium) for 2 hrs at 37°C, and then replenished with complete growth medium following washing with MEM. The inoculated cells were incubated at 37°C with 5% CO2 for 5 to 7 days and monitored daily for the development of cytopathic effects (CPE). When CPE was observed in more than 90% of the cells, the flasks were frozen at −80°C, freeze–thawed five times to facilitate cell lysis and virus release, and then centrifuged at 3,000rpm for 10 min to remove cellular debris. The isolate was then passaged five times in Vero cells, titrated, and stored at −80°C. The virus stock was designated as PC-08 (PENN CHIKV strain - 08).
Viral RNA from patient serum was extracted using QIAamp Viral RNA mini kit according to manufacturer's instructions (Qiagen Inc, Valencia, CA). A one-step RT-PCR test was carried out using Qiagen One step RT-PCR kit on a block thermo cycler (PTC-100, MJ Research, Waltham, MA, USA); 5 pmoles of each primer (CHIK-forward 5′-TATCCTGACCACCCAACGCTCC-3′ and CHIK-reverse 5′-ACATGCACATCC CACCTGCC-3′ amplify a 305 bp region within the gene coding for the viral envelope protein E2) were used with 10 µl 5X RT PCR buffer, 2 µl dNTP mix, 5 µl Q solution, 1 µl enzyme mix, 25 µl water, and 5 µl extracted RNA for a total reaction volume of 50 µl [38]. Thermocycler conditions were as follows: 50°C for 30 min, 94°C for 2 min, then 40 cycles of 94°C for 15 s, 55°C for 30 s and 68°C for 2 min 20 s with a final extension at 68°C for 5 min. PCR products were purified by gel extraction (Qiagen Inc, Valencia, CA) and sequenced at the University of Pennsylvania DNA Sequencing Facility.
BALB/c mice (n = 14/each group) were immunized with the pCHIKV-Capsid, the pCHIKV-Envelope (pMCE321) constructs, or control pVax1 (25 µg) 3 times at 2-week intervals, according to a standard DNA immunization protocol. All injections were delivered into the quadriceps muscles in a total volume of 25 µl followed by i.m. electroporation (Inovio Biomedical Corporation, Blue Bell, PA) as described previously [30], [39]. After the last immunization, 4 mice from each group were sacrificed for immunology assays (IFN-γ and Abs ELISA), while the remaining mice (n = 10) were used for the challenge studies.
Mice were challenged with 7 log10 PFU of the CHIKV isolate (PC-08) by intranasal infection (i.n.) in a total volume of 25 µl and animals were checked daily for clinical signs of infection, such as lethargy and hind limb weakness. Additionally, body weight was monitored [17], [31], [40], [41]. Animals were then sacrificed either 14 days p.i. or earlier if a weight loss of more than 30% was observed.
Non-human primate studies were conducted under a contract at Bioqual Inc, MD. The animal management program of this institution is accredited by the American Association for the Accreditation of Laboratory Animal Care and meets NIH standards as set in the guide for the Care and Use of Laboratory Animals. This institution also accepts as mandatory the PHS policy on Humane Care of Vertebrate Animals used in testing, research and training. Animals were allowed to acclimate for at least 30 days in quarantine prior to any immunization. Four rhesus macaques were immunized at weeks 0, 4, and 8 with 1 mg/construct (at a concentration of 10 mg/ml) of CHIKV envelope (pMCE321) and 3 rhesus macaques were immunized with pVax1 vector. DNA was delivered into the quadriceps muscle (intramuscularly (i.m.) followed by in vivo electroporation as previously described [39]. Animals were bled every 2 weeks. Five ml of blood was collected for serum studies and ten ml of blood was collected in EDTA tubes, and peripheral blood mononuclear cells were isolated by standard Ficoll-Hypaque centrifugation and resuspension in complete culture medium (RPMI 1640 with 2 mM/liter L-glutamine, 10% heat-inactivated fetal bovine serum, 100 IU/ml penicillin, 100 µg/ml streptomycin, and 55 µM/liter β-mercaptoethanol). Red blood cells were lysed with ammonium chloride-potassium (ACK) lysis buffer (Invitrogen, CA).
The 50% tissue culture infectivity dose (TCID50) was calculated and a standard concentration of virus was used for the micro-neutralization test throughout these studies. Microneutralization assays were performed with human patient samples as well as using the mouse sera from pCHIKV-E1/pCHIKV-E2/and pCHIKV-Env (pMCE321) immunized animals, as described previously [42]. Briefly, the patient, mouse, or monkey sera were serially diluted in MEM (1∶10 to 1∶10,240) and incubated with an equal volume of CHIKV (100 TCID50) at 37°C. After 90 min, the virus-serum mixture was added to a monolayer of Vero cells (100,000 cells (for patient and mouse samples) and 15,000 cells (for monkey samples) per well) in a 96-well flat bottom plate and incubated for 5 days at 37°C in a 5% CO2 incubator. The highest titer at which no CPE was observed was recorded as the nAb titer.
HI assays were performed as described previously for Arboviruses [43] and CHIKV virus isolate was used as the antigen. Kaolin-treated sera from human patient samples or immunized mice were diluted and tested at serial 2-fold dilutions from 1∶10 to 1∶1,280 at pH 6.3, using eight hemagglutination (8HA) units of antigen (CHIKV) and 0.4% goose erythrocytes. The highest dilution of the serum that inhibited hemagglutination was recorded as the HI titer. HI titers greater than or equal to 20 were considered positive [41].
ELISpot assays were performed as previously described [33], [39]. Briefly, 96-well ELISpot plates (Millipore) were coated with anti-mouse IFN-γ capture Ab (R&D Systems) and incubated for 24 h at 4°C. The following day, plates were washed with PBS and blocked for 2 h with 1% BSA. Two hundred thousand splenocytes from the pMCE321 Env-immunized mice were added to each well and incubated overnight at 37°C in 5% CO2 in the presence of media alone (negative control), media with Con A (positive control), or media with peptide pools (10 µg/ml) consisting of 15-mers overlapping by 9 amino acids and spanning the length of the appropriate protein. After 12 h, the cells were washed and then incubated for an additional 24 h at 4°C with biotinylated anti-mouse IFN-γ Ab (R&D Systems, Minneapolis, MN). Streptavidin–alkaline phosphatase was added to each well after washing and then incubated for 2 h at room temperature. The plates were washed, and then 5-bromo-4-chloro-3′-indolylphosphate p-toluidine salt and nitro blue tetrazolium chloride (chromogen color reagent; R&D Systems) were added to the wells. Lastly, the plates were rinsed with distilled water, dried at room temperature, and spot forming units (SFU) were quantified by an automated ELISpot reader (CTL Limited). For each sample, the raw values were normalized to SFU per million splenocytes.
CD8+ T-cell depletion studies were carried out following immunomagnetic cell separation. Dynabeads (Dynal Biotech) monkey CD8 (clone BW135/80) was used as the method for separation, resulting in 90% depletion in 20 min using 1×107 beads/ml for 2.5×106 splenocytes. Depletion was conducted as described by the manufacturer. The negatively isolated cells (CD8+ T cell depleted) were transferred to a second tube for further use in the ELISpot assays [44], [45].
The proinflammatory cytokines levels following CHIKV infection were determined in the culture medium using a commercially available ELISA kit following the manufacturer's instructions (R&D Systems Inc, MN). All samples were analyzed in triplicate [33].
Data was collected from cellular assays and presented as the mean +/− standard deviation which was calculated from triplicate wells of pooled samples from each experimental group. Prior to all statistical analysis, the normality of the data was confirmed with Levine's test. Analysis between groups was performed using independent samples t-test. Comparisons among three groups were performed with ANOVA with a post-hoc Fisher's Least Significant Difference (LSD) test to correct for multiple comparisons between groups. In each case, p≤0.05 was considered to be significant. All statistical analysis was carried out using the Statistical Package for the Social Sciences (SPSS).
The clinical manifestations caused by the Chikungunya outbreaks in 2005 to 2007 appeared varied and somewhat divergent from those observed in the early outbreak in 1953. In particular, the hemorrhagic tendency of CHIKV infections is not as predominant as that of past outbreaks [46]. Accordingly, we sought to study serum samples from the recent CHIKV outbreaks and characterize viruses and the humoral responses from acute and convalescent sera of infected patients. During the recent outbreak in India, serum samples were regularly collected from patients with complaints of high fever, chills, headache, vomiting and severe abdominal pain (outpatient department, Sri Ramachandra Medical Centre, Chennai, India & at Regional Medical Research Centre (ICMR), Port Blair, India). From this population, thirty patient samples suspected for CHIKV were randomly selected and included in this study; all of the 30 patients from the outbreak area experienced fever lasting 2–15 days with high-grade temperature (38°C to 40°C), and in some patients (23%) accompanied by chills and rigors. Vomiting and abdominal pain (20% of the patients) as well as headache (13% of cases) was also observed. While myalgia or joint pain was seen only in 3% of the patients, rashes or hemorrhages were not observed in this patient population. Furthermore, sore throat and retro-orbital pain as seen in other common viral infections was not prominent in this outbreak. A summary of the clinical and laboratory observations in this CHIKV-infected patient population is listed in Table 1. Importantly, 13 samples (43%) were RT-PCR positive to CHIKV primers demonstrating that some contained Chikungunya virus.
While samples collected during the first 48–72 h of infection are typically ideal for virus isolation [28], [37], we were able to isolate CHIKV successfully from RT-PCR positive and symptomatic patient samples collected at the 3–4 days post CHIKV infection. Isolation of virus from the serum of a CHIKV positive patient who had a high-grade fever (40°C) lasting for 2 days was verified by the observance of massive cell death (Cytopathic effect: CPE) in Vero cells and by RT-PCR (Fig. 1). As seen in Fig. 1A, the isolated virus induced CPE indicating the presence of infectious CHIKV and successful virus production. To further confirm the identity of the virus that caused CPE, we extracted RNA from the infected cell culture supernatant and performed a one step RT-PCR to amplify a part of the CHIKV E2 gene by reverse transcriptase PCR (Fig. 1B) and electron micrographs of CHIKV viral isolates (Fig.1C) [38]. The E2 gene was selected as the target region for the RT-PCR because this gene shows a high degree of divergence among the alphaviruses [2] and harbors virus-specific nucleotide stretches suitable for primer design. The sequences from the reaction amplicons were then analyzed via a Genbank BLAST search and showed sequence similarity with CHIKV strain Ross (Fig. 1D). This virus was designated as PC-08 (PENN CHIKV strain-2008).
Previous studies from our laboratory using the envelope E2 and E1 DNA vaccine constructs showed the induction of cellular and humoral responses in vaccinated mice [30]. In the present report, we modified the previous vaccine to optimize its capacity for induction of neutralization Abs by designing a single consensus envelope vaccine construct that expresses all three of the CHIKV envelope glycoproteins (E3+E2+E1). Consensus sequences were optimized for expression, including codon and RNA optimization [30], [47] insertion of a novel leader sequence [36] as well as furin cleavage sites between envelopes to facilitate envelope processing as previously reported [34] and inserted into the pVax1 expression vector and verified by sequencing and designated as pMCE321 (Fig. 2A).
Expression of the pMCE321 vaccine constructs in vitro was verified by immunoblotting and immunofluorescence techniques. The vaccine constructs expressed strongly in the transfected BHK-21 and Vero cells and the envelope glycoproteins were detected in the pMCE321 transfected lysates by Western blot using envelope E1 antiserum (Fig. 2B). Further, to evaluate the expression of envelope proteins immunofluorescence techniques were performed with pMCE321 immunized sera in Vero cells transfected with pVax1, pCHIKV-E1, pCHIKV-E2 or pMCE321. Immunofluorescence showed envelope staining of the expressed proteins in the cytoplasm, which strongly suggested immune reactivity to each envelope component of the fusion protein (Fig. 2C). Further, to visualize the expression of pMCE321, pCHIKV-E1 and pCHIKV-E2, we performed a parallel FACS analysis from transfected cells, and studied the surface expression of envelope proteins. Interestingly, the pCHIKV-E1 and E2 expression profile was almost identical (Fig. 2D). Unlike E1 and E2 sera expression, pMCE321 sera respond to strong envelope expression in transfected cells. These findings demonstrated the ability of the pMCE321 construct to potently express in mammalian cells and that the Abs induced by these constructs were able to react with the individual envelope glycoproteins E1 and E2.
The ability of pMCE321, to induceCD8+ CTL responses in mice after three immunizations was determined by IFN-γ ELISpot assay. Expression of all three of the envelope glycoproteins from the single DNA vaccine construct induced detectable cellular immune responses against CHIKV envelope specific peptide pools (Fig. 3A). The results of the IFN-γ ELISpot assay 1 week following the third i.m. immunization (mean count, 1,613±117) for pMCE321 against both peptide pools showed strong cellular immune responses to administered envelope Ag in contrast to the control. pCapsid- or pVax1-immunized mice showed no envelope-specific Ag-specific responses at any point during the study. Capsid-specific T cell responses were induced to the Capsid vaccine. Because of increased T cell responses to the combined CHIKV-Envelope vaccine observed in mice, we anticipated that the Ab responses to combined envelopes would be similarly increased. To examine this, sera collected one week after each immunization were tested by ELISA to detect the induction of Envelope-specific IgG. Interestingly, mice immunized with the pCME321 complex displayed significantly higher levels of envelope-specific serum IgG than mice immunized with E1, E2 or E3 alone (Fig. 3B). As expected, the control plasmid, pVax1 did not elicit any detectable Ab responses as determined by ELISA.
The combination of CHIKV envelope glycoprotein genes into one vaccine construct, pMCE321, induced measurable levels of neutralizing and HI antibodies which are significantly greater than responses induced by pCHIKV E1 and E2 constructs alone (Fig. 3C–D; p<0.001). Interestingly, we also observed that the E2 and E1 constructs when delivered individually were able to induce high levels of neutralizing Ab responses or HI titers respectively, but not both showing segregation of these functions. Importantly, the pMCE321 construct was able to drive both neutralizing and HI Ab responses at levels higher than either construct on its own. Taken together, these data demonstrate that immunization with the pMCE321 DNA vaccine induced both cellular and humoral immunity in mice.
We next addressed whether levels of CHIKV vaccine-elicited immunity were able to confer protective immunity by virtue of its cellular and neutralizing/HI Ab responses in mice. A virus challenge study was conducted to assess protective efficacy of the pMCE321 envelope vaccine in comparison with a CHIKV-Capsid vaccine which induced cellular responses, but no neutralization or HI responses. CHIKV-challenged mice were monitored daily for 14 days p.i. and outcome of the challenge was evaluated based on the common signs of Chikungunya infection in mice such as reduction in body weight, survival, lethargy, and hind limb weakness reported in previous studies [31], [41]. A recent study by Ng et al., strongly suggested that proinflamatory cytokines such as IL-1β, TNF-α and IL-6 are biomarkers that have utility in measuring disease severity during CHIKV viral infection [48]. Therefore we also analyzed the production of these pro-inflammatory cytokines in naïve and vaccinated mice post infection and compared the levels to naïve-uninfected mice. Neither IL-6 nor TNF-α were detected in significant amounts in naïve mice; conversely, the production of IL-6 and TNF-α in CHIKV-infected mice (virus) was significantly greater than that in naïve mice (Fig. 4A–B). Similarly, secretion of proinflammatory mediators including IL-6 and TNF-α was measured in vaccinated mice and found to be strongly activated in pCapsid-immunized mice. In contrast remarkably lower levels were detected in pMCE321-immunized mice (Fig. 4A–B) (p<0.001 versus pCapsid). These data suggest that CHIKV Envelope DNA immunization, and not Capsid vaccination, was more effective in minimizing the secretion of proinflammatory cytokines that are commonly associated with viral pathology.
Further, signs of disease in the naïve and vaccinated mice infected with the CHIKV isolate were detected and all mice showed a reduction in body weight for a period of 3 days p.i. However, after this initial period, pMCE321-immunized animals exhibited a restoration in body weight on average when compared with either naïve or Capsid-vaccinated animals. While none of the animals died naturally due to infection, the Capsid-immunized and naïve groups continued to lose body weight and were subsequently euthanized over the period of 7–12 days p.i when their body weight loss was greater than 30% of the pre-challenge weight (Fig. 4C). In contrast 100% of the pMCE321 vaccinated animals recovered and survived beyond day 12 p.i. Viremia on day 5 after i.n. infection was also measured in five unvaccinated animals infected with the CHIKV virus (mean titer log10 1.5 PFU/ml ±0.42), and vaccinated animals (mean titer log10 0.58 PFU/ml ±0.17). The vaccinated animals had significantly lower viremia (p< 0.001) with no signs of infection and remained apparently healthy (Fig. 4D).
Histopathological studies were conducted in vaccinated and naïve animals in the brain for neurological manifestations, liver (the initial site of virus replication), lungs (the portal of entry alternate to skin), heart and kidneys.
Nonhuman primate studies were performed to determine whether CHIKV DNA vaccination with pMCE321 could elicit cellular as well as humoral responses, characterized by the elicitation of nAb responses. Four rhesus macaques were vaccinated with pMCE321 DNA delivered by in vivo EP. As negative controls, three monkeys were vaccinated with the pVax1 control vector. The animals were then monitored for the development of CHIKV Envelope-specific CD8+ T-lymphocyte and nAb responses. Two weeks after the fifth DNA immunization, cells and serum were collected and tested for immunogenicity. Three of the four CHIKV pMCE321 DNA-immunized monkeys had detectable envelope -specific functional CTL activity as measured by IFN-γ ELIspot (Fig. 6A). The control plasmid-immunized monkeys remained negative throughout the course of the experiment. Furthermore, of the four monkeys immunized with the pMCE321 DNA vaccine, all four monkeys developed nAb titers. These averaged 570 and ranged from 80-1,280 titers (Fig. 6B).
We next compared macaque nAb titers with those observed in human patient convalescent sera. Understanding the correlates of immunity during CHIKV infection is likely important for the rational design and development of an effective vaccine. While it is clear that a nAb response appears to be critical for protection against CHIKV infection, as with many infectious diseases like influenza and hepatitis, the levels required to induce sterilizing immunity or protection from disease-related morbidity are currently unknown. Therefore, we tested serum samples from CHIKV-infected individuals to measure levels of nAb activity. Among the thirty patients tested in this study, sixteen patients (53%) showed nAb titers to CHIKV, (titers above 20 titers were considered as positive; [41]), ranging from 40–640 titers (Fig. 7A). Further, the presence of HI Abs to arboviruses including CHIKV was observed in seroepidemiological studies [18], [32]. However the importance of HI during active CHIKV infection is not well understood. During the recent outbreak investigation, HI Abs were observed in eighteen (60%) of the thirty patients tested in the study. The HI titers varied from 20-1,280 titers (Fig 7B). Indeed, the HI titer also directly correlated with the levels of nAb in CHIKV infected patients (r = 0.9424; p<0.001) (Fig. 7C), suggesting that the neutralizing and HI antibodies to CHIKV correlate with the ability of the host to clear the infectious virus during the course of natural infection. The neutralization titer was defined as the highest dilution of serum that prevented virus propagation as determined by CPE (Fig. 7D). Interestingly, the post-infection nAb titers in convalescent humans were in the same range as the vaccine- induced titers we observed in macaques suggesting its value as a potential model for CHIKV vaccine development.
The mouse and primate immunological data taken together demonstrate that vaccination with pMCE321 induces a strong CD8+ T cell-mediated cellular response as well as a humoral response (nAb and HI titers) capable of protecting the animals against a lethal challenge. In convalescent human samples we report the presence of significant nAb titers. Taken together, the data is suggestive of the role of antibody responses in protecting against CHIKV.
CHIKV is an emerging pathogen and an important public health concern [2], [5], [12], [13], [49]. Since no licensed vaccine or treatment is available for the pathogen, there is an urgent need for an effective vaccine [30]. In this report, we describe the development of a DNA vaccine construct from our laboratory that expresses three of the CHIKV envelope proteins (E3, E2, and E1), is immunogenic in mice and nonhuman primates, provides protection in mice, and drives neutralizing titers in primates similar to those observed in human patient convalescent sera.
During the recent outbreak in India, sera were obtained from human patients with suspected CHIKV disease. All patients were from outbreak areas and a total of thirty sera were randomly selected and included in this study. However, the clinical picture reported herein had a different pattern with respect to previous reports [50]; this cohort had less severe disease symptoms and the reasons attributed to this could be multifactorial, such as the time of sample collection, the magnitute and types of immune responses mounted by these individuals, and the presence of pre-existing immunity in the outbreak area. RT-PCR confirmed the laboratory diagnosis and most of the patients in the study group were found to be positive, additionally having neutralizing/HI Abs to CHIKV. Virus isolation was successfully accomplished from the serum of a febrile patient and the virus isolate was identified and confirmed as CHIKV by RT-PCR amplifying the partial region of the E2 gene. Further, phylogenetic tree analysis with this sequence revealed a similarity to the Chikungunya strain Ross.
The importance of nAb against a viral infection has been well established and was reported earlier in a recent CHIKV study and also an RRV study, a virus that is similar to CHIKV [32], [51]. Viral clearance was associated with the rapid induction of nAb in the acute phase of infection and loss of nAb after recovery from infection [52]. The same concepts may also extend to immunity to CHIKV infection. For example, previous reports in populations with high levels of nAb against CHIKV showed low infection rates, possibly due to protective immunity due to prior exposures, thus resulting in subsequent protective immunity [2], [17], [48]. For instance, a study in northern Malaysia found that 35% of adults had nAbs against CHIKV even though there were no reports of a CHIKV outbreak in Malaysia, during this period of time. Furthermore, serologic surveys in India not linked to a recognized outbreak found a prevalence of 4.4% in the Calcutta metropolis and 15.3% in Andaman and Nicobar Island [7], [16], [53]. Hence, it is likely that nAb aid in the reduction of symptoms either by aiding in clearing the virus or by preventing the pathological damage caused by the virus. In the current study, the range of nAb titers observed in the study cohort ranged between 40 and 640. Similarly, the HI titers induced in patients were in parallel with the neutralizing titer with HI titers ranging from 20 to 1,260.
The findings of significant nAb titers in human patient sera during active infection encouraged us to compare the capacity of our CHIKV DNA vaccine to generate nAb. Our previous CHIKV DNA vaccine consisting of the co-delivery of two different plasmids, pCHIKV-E1 and pCHIKV-E2, was capable of inducing levels of neutralizing and HI titers; pCHIKV-E2 induced the production of nAb and pCHIKV-E1 elicited HI Abs response in immunized mice. These findings led us to combine the predominant genes that constitute the entire CHIKV envelope in a single envelope construct (pMCE321) for vaccination. We envisioned such an approach may lead to simultaneously increased cellular responses and humoral responses and likely inducing both neutralizing and hemagglutination inhibition Abs. Indeed, the novel envelope construct pMCE321 vaccine drove the cellular response to both E1 and E2 glycoproteins and the magnitude of this response was higher than that seen with individual E1/E2 gene constructs. Moreover, the humoral responses to pMCE321 were also increased when compared to that of the individual pCHIKV-E1 and pCHIKV-E2 constructs.
To assess the protective efficacy of this novel vaccine construct capable of generating both strong cellular and humoral immunity, we conducted a CHIKV challenge study in mice and observed that the PC-08 strain of virus was pathogenic in mice. Specifically, we observed a reduction in body weight, lethargy, hind limb weakness and high levels of proinflammatory cytokines such as IL-1β, TNF-α and IL-6 [31], [40]. Following challenge with CHIKV, all of the naïve mice showed severe weight loss from 1-3 days p.i. and showed clinical symptoms like lethargy and hind limb weakness by day 6. However, none of the envelope-based vaccine-immunized mice showed clinical symptoms as pronounced as in the unvaccinated mice. While control mice continued to lose weight over the course of the study and had to be euthanized, the Env-immunized mice rapidly regained weight after the initial 3 days and returned to their normal, pre-challenge state. Overall, these data showed that the new envelope-based vaccine construct pMCE321 was highly effective at protecting against morbidity and mortality in this model. Furthermore, our study also demonstrated an inverse relationship between vaccination and the resulting viremia. Interestingly, we observed a reduction in the amount of virus in the vaccinated mice compared to the unvaccinated mice post-challenge. Similarly, the histopathological evaluation of tissues from the brain, liver, heart, lung and kidney in immunized mice showed minimal or no damage when compared to naïve infected mice. Furthermore, Capsid-immunized mice exhibited symptoms of morbidity as well as mortality likely due to the lack of induction of nAb responses. This protective efficacy may likely be attributed to the high titers of nAb produced in the vaccinated animals similar to the findings in humans where high neutralizing titers have been correlated to better disease prognosis and protection [15], [17], [32], [54].
Due to the induction of strong immunity and protection in mice, we next wanted to assess whether the envelope DNA vaccine (pMCE321) was immunogenic in nonhuman primates. Four rhesus animals were immunized and the cellular and humoral responses were measured. Similar to the immune responses observed in vaccinated mice, CHIKV envelope-specific T cell responses were induced as well as nAb responses in the nonhuman primate cohort. Importantly, the range of neutralization titers observed in these animals was similar to the levels observed in humans during active CHIKV infection. These data demonstrate that the pMCE321 DNA vaccine is immunogenic in nonhuman primates, and was capable of producing titers of nAb which are thought to be protective in humans against disease.
In summary, there are several important findings in this manuscript. We report the isolation of a new isolate of CHIKV from the southern regions of India which we have named PC-08. This isolate is cytopathic in several cell lines as well as primary immune cells. The virus can induce pathogenesis in a mouse challenge model through i.n. challenge and thus should provide a useful in vivo model for further study. Furthermore, this viral stock allowed us to scale up a Neutralization assay for CHIKV study. We also report development of a novel single-plasmid envelope-based DNA vaccine; pMCE321 is protective in the mouse challenge model and drives relevant titers of nAb in a macaque model system. Further study of this novel vaccine and protective immunity is warranted.
|
10.1371/journal.ppat.1002219 | Genomic Insights into the Origin of Parasitism in the Emerging Plant Pathogen Bursaphelenchus xylophilus | Bursaphelenchus xylophilus is the nematode responsible for a devastating epidemic of pine wilt disease in Asia and Europe, and represents a recent, independent origin of plant parasitism in nematodes, ecologically and taxonomically distinct from other nematodes for which genomic data is available. As well as being an important pathogen, the B. xylophilus genome thus provides a unique opportunity to study the evolution and mechanism of plant parasitism. Here, we present a high-quality draft genome sequence from an inbred line of B. xylophilus, and use this to investigate the biological basis of its complex ecology which combines fungal feeding, plant parasitic and insect-associated stages. We focus particularly on putative parasitism genes as well as those linked to other key biological processes and demonstrate that B. xylophilus is well endowed with RNA interference effectors, peptidergic neurotransmitters (including the first description of ins genes in a parasite) stress response and developmental genes and has a contracted set of chemosensory receptors. B. xylophilus has the largest number of digestive proteases known for any nematode and displays expanded families of lysosome pathway genes, ABC transporters and cytochrome P450 pathway genes. This expansion in digestive and detoxification proteins may reflect the unusual diversity in foods it exploits and environments it encounters during its life cycle. In addition, B. xylophilus possesses a unique complement of plant cell wall modifying proteins acquired by horizontal gene transfer, underscoring the impact of this process on the evolution of plant parasitism by nematodes. Together with the lack of proteins homologous to effectors from other plant parasitic nematodes, this confirms the distinctive molecular basis of plant parasitism in the Bursaphelenchus lineage. The genome sequence of B. xylophilus adds to the diversity of genomic data for nematodes, and will be an important resource in understanding the biology of this unusual parasite.
| Bursaphelenchus xylophilus is an important plant pathogen, responsible for an epidemic of pine wilt disease in Asia and Europe. B. xylophilus has acquired the ability to parasitise plants independently from other economically important nematodes and has a complex life cycle that includes fungal feeding and a stage associated with an insect, as well as plant parasitism. We have sequenced the genome of B. xylophilus and used it as a resource to understand disease mechanisms and the biological basis of its complex ecology. The ability to break down cellulose, the major component of the plant cell wall, is a major problem for plant parasitic nematodes as few animals can produce the required enzymes (cellulases). Previous work has shown that other plant parasitic nematodes have acquired cellulases from bacteria but we show that all Bursaphelenchus cellulases were most likely acquired independently from fungi. We also describe a complex set of genes encoding enzymes that can break down proteins and other molecules, perhaps reflecting the range of organisms with which B. xylophilus interacts during its life cycle. The genome sequence of Bursaphelenchus represents an important step forward in understanding its biology, and will contribute to efforts to control the devastating disease it causes.
| The nematode Caenorhabditis elegans was the first multicellular organism for which a complete genome sequence was available, and subsequent genomics research on a wider range of nematodes has provided information on many important biological processes and is underpinned by the information developed for C. elegans [1]. While C. elegans is a free-living bacterial feeder, nematodes exhibit a wide range of ecological interactions, including important parasites of humans and livestock that have huge agricultural and medical impacts [2].
Plant parasitic nematodes cause damage to crops globally. Within the Nematoda, the ability to parasitise plants has evolved independently on several occasions [3] and nematodes use a wide range of strategies to parasitise plants. Some nematodes are migratory ectoparasites which remain outside the roots and have a very limited interaction with their hosts. Migratory endoparasitic nematodes invade their host and cause extensive damage as they move through the host and feed. Sedentary endoparasitic nematodes, including the cyst nematodes and root knot nematodes, have highly complex biotrophic interactions with their hosts. These are the most damaging plant nematodes and consequently genomes for two species of root knot nematode have been reported [4], [5] with others in progress for cyst nematodes. Currently there is no genome sequence for any migratory endoparasitic nematode.
The pine wood nematode Bursaphelenchus xylophilus is a migratory endoparasite that causes severe damage to forestry and forest ecosystems (reviewed in [6]). B. xylophilus is native to North America such that trees there have evolved tolerance or resistance to the pathogen. However, at the start of the 20th Century it was introduced into Japan and has subsequently spread to other countries in Asia where no natural resistance is present, causing huge damage in an on-going epidemic of pine wilt disease. Despite global quarantine efforts B. xylophilus was recently introduced into Portugal [7] and has now also spread to Spain.
Most species within the Bursaphelenchus genus, including the closest relatives of B. xylophilus [8], are fungal feeders that are transmitted by vector insects only to dead or dying trees during oviposition. B. xylophilus and the few other pathogenic species described to date are unique in their capacity to feed on live trees as well as on the fungi that colonise dead or dying trees, so these species may represent a relatively recent, independent origin of plant parasitism. The nematode is a member of the Aphelenchoididae and belongs to clade 10 [3] while most other major plant parasites including Meloidogyne species belong to clade12 [3] (Figure 1). The life cycle of B. xylophilus is summarised in Figure 2 and the infection and disease process has been reviewed by Mamiya [9] and by Jones et al. [6].
Little was known about the molecular basis of the interactions between B. xylophilus and its host plants. A series of advances have been made in the last few years, underpinned by a relatively small scale Expressed Sequence Tag (EST) project on this nematode [10]. Genes involved in parasitism were identified and characterised, including those encoding a wide range of plant cell wall degrading or modifying proteins [11]–[13]. As in other plant parasitic nematodes, it is now clear that horizontal gene transfer has played an important role in the evolution of plant parasitism in B. xylophilus [14].
In order to shed further light on the mechanisms of parasitism used by B. xylophilus and to investigate the genetic and genomic factors involved in the evolution of parasitism, we have produced a high quality genome sequence from an inbred line of this nematode. We describe the assembly and initial annotation and characterisiation of the genome sequence, then interrogate this dataset to identify genes involved in key biological processes, including those associated with chemosensation, neurotransmission, alimentation, stress-responses and development. Further, we identify genes potentially important in functional genetics such as the RNAi pathway and putative control targets such as G-protein coupled receptors (GPCRs), peptidases and neuropeptide genes. This genome sequence allows a comparison of genes involved in plant parasitism across nematode clades and expands our knowledge of the role played by horizontal gene transfer in the evolution of plant parasitism by nematodes.
The B. xylophilus Ka4 population, which originated in Ibaraki prefecture Japan and has been maintained for over 15 years in the Nematology Lab in FFPRI was previously used to generate biological material for ESTs. The Ka4C1 inbred line was established by sister-brother mating over 8 generations from the Ka4 population and was used to generate of material for genome sequencing.
Nematodes were cultivated for 10 days on Botrytis cinerea grown on autoclaved barley grains with antibiotics (100 µg/ml streptomycin and 25 µg/ml chloramphenicol). The nematodes were collected using a modified Baermann funnel technique for 3 h at 25°C and cleaned by sucrose flotation [15] followed by three rinses in 1x PBST. The cleaned nematodes were incubated in 1x PBST containing antibiotics (100 µg/ml streptomycin and 25 µg/ml chloramphenicol) at 23°C for 8 hours to allow nematodes to digest fungal residues remaining in their guts before use. Genomic DNA was extracted from nematodes using GenomeTip-100G (Qiagen) following the manufacturer's instructions.
Poly-(A) + RNA was extracted from mixed-stage nematodes or fourth-stage dispersal larva (DL4 or LIV) collected from the vector insect beetle as described previously [10] and used for the construction of EST libraries.
To determine the B. xylophilus chromosome number, chromosomes were observed in early embryos by DAPI staining and confocal laser-scanning microscopy [16]. The genome size of B. xylophilus was estimated using real time PCR as described in Welhen et al [17]. Translation elongation factor 1 alpha (genbank no GU130155) was used as a reference. DNA concentration was calculated using Qubit (Invitrogen) and the real time PCR reaction was performed using the StepOnePlus system (Applied BioSystems) with SYBR Green I. These protocols can be extremely sensitive and consequently the experiments were repeated in triplicate.
Sequence data from 454 FLX and Illumina GAI were assembled using the Newbler de novo assembler (version 2.3) and Velvet assembler (version 1.0.12) [18] respectively. The result from each assembly was combined using Minimus2 (sourceforge.net/projects/amos/) and contigs supported by both assemblies were used as fake unpaired capillary reads in the subsequent Newbler assembly. The resulting assembly was improved using three different methods: AbacasII [19] to merge small contigs; IMAGE [20] to iteratively map and assemble short reads to close gaps; and iCORN [21] to iteratively correct single base substitutions and small indels.
As an indirect measure of completeness of the assembly, a search for orthologues was performed by CEGMA (ver. 2.0) software using CEGs (core eukaryotic genes); a set of 248 extremely highly conserved genes thought to be present in almost all eukaryotes in a reduced number of paralogues [22].
EST clustering was performed using PartiGene (ver. 3.0), a software pipeline designed to analyze and organize EST data sets [23]. Briefly, EST sequences were clustered into groups (putative genes) on the basis of sequence similarity and then clusters were assembled to yield consensus sequences using Phrap (P. Green, unpublished). Mapping of individual ESTs or clustered ESTs to the genome assembly was performed using PASA software [24].
EST data generated in this study and previously obtained by capillary sequencing [10] were used to predict genes in the assembly. The total number of ESTs used was 82,100.
A reference dataset of 565 B. xylophilus protein encoding genes was manually curated from EST clusters and predictions of highly conserved genes using CEGMA [25]. 365 of these were used to train ab-initio gene predictors Augustus [26] and SNAP [27] and 200 were used to evaluate accuracy of the predictions. EVidenceModeler [24] was used to combine all predictions from gene predictors Augustus, SNAP and GeneMark.hmm [28], EST mapping data from PASA [24] and protein homology data against the Pfam database using GeneWise2 [29]. Gene prediction accuracy was computed at the level of nucleotides, exons and complete genes on 200 manually-curated gene models (Figure S1) as described previously [30].
Initial functional annotation was performed using InterProScan to search against the InterPro protein family database, which included PROSITE, PRINTS, Pfam, ProDom, SMART, TIGRFAMs, PIR SuperFamily and SUPERFAMILY [31]. The latest version Pfam search (ver. 24.0) [32], which uses HMMER3 and is more sensitive than the previous version packed in InterProScan, was also performed independently for B. xylophilus genome. Gene Ontology annotation was derived using Blast2GO software [33] based on the BLAST match against NCBI non-redundant (NR) proteins with an E-value cutoff of 1e-10 and InterProScan results.
Assignments to conserved positions in metabolic and regulatory pathways were performed using KOBAS software [34] based on the KEGG annotation resource [35]. KEGG genes and KO term annotations were assigned based on similarity searches with a 1e-5 E-value cutoff and a rank cutoff 5. Significantly enriched pathway terms between two organisms were identified by frequencies of terms with chi-square and FDR correlation tests.
To study the evolution of gene families across nematodes within the order Rhabditida, we used the predicted protein sets from all 9 genomes available in WormBase release WS221 (www.wormbase.org) – C. elegans, C. brenneri, C. briggsae, C. japonicum, C. remaneri, Meloidogyne hapla, M. incognita, Pristionchus pacificus and Brugia malayi, together with predicted proteins of B. xylophilus. Version 2.0 of the OrthoMCL pipeline [36] was used to cluster proteins into families of orthologous genes, with default settings and the BLAST parameters recommended in the OrthoMCL documentation. We reconstructed the evolution of gene families on a phylogeny for these 10 species, based on aligning amino-acid sequences from single-copy gene families using Muscle v3.6 [37], and constructing coding-sequence nucleotide alignments based on these. Phylogenetic inference was performed using BEAST v.1.6.1 [38] from 10 million Markov Chain Monte Carlo (MCMC) generations under a strict clock model using the SRD [39] model for each gene partition. Three independent MCMC runs converged to the same posterior probability. Birth and death of gene families was inferred under Dollo parsimony using the Dollop program from v3.69 of the Phylip package [40].
To look for potential horizontal gene transfers specifically into the B. xylophilus lineage, we used BLASTP to compare predicted B. xylophilus protein sequences against the NCBI NR database, producing a candidate set of laterally transferred genes that either had no significant BLAST hit (E-value≤10−5) to any nematode sequence or had significant hits only to genes from Aphelenchoidoidea and no other nematode. For each of these candidates, amino acid sequence data was extracted from the NCBI database, aligned using Muscle v3.6 and ML phylogenetic trees estimated using the best-fitting model under the AICc criterion in RaxML v7.2.8. [41]. Clade support was estimated using 100 non-parametric bootstrap replicates in RaxML, and approximately unbiased (AU) statistical tests of tree topology were performed in Consel v1.19 [42].
The detection and annotation of ncRNA molecules was performed using the LeARN pipeline [43]. This pipeline contains four independent methods: tRNAscanSE for transfer RNA (tRNA) gene detection, NCBI-BLASTN versus a ribosomal RNA (rRNA) sequence database, the Rfam database (release 9.0) to detect common ncRNA families and a mirfold-based pipeline using the mirBase library as a source of micro RNA (miRNA) candidates.
Transposable elements (TEs) in the assembly were identified using two approaches. The first stage consisted of de novo identification of repeat families in the assembly based on signatures of transposable elements and assuming fragments of TEs are present throughout the genome. Long terminal repeat (LTR) retrotransposons were identified using LTRharvest which searches for two near-identical copies of an LTR flanked by target site duplications that are close to each other. We also used RepeatModeler (http://www.repeatmasker.org/RepeatModeler.html) which aims to construct repeat consensus from two de novo detection programs (RepeatScout and RECON). Repeats present at less than 10 copies in the genome or that were less than 100 bp were excluded from further analysis. The second approach used homology searching of the assembly sequence against curated TEs using TransposonPSI (http://transposonpsi.sourceforge.net/). UCLUST was used to cluster the candidate sequences (with 80% identity) and create a non-redundant library of repeat consensus sequences.
The annotation of repeat candidates involves a search against RepBase and NCBI non-redundant library. Some of these candidates that have some annotations available from program output (for example, from TransposonPSI) were further checked this way. Manual curation of the candidates was carried out to determine coding regions on intact TEs that are potentially active. We found the majority of unannotated elements contained no ORFs longer than 30 bp nor had any significant matches to repetitive elements in the databases.
RepeatMasker (v3.2.8) was used to calculate the distribution of each repeat and its abundance. Custom perl scripts were used to choose the best match from overlapping matches in RepeatMasker output to avoid calculating the same region twice or more when considering repeat content of the genome.
The CAZymes Analysis Toolkit (CAT) [44] was used to detect B. xylophilus carbohydrate active enzymes (CAZymes) based on the CAZy database. An annotation method “based on association rules between CAZy families and Pfam domains” was used with an E-value threshold of 0.01, a bitscore threshold of 55 and rule support level 40. The annotation was supplemented and confirmed manually using BLAST search similarities and protein length matches. Expansin-like genes were detected by BLAST search using core modules of known expansin proteins as queries. Putative functions of the proteins were predicted by similarity to known protein modules and presence of catalytic sites using BLASTP search against NCBI's Conserved Domain Database service and InterProScan (www.ebi.ac.uk/Tools/InterProScan).
The MEROPS server was used to identify B. xylophilus putative peptidases. The peptidase candidates were derived from MEROPS batch BLAST [45]. The candidates were manually examined in terms of similarity (E-value cutoff 1e-10) to MEROPS proteins and presence of all catalytic sites.
BLASTP was used to search for B. xylophilus homologues of effectors from M. incognita [46], Heterodera glycines [47] and Globodera pallida [48]. An E-value cutoff of 1e-5 was used to identify significant matches. In addition, candidate effectors were sought from the B. xylophilus protein set using a bioinformatic approach. Secreted proteins were identified as those having a potential signal peptide at their N-terminus predicted by SignalP 3.0 [49] and no transmembrane domain within the mature peptide as predicted using TMHMM 2.0 [50]. Novel candidate effectors within this secreted protein dataset were identified as those with no BLASTP matches against the NR protein database (E-value cutoff 0.001).
Orthologues of C. elegans genes necessary for unequal cell divisions were identified from the B. xylophilus protein set by BLASTP search using protein sequences retrieved from WormBase as queries. Their structures were confirmed manually using NCBI and WormBase BLAST.
Seventy-eight proteins known to be involved in core aspects of the C. elegans RNAi pathway were identified from the literature. Protein sequences were retrieved from WormBase (release WS206) and used as queries in TBLASTN and BLASTP searches against the predicted protein and contig databases. All primary BLAST hits returning with a bitscore ≥40 and an E-value≤0.01 were manually translated to amino acid sequence in six reading frames (www.expasy.ch/tools/dna.html), and analysed for identity domain structure by BLASTP (through NCBI's Conserved Domain Database service) and InterProScan. The appropriate reading frame in each case (determined empirically on a case by case basis, but usually that with the largest uninterrupted open reading frame) was then subjected to reciprocal TBLASTN and BLASTP against the C. elegans non-redundant nucleotide and protein databases on the NCBI BLAST server (http://www.ncbi.nlm.nih.gov/BLAST), using default settings. The identity of the top-scoring reciprocal BLAST hit was accepted as identity of the relevant primary hit, as long as that identity was also supported by domain structure analysis.
The predicted B. xylophilus AGOs were analysed further through annotation of conserved RNase-H-like catalytic residue sites and the MID sub-domain which was then used for a further BLAST search against the C. elegans nr protein set (NCBI). The MID domain is highly conserved in functionally comparable AGOs [51], [52].
Neuropeptide genes were identified using the BLAST search tool available through the genome consortium website. FLP and NLP search strings were constructed from orthologous flp and nlp transcript sequences taken from C. elegans in the first instance, or another plant-parasitic nematode where a C. elegans orthologue was unavailable. Search strings were constructed by concatenating the mature peptides (including basic cleavage sites) encoded by each orthologue. Where this concatenation resulted in a search string of less than 20 characters, the string was repeated end-to-end at least once. Search strings were entered into BLASTP searches of the predicted proteins, and tBLASTn searches of the genome scaffolds, with E-value set to 1000. INS search strings were created by concatenating functionally conserved A and B peptide regions from the C. elegans ins gene complement, in addition to a series of mammalian and molluscan orthologues. These were used to search the predicted protein set of B. xylophilus, using a higher E-value threshold of 1,000,000. The BLAST returns were annotated for both A and B peptides, which were isolated, concatenated and used as BLAST search strings against the C. elegans nr protein set (NCBI) with an increased E-value of 1,000,000. All reciprocal BLAST returns with a Bit score ≥25 and an E-value≤10 were annotated for conserved A and B peptides, and the C. elegans orthologue with the highest similarity to these domains was accepted as identity. Putative INS orthologues which did not meet the reciprocal BLAST criteria, but which closely resemble the structure of INS A and B peptide domains are included as putative variant (Var) INS orthologues (≤25, ≥20 bit score; E-value ≥10, ≤20). All hits were analysed by eye for the presence of neuropeptide precursor sequences, dibasic cleavage sites, and analysed for the presence of secretory signal peptides using SignalP 3.0 [49].
Orthologues of C. elegans genes were identified from B. xylophilus protein set using BLASTP. The families of GSTs, UGTs, CYPs, and ABC transporters were identified with InterProScan. Their structures were confirmed manually with BLAST on NCBI and WormBase.
Protein sequences known to be involved in dauer formation and maintenance were retrieved from WormBase and used as search strings in a series of tBLASTn and BLASTP searches against B. xylophilus genome and protein sequences. An E-value cutoff of 1e-10 was used to identify significant matches.
Potential chemoreceptors included 7-transmembrane, G-protein coupled receptors (GPCRs), e. g. Str, Sra and Srg gene superfamilies in C. elegans [53], and other receptors with transmembrane structures, e. g., gustatory receptors (GURs, orthologues of insect gustatory receptors) in C. elegans, ionotropic glutamate receptors (IRs) in Drosophila melanogaster [54] and transient receptor potential (TRP) channels [55], [56]. Other putative GPCRs including those for neurotransmitters (e.g., bioorganic amines) and other signal transduction pathways were also searched. BLASTP and InterProScan were used to search for B. xylophilus orthologues of these proteins. All primary BLASTP hits returning with an E-value ≤ 0.0001 and coverage ratio ≥ 0.7 (apart from insect GURs, which included only hits of very low similarity to nematode GURs) were analysed for identity and domain structure by BLASTP (NCBI's Conserved Domain Database service), WormBase (WS221) and TMHMM 2.0 [50].
Molecular phylogenetic trees of serpentine receptor and gustatory receptor proteins were built by maximum likelihood method using MEGA5 with the JTT matrix-based model [57]. A few receptors were removed in some families based on very long branch lengths in a preliminary maximum likelihood tree. A maximum likelihood tree with culled proteins was drawn to a scale, with branch lengths measured in the number of substitutions per site. All positions containing gaps and missing data were eliminated.
The contigs resulting from Bursaphelenchus xylophilus assembly were deposited in the EMBL/Genbank/DDBJ databases under accession numbers CADV01000001-CADV01010432.
Our assembly strategy assembled the 6 pairs of nuclear chromosomes (Figure S2) into 1,231 scaffolds, totaling 74.5 Mb, with half of these nucleotides present in scaffolds of at least 1.16 Mb (Table 1). The size of the assembly is in good agreement with our experimental estimate of 69.0±5.5 Mb (see Text S1, Table S1 in Text S2). The genome size of 74.5 Mb is smaller than that of C. elegans and other published nematode genomes except that of Meloidogyne hapla. The GC content of the genome was 40.4%, higher than that of other nematodes except for Pristionchus pacificus. Most of the mitochondrial genome was assembled in a single 13,410 bp scaffold, and shows a similar gene content and organization to that of C. elegans (Figure S3). Analysis of conserved eukaryotic genes (CEGs) showed that 96.77% and 97.98% of CEGs were present as full or partial genes respectively, with an average of 1.08―1.09 genes per CEG (Table 1), suggesting high completeness of the assembly.
The assembly is approximately 22% repetitive, of which only around 1.3% had characteristics of transposable elements (TEs, Table 2). A complete set of tRNA (Table S2 in Text S2) and rRNA genes were found in the genome. In common with other parasitic nematodes, including B. malayi and M. incognita [58], SL2 trans-splicing appears not to exist in Bursaphelenchus, but we find 25 SL1-like sequences, found in the same tandem repeats as the 5S rRNAs, as in C. elegans.
Analysis of chromosomal rearrangements between B. xylophilus and C. elegans identified a similar pattern of macrosynteny to that found between the more distantly related Trichinella spiralis and C. elegans [59]. Large B. xylophilus scaffolds largely contain genes orthologous to those from a single C. elegans chromosome. These genes are, however, interspersed by genes orthologous to those from other chromosomes (see Text S1, Figure 3).
A total of 18,074 protein coding genes were predicted in the assembly (Table 1). This is fewer than the 20,416 in C. elegans (WormBase WS221) and 19,212 in M. incognita, although it is higher than the number for M. hapla. The average protein length is similar to that of other nematodes, but B. xylophilus displays the largest average exon size (289 bp) and the smallest average number of exons per gene (4.5). The Bursaphelenchus genome shows a number of characteristics of compact parasite genomes, for example having relatively few, short introns like M. hapla, but has a similar repetitive element content to other published nematode genomes, and is overall only slightly smaller.
Our automated annotation of the B. xylophilus proteome assigned some functional information to a total of 12,483 proteins (69%) (see Text S1 Figure S4). The top 20 Pfam hits in B. xylophilus are shown in Figure 4, and compared with hits in C. elegans. As part of our annotation approach, B. xylophilus proteins were mapped to pathways defined by KEGG, and pathways that are under- and over-represented in this genome compared to C. elegans are shown in Table S3 in Text S2 and are discussed in sections focusing on particular biological features below.
The combined predicted proteins from B. xylophilus and 9 other nematode species were grouped into 27,547 families of orthologues and an additional 51,942 singleton proteins. We used a molecular phylogeny based on single-copy gene families to reconstruct the distribution and evolutionary dynamics of these gene families (Figure 5), and find that the B. xylophilus genome has been relatively conserved over the long divergence from other plant parasitic nematodes. The pattern of sharing of gene families between genomes (Figure 6) shows little obvious phylogenetic pattern, but identifies relatively small numbers of genes - 202 genes shared by the two plant parasitic genomes, and 144 genes shared by Pristionchus and Bursaphelenchus that both show a close association with insects during their lifecycle – that could be implicated in these particular specializations.
The plant cell wall is the primary barrier faced by most plant parasites and the production of enzymes able to break down this cell wall is thus of critical importance. A summary of the carbohydrate active enzymes (CAZymes) and expansin-like proteins which may modify plant cell walls detected in B. xylophilus and other nematodes is shown in Table 3. B. xylophilus contains 34 putative plant cell wall modifying enzymes but compared to other plant parasitic nematodes its composition is unique. Most interestingly, glycoside hydrolase family 45 (GH45) cellulases are present only in B. xylophilus. Other plant parasitic nematodes have GH5 proteins that degrade cellulose but no such genes are present in B. xylophilus. In addition, GH30 xylanases, GH43 arabinases and GH28 pectinases were also absent in B. xylophilus.
Cell wall degrading CAZymes found in plant parasitic nematodes are thought to have been acquired via horizontal gene transfer (HGT) because similar genes are absent in almost all other nematodes and because they are most similar to genes from bacteria or fungi [60], [61]. GH5 cellulases have been found in many plant parasitic Tylenchoidea including Meloidogyne, Globodera and Heterodera. Recently, the genome sequence of P. pacificus revealed that this nematode also has GH5 cellulases [62]. However phylogenetic analysis suggested that these cellulases were not closely related to those in the Tylenchida and that they are likely to have been acquired independently from different sources [62]. B. xylophilus GH45 cellulases have not been found in any other nematode genus and are most similar to those from fungi. Thus these genes have been hypothesised to be acquired via HGT from fungi [11]. In a phylogenetic analysis all 11 GH45 proteins found in the B. xylophilus genome are grouped in a highly supported monophyletic group and embedded within a clade of fungal homologues (Figure S5). This supports the idea of HGT from fungi with subsequent duplication within the B. xylophilus genome. Recent analysis has revealed that distribution of GH45 proteins is limited to the genus Bursaphelenchus and its sister genus Aphelenchoides (T Kikuchi unpublished results). The absence of GH5 genes in the B. xylophilus genome and the absence of GH45 proteins in Tylenchida nematodes support these hypotheses and suggest that HGT events have repeatedly played important roles in the evolution of plant parasitism in nematodes.
A systematic evolutionary study of plant cell wall modifying genes in Tylenchoidea concluded that those genes were acquired by multiple HGT events from bacteria closely associated with the ancestors of these nematodes followed by gene duplication [61]. B. xylophilus is closely associated with fungi and it is likely that this feeding strategy is ancestral for this group as most Bursaphelenchus species are solely fungal feeders.
In addition to plant cell wall degrading enzymes, CAZymes are present in B. xylophilus which potentially degrade the fungal cell wall. Chitin is one of the main components of fungal cell walls. Proteins related to chitin degradation were identified in the B. xylophilus genome. In comparison to the two Meloidogyne species, P. pacificus and B. malayi the number of chitin-related CAZymes in B. xylophilus is increased, likely reflecting its fungal feeding activity (Table 4). C. elegans has a much larger number of GH18 proteins than B. xylophilus. This may be because those proteins have been used in C. elegans for specific biological features such as bacterial feeding. Interestingly, six GH16 proteins which may degrade beta-1,3-glucan, another core component of the fungal cell wall, have been identified in the genome while no homologues have been found in other nematodes (Table 4). Because the B. xylophilus GH16 β-1,3-glucanase genes are most similar to those from bacteria, it has been suggested that they were acquired from bacteria which were closely associated with its ancestor [63]. This suggests that HGT processes, similar to those associated with plant cell wall modifying enzymes of other parasitic nematodes, have enhanced the ability of Bursaphelenchus spp. to feed on fungi.
Not all of the carbohydrate-active enzymes over-represented in the B. xylophilus genome are involved in cell-wall degradation – several other CAZyme families are substantially expanded in B. xylophilus compared to other nematodes sequenced to date (Table 4). For example, the genome also has more glycosyl transferases family 43 (GT43) genes than other nematodes. These proteins may have beta-glucuronyltransferase activities, but the reason for the increase in these genes in B. xylophilus remains unclear.
Peptidases (proteases) catalyse the cleavage of peptide bonds within proteins, play important functions in all cellular organisms and are involved in a broad range of biological processes. In nematodes, peptidases play critical roles not only in physiological processes including embryogenesis and cuticle remodeling during larval development but also in parasitic processes such as tissue penetration, digestion of host tissue for nutrition and evasion of the host immune response.
In our analysis 581 peptidase genes were identified in B. xylophilus, which is the largest number in any characterized nematode genome (Table 5), with peptidase families involved in extracellular digestion and lysosomal activities particularly expanded (see Text S1, Table S4 in Text S2). One family of endopeptidases appears to have been acquired by HGT from an ascomycete fungus (Table 6). In addition, B. xylophilus contains an expanded number of GH27 proteins homologous to the gana-1 gene of C. elegans (Table 4), which has α-galactosidase and α-N-acetylgalactosaminidase activities and is localized to lysosomes [64].
The gut granules of intestinal cells in C. elegans are intestine-specific secondary lysosomes, so lysosomal enzymes play important roles in the digestion of food proteins in nematodes. B. xylophilus has an expanded repertoire of peptidases and other digestive enzymes that are either secreted or localised in lysosomes and so may play important roles in food digestion. Genes in the lysosome pathway were the most significantly over-represented in B. xylophilus (Table S3 in Text S2). B. xylophilus uses food sources such as fungi and woody plants that may be difficult to digest and the expansion of digestive peptidases in the nematode is therefore consistent with its unusual life style.
Plant parasitic nematodes produce a variety of secreted proteins that mediate interactions with their hosts – these encompass a variety of functions and include the cell-wall modifying enzymes discussed above. Such proteins have variously been termed “parasitism genes” or “effectors” and encompass any protein secreted by the nematode into the host that manipulates the host to the benefit of the nematode. For example, cyst nematodes produce effectors that mimic plant peptides and which may help initiate the formation of the biotrophic feeding structures induced by these nematodes [65], as well as proteins that suppress host defence responses [66]. We found that the majority of effectors from other plant parasitic nematodes have no homologues in B. xylophilus. Some significant matches to effectors from all three species were found (Table 7). However, the B. xylophilus sequences that matched these effectors, except cell wall degrading enzymes, either did not have predicted signal peptides or, if a signal peptide was predicted, homologues were also present in a wide range of other species including C. elegans and animal parasitic nematodes. Both these lines of evidence suggest that the B. xylophilus homologues identified in this analysis are not true effectors that play a role in parasitism. These findings are consistent with the differing biology of the various nematode groups; root knot and cyst nematodes are biotrophic species whereas B. xylophilus is a migratory endoparasite that does not rely on biotrophy.
There are two possible exceptions. B. xylophilus contains homologues of venom allergen proteins. These proteins are present in all nematodes investigated to date and are thought to be important for the parasitic process of animal and plant parasites (e.g. [67], [68]). Several venom allergen proteins from B. xylophilus have been characterized and are known to be expressed in the oesophageal gland cells [69]. Our HGT analysis identified a putative cystatin, or cystein protease inhibitor, apparently acquired from a bacterium (Table 6). Cystatins are well known as immunomodulatory pathogenicity factors in the animal parasitic filarial nematodes [70], so this protein could potentially play a role in parasite-host interaction in a plant parasitic nematode. However proteins from this family are involved in regulating a variety of endogenous proteinase activities in many cellular roles and, for example, are described as having anti-fungal properties [71], so the function of this protein will require experimental verification. We also identified a total of 923 predicted secreted proteins in the B. xylophilus genome that show no significant similarity to proteins from other species (Dataset S1), representing a pool of candidates that may play a role in the interaction between B. xylophilus and the other organisms with which it interacts. Very few (5) of these sequences produce matches against other plant parasitic nematode ESTs, consistent with previous studies which have shown that secreted proteins of parasitic nematodes often bear a high proportion of novel genes [72].
Detoxification of potentially damaging compounds is an important process for any organism to cope with its environment and may be particularly crucial for parasitic organisms, which come under attack from host responses to infection. In particular, plant parasites must cope with a wide range of secondary metabolites that plants generate in order to protect their tissues [73].
B. xylophilus principally inhabits the resin canals of its pine hosts. The resin to which it is exposed – a complex mixture of compounds, including terpenoids [74] and cyclic aromatic compounds [75] – is likely to have nematocidal activity and, like the detoxification of xenobiotics by C. elegans [76], would be expected to proceed in three distinct phases: (I) the addition of functional groups to molecules, making them more suitable substrates for downstream; (II) the actual detoxification reactions; and (III) efflux. Cytochrome P450s (CYPs) represent the most important group of phase I proteins, and B. xylophilus encodes a similar number of CYPs to that found in C. elegans (Table 8). Of the two main families of phase II detoxification enzymes – the glutathione S-transferases (GSTs) and UDP-glucuronosyl transferases (UGTs), we identified 41 full-length and 26 partial GSTs, and 60 UGTs, similar numbers to those found in C. elegans (Table 8, Figure S6). The final phase of the detoxification process involves ATP-binding cassette (ABC) transporters actively exporting detoxified xenobiotics. A total of 106 ABC transporters were detected in the B. xylophilus genome; this number was about twice that for C. elegans and about three times that for M. incognita (Table 8), suggesting that B. xylophilus is particularly enriched in genes responsible for the efflux of detoxified molecules.
Finally, we investigated genes involved in regulating the detoxification process. In C. elegans, the transcription factor SKN-1 regulates expression of many detoxification enzymes [77], and SKN-1 activity is in turn controlled by a number of different pathways (see Figure S7). In the presence of oxidative stress or electrophilic compounds, SKN-1 induces the expression of many phase II detoxification enzymes. Orthologues of all these regulatory pathways can be identified in B. xylophilus, suggesting that the regulation of xenobiotic degradation may be conserved in nematodes.
There are other signs that Bursaphelenchus may have an unusual repertoire of genes involved in the defence against or utilisation of complex pine tree metabolites – in our KEGG pathway analysis, xenobiotic and drug metabolism through CYPs were among the pathways showing most significant enrichment in gene copy number over C. elegans, confirming that other genes, likely to be involved downstream of the CYP genes themselves, as well as efflux effectors, are notably enriched (Table S3 in Text S2). Furthermore, our search for carbohydrate active enzymes reported a number of genes classified into the GH109 family of glycosyl hydrolases (Table 5) that on closer inspection proved to be most similar (approx 39% identity and E-values<1E-50) to enzymes displaying trans-1,2-dihydrobenzene-1,2-diol dehydrogenase activity, which is involved in the pathway downstream of cytochrome P450 in the metabolism of naphthalene and other polycyclic aromatic hydrocarbons (PAHs) [78] (Table S3 in Text S2). While PAHs, including naphthalene itself, are known to be produced in small quantities by a few plant species [79], they are not known from pines, and it seems more likely that this enzyme is homologous to naphthalene-degrading enzymes but acts on some of the many other aromatic molecules generated by plants [73].
It seems likely that B. xylophilus has a larger number of detoxification enzymes than other plant parasitic nematodes (M. incognita and M. hapla), with similar or expanded repertoires of such genes to those reported for the free-living C. elegans and the necromenic P. pacificus [62] for the various components of the detoxification process. This expansion in detoxification process components may reflect the variety of stressful environments that it encounters during its life cycle, and perhaps the particular challenges of inhabiting living tissues in a plant host that produces diverse secondary toxic metabolites.
B. xylophilus embryos seem to form the anterior-posterior axis quite differently from those of C. elegans as the point of sperm entry becomes the future anterior end of the animal [80]. Surprisingly, however, other early events in B. xylophilus embryos, such as pronuclear meeting and posterior spindle movement followed by the unequal first cell division are quite similar [80]. Therefore, it is informative to compare and contrast the proteins that control these processes in these two species. Orthologues of the majority of C. elegans proteins involved in these processes were identified in B. xylophilus and appear to be highly conserved. However one putative homologue of the serine/threonine kinase protein PAR-1 was quite distinct in B. xylophilus from that in C. elegans in that the former was considerably smaller (467 AA compared to 1,192 AA in C. elegans); the implications of this difference are unknown.
The formation of dauer (or infective) larvae specialized for surviving adverse conditions or for invading host organisms is an important life stage for many nematodes. In B. xylophilus, we identified orthologues of most genes involved in pathways which regulate dauer larva formation and recovery in C. elegans [81] (Table S5 in Text S2). We also identified orthologues of genes involved in C. elegans dauer pheromone synthesis (see Text S1). As C. elegans, adverse conditions trigger B. xylophilus to enter the third-stage dauer larva (DL3 or dispersal third-stage larva LIII) (Figure 2). Pathways that respond to these environmental cues may be more conserved in B. xylophilus than in other parasitic nematodes, most of which use different cues when forming a dauer (infective) stage. In addition to DL3, B. xylophilus has a specialized stage called the fourth-stage dispersal larva (DL4 or LIV). B. xylophilus DL3 develop into the DL4 when stimulated by the presence of the vector beetle Monochamus alternatus and become ready to board the vector [82], [83]. Previous studies showed that several novel genes are expressed specifically in the DL4 nematodes [10], suggesting that B. xylophilus responds to different environmental stimuli, and likely uses distinct pathways and proteins to control this part of the lifecycle.
Nematode neuropeptides are encoded on flp (FMRFamide-like peptide), nlp (neuropeptide-like protein) or ins (insulin-like peptide) genes [84], [85]. Diverse arrays of neuropeptides exist within every nematode species that has been studied, and neuropeptide receptors are promising potential drug targets [86]. The complexity of this peptidergic signalling environment likely aids behavioural diversity and plasticity in spite of the structurally simple nematode nervous system. We find B. xylophilus's flp and nlp gene complements are typical of those seen in other parasitic nematode species [87], [88], although the absence of two flp genes - flp-30 and -31 - is noteworthy (Table S6 in Text S2), as these genes have previously been considered unique to Meloidogyne spp. [4], [87]. Their absence from B. xylophilus suggests they may associate with an obligate parasitic lifestyle. The discovery of seven ins-like orthologues in the B. xylophilus genome is significant as the first description of nematode INS-like peptides outside C. elegans (Table S6 in Text S2, Dataset S2).
Chemoreception governs essential aspects of the life of many invertebrates, including the search for mates and hosts and the timing of critical steps in their life cycles. Chemoreceptors constitute one interface between the animal and its world, and could be expected to exhibit local adaptations to the specific chemosensory niche of each organism. The main group of putative chemosensory genes in nematodes is represented by serpentine receptors, which are GPCRs, include a large number of families and are also important drug targets. We find representatives of most C. elegans serpentine receptor families in the B. xylophilus genome, but many represent specific expansions, so the two species have related but largely distinct repertoires. The total number of serpentine receptor genes identified from B. xylophilus represents only 10–20% of the number found in C. elegans [53] but 35–45 times of those in M. hapla [5]. It is unclear whether these striking differences represent a reduced and/or expanded chemosensory systems in various nematodes, or whether additional gene families have been expanded to cover some of the chemosensory spectrum in B. xylophilus and other species. Other chemosensory genes identified include gustatory receptors, GPCR receptors for a range of neurotransmitters that could have chemosensory roles, and members of the ionotropic glutamate receptor family [54], among others (see Text S1, Figure S8–S10).
The B. xylophilus genome encodes more predicted orthologues of C. elegans RNAi pathway effectors (37 of a potential 78) than found in M. incognita (27) and M. hapla (28) (unpublished data). Whilst B. xylophilus has orthologues of eight of the nine small RNA biosynthetic protein-encoding genes considered, dsRNA uptake and spreading genes are not well represented, e.g. no sid gene orthologues were identified with rsd-3 the only representative gene identified. RNA-dependent RNA polymerases (RdRps) are expanded relative to C. elegans, with four ego-1-, two rrf-1-, and three rrf-3-like orthologues. Sixteen Argonaute (AGO) genes were identified relative to the 27 of C. elegans and, for some of these, there was divergence within the catalytic and RNA-binding MID subdomains; other RNA-induced silencing complex (RISC) cofactors were identified (ain-1, tsn-1 and vig-1). Whilst the short interfering RNA (siRNA) inhibitor eri-1 was not found, microRNA (miRNA) inhibitors (somi-1; xrn-2) were identified. Nuclear effectors were reasonably well represented such that most of the components of a functional RNAi pathway were identified within the B. xylophilus genome. See Text S1 and Table S7-S11 in Text S2 for details.
In addition to its status as an economically important plant pathogen, B. xylophilus is remarkable for its unusual biological traits that relate to its complex ecology. During its life cycle it occupies two distinct habitats – an insect and a tree – where it exploits a number of different food sources, including plant tissues and a wide variety of fungi. This adaptability to a number of different niches is reflected in its genome sequence. The presence of a rich repertoire of detoxification enzymes and transporters reflects B. xylophilus' habitat in the resin canals of its host trees, where it is exposed to a cocktail of secondary metabolites, and its elaboration of a narrow subset of carbohydrate metabolizing enzymes reflects it adaptation to break down plant cell walls. The unique complement of genes involved in cellulose degradation, and other catabolic enzymes and the absence of effectors previously known to function at the host-parasite interface, confirms that B. xylophilus has a mode of parasitism that is distinct from other plant parasitic nematodes. This parasitism is mediated by a unique suite of parasitism-related genes, assembled through a combination of gene duplication and horizontal gene transfer. The genome provides strong evidence of multiple independent horizontal gene transfer events and these have shaped the evolution of this group.
Most importantly the genome sequence will act as a foundation for functional studies using a wide range of techniques and will directly inform efforts aimed at controlling this parasite. The identification of genes involved in nematode invasion and feeding from the plant will empower efforts to understand the interaction of B. xylophilus's with its host. One exciting possibility is the potential for genomic information from the hosts of B. xylophilus to facilitate understanding of the host-parasite interaction and associated pathology; host genetics is likely to play a key role in the disease as B. xylophilus is non-pathogenic to American pine species. We have identified genes involved in a range of crucial biological processes, many of which, such as neuropeptides, GPCRs and developmental genes could be viable control targets. The presence of a rich set of RNAi pathway effector genes gives much hope that reverse genetics will underpin future functional genomics efforts in this species. In this way, the genome sequence provides the opportunity to identify and validate putative control targets without the need to rely on C. elegans and make assumptions on conserved functionality/importance between nematodes from different clades.
To our knowledge, only seven nematode genomes have previously been published, and data are available for only a handful more, mostly from the genus Caenorhabditis. Given the breadth of the nematode phylum, genomic information from any new nematode species is an important advance but, B. xylophilus, in particular, is the first species to be sequenced from clade 10, and the first from the order Aphelenchoidea. We hope that with the other imminent nematode genomes being sequenced, B. xylophilus will serve as an important comparator. These data provide a rich resource for those trying to develop novel control strategies directed against B. xylophilus. In addition, the parasite's unusual life cycle makes this genome sequence a unique resource to investigate the association between genome structure and lifestyle, casting new light on the many conserved processes for which the free-living non-parasitic C. elegans remains the pre-eminent model.
|
10.1371/journal.ppat.1002275 | Strain Specific Resistance to Murine Scrapie Associated with a Naturally Occurring Human Prion Protein Polymorphism at Residue 171 | Transmissible spongiform encephalopathies (TSE) or prion diseases are neurodegenerative disorders associated with conversion of normal host prion protein (PrP) to a misfolded, protease-resistant form (PrPres). Genetic variations of prion protein in humans and animals can alter susceptibility to both familial and infectious prion diseases. The N171S PrP polymorphism is found mainly in humans of African descent, but its low incidence has precluded study of its possible influence on prion disease. Similar to previous experiments of others, for laboratory studies we created a transgenic model expressing the mouse PrP homolog, PrP-170S, of human PrP-171S. Since PrP polymorphisms can vary in their effects on different TSE diseases, we tested these mice with four different strains of mouse-adapted scrapie. Whereas 22L and ME7 scrapie strains induced typical clinical disease, neuropathology and accumulation of PrPres in all transgenic mice at 99-128 average days post-inoculation, strains RML and 79A produced clinical disease and PrPres formation in only a small subset of mice at very late times. When mice expressing both PrP-170S and PrP-170N were inoculated with RML scrapie, dominant-negative inhibition of disease did not occur, possibly because interaction of strain RML with PrP-170S was minimal. Surprisingly, in vitro PrP conversion using protein misfolding cyclic amplification (PMCA), did not reproduce the in vivo findings, suggesting that the resistance noted in live mice might be due to factors or conditions not present in vitro. These findings suggest that in vivo conversion of PrP-170S by RML and 79A scrapie strains was slow and inefficient. PrP-170S mice may be an example of the conformational selection model where the structure of some prion strains does not favor interactions with PrP molecules expressing certain polymorphisms.
| Transmissible spongiform encephalopathies (TSE) or prion diseases are infectious fatal neurological diseases that affect many mammals, including humans. In these diseases a misfolded form of host prion protein (PrP) leads to brain degeneration and death. The genetic code of PrP in individual animals or humans has minor variations, which in some cases are associated with altered susceptibility to disease. In humans a variation at residue 171 (N171S) has been found in people mainly of African descent. However, due to the low incidence of the variation and difficult accessibility of these individuals, studies of prion diseases in these populations have not been carried out. Therefore, to create a laboratory animal model to study the effect of this variation on prion diseases, we generated transgenic mice expressing the mouse version of the human PrP variation at residue 171. We then studied the susceptibility of these mice to 4 strains of mouse-adapted scrapie. In our experiments these transgenic mice were uniquely resistant to two scrapie strains, but showed high sensitivity to two others. This resistance appeared to be related to a slow or inefficient generation of the aggregated disease-associated form of PrP in these mice, and was not duplicated using in vitro assays. In summary, transgenic mice expressing this variant PrP provide an interesting model to study differences among prion strains and their interactions with PrP in vivo.
| TSE or prion diseases are transmissible neurodegenerative diseases occurring in a variety of mammalian species including domestic and wild animals as well as humans [1]. Examples include scrapie in sheep, chronic wasting disease (CWD) in cervids and bovine spongiform encephalopathy (BSE) in cattle. Human diseases include sporadic and variant Creutzfeldt-Jakob disease (sCJD and vCJD) and familial diseases such as Gerstmann–Sträussler–Scheinker syndrome (GSS), Fatal Familial Insomnia (FFI) and familial CJD. A hallmark of prion diseases is the conversion of the normal protease-sensitive host prion protein (PrPsen) into a misfolded partially protease-resistant form (PrPres) which may be in part responsible for the generation of the disease [2]. PrP expression is required for prion disease, and PrP knockout mice are resistant to both infection and disease [3].
PrP gene polymorphisms occur naturally in many species. In both sheep and in mice such polymorphisms have been found to influence susceptibility to scrapie infection and disease. For example, sheep expressing Alanine, Arginine, Arginine at positions 136, 154 and 171 (A136R154R171) respectively, are highly resistant to most strains of sheep scrapie [4], [5], [6]. However, recent observations indicate that sheep with the A136R154R171 genotype can be susceptible to atypical scrapie (Nor98) suggesting that resistance might be strain specific [7], [8]. Similarly in inbred mice there are two known PrP alleles (Prnpa and Prnpb) which are characterized by amino acid differences at PrP residues 108 and 189 [9], [10], [11]. Mice with Prnpa have short incubation times with one set of scrapie strains and prolonged incubation times with another set, while mice with Prnpb show an opposite pattern of incubation times with these same scrapie strains [12], [13]. Thus there is a different strain-specific pattern of susceptibility associated with each of these Prnp genotypes.
In humans, the methionine vs. valine polymorphism at PrP residue 129 appears to influence the clinical phenotype of familial prion disease associated with the D178N PrP mutation [14], [15], as well as susceptibility to sCJD, vCJD and kuru [16], [17], [18], [19], [20]. In addition, the human polymorphism E219K may be associated with resistance to sCJD [21], but does not appear to correlate with resistance to vCJD in humans or mouse models [22], [23]. The human PrP polymorphism, G127V, was recently shown to be associated with resistance to kuru [24]. Other naturally occurring human polymorphisms, such as G142S and N171S, have not been tested for prion disease susceptibility [25].
N171S is a PrP polymorphism found in Sub-Saharan Africans, Jamaicans and Sardinians [26]. The rarity of this polymorphism and the fact that it is present mainly in geographical regions with limited CJD surveillance make it difficult to detect possible associations with prion disease. Therefore, in order to initiate laboratory studies of the possible effects of this polymorphism on TSE diseases and possibly also other CNS disorders, we generated transgenic mice expressing mouse PrP-170S, the mouse homolog of human PrP-171S (Table 1).
This approach of using transgenic mice expressing mouse or hamster PrP with human PrP mutations and/or polymorphic residues at homologous sites has been taken in numerous previous studies of prion diseases. These include important studies of models of Gerstmann-Sträussler-Scheinker syndrome (GSS) [27], [28], [29], [30], [31], [32], FFI [33] and familial CJD [34], [35], as well as a non-infectious neurodegenerative disease associated with expression of prion protein with a nine octapeptide insertion [36]. In addition, transgenic mice expressing mouse PrP with cervid PrP residues associated with a “rigid loop” in PrP have also revealed interesting in vivo pathogenic effects [37], [38].
Because PrP variations in animals and humans can have different effects on different TSE strains, we tested our mice with four different scrapie strains which have been maintained in continuous mouse passage for many years. In these studies PrP-170S expressing mice were highly susceptible to scrapie strains 22L and ME7, but were markedly resistant to scrapie strains RML and 79A. In contrast, control mice were susceptible to all four strains. This in vivo strain-specific influence on scrapie susceptibility was not reproduced in cell-free in vitro PrP conversion studies, suggesting that in vivo conditions not replicated in our in vitro system were required. PrP-170S transgenic mice appear to be an interesting new model to study the interactions between TSE agent strains and PrP.
All mice were housed at the Rocky Mountain Laboratories (RML) in an AAALAC-accredited facility and experimentation followed NIH RML Animal Care and Use Committee approved protocols (NIH/RML Protocol #2007-31). This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health.
To study the effect of the human N171S PrP polymorphism (Ref SNP#rs16990018) we constructed transgenic mice expressing serine instead of asparagine at mouse PrP residue 170 (homologous to human PrP residue 171) (Table 1). The mutation was made at residue 170 in mouse PrP using a cDNA clone of mouse PrP, p1-5 (E48-16) [39] using oligonucleotides 2077U (5′- CCA GTG GAT CAG TAC AGC AGC CAG AAC AAC TTC GTG C -3′) and 2078L (5′- GCA CGA AGT TGT TCT GGC TGC TGT ACT GAT CCA CTG G -3′) with a site directed mutagenesis kit (Stratagene/Agilent, Santa Clara, CA). The plasmid with the mutation was recloned, and the mutation was confirmed by sequencing. A DNA fragment from PshAI to SfoI with PrP sequences containing the mutation was excised, purified and religated into a subclone, p44-3 (E58-6) [39], [40], derived from the pHGPrP half-genomic clone [41] by digesting with AgeI and SfoI to remove a portion of the PrP ORF and replacing this segment with an oligonucleotide polylinker containing these and other sites. Upstream sequences previously excised between two BamHI sites to remove an unwanted SfoI site were replaced by digestion with adjacent sites NotI and BspEI and the original 6.2kb fragment from PHGPrP was reinserted. The resulting plasmid, p188-6, was digested with NotI and SbfI to separate mouse sequences from bacterial plasmid sequences, and the fragment containing the mouse sequences was used to inoculate C57BL/6 mouse eggs to generate transgenic mice expressing PrP-170S [39]. Of the five transgenic founding lines produced, three lines (Tg330, Tg340 and Tg290) were selected for experimentation. These transgenic lines were hemizygous for the transgene and homozygous for Prnp, the gene encoding normal mouse PrP (MoPrP). The Prnp gene encoding PrP-170N in C57BL/6 mice was removed by serial backcrossing to C57BL/10Sn mice with a knocked-out Prnp gene (B10 PrP-/-) derived from the original Edinburgh Prnp knockout mouse as previously described [39], [42]. Transgenic lines were maintained as transgene heterozygotes by crossing to C57BL/10 PrP-/- mice and selection of transgene positive mice by PCR analysis of tail DNA.
Genotyping was conducted using standard PCR reactions as previously described [39]. Briefly, detection of the modified knock-out version of Prnp and the neo cassette in the PrP-/- mice was accomplished using primers RK1 and Mut217 as previously described [43]. These primers yielded a C57BL/10 MoPrP product of approximately 700 bp and a PrP-/- neomycin cassette gene product of approximately 1700bp. Detection of the half-genomic PrP transgene, expressing the N170S PrP construct, was accomplished using primers pE2+ and Mut217 as described previously [43]. PCR products were visualized via electrophoresis in a 2% agarose gel.
All mice were housed at the Rocky Mountain Laboratories (RML) in an AAALAC-accredited facility and experimentation followed NIH RML Animal Care and Use Committee approved protocols (NIH/RML Protocol #2007-31). This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Mice were bred and genotyped at RML. For use as controls, weanling C57BL/10Hsd mice were obtained from Harlan Sprague Dawley, Madison, WI. Transgenic tga20 mice were obtained from EMMA (Munich, Germany) [41].
Mice were injected intracerebrally (i.c.) with 50μl of a 1% (wt/vol) dilution of brain homogenate pools from C57BL mice terminally ill from 22L, RML, 79A or ME7 scrapie. Two stocks of RML scrapie (RML-06, RML-81) were used. Titers of scrapie stocks were determined in previous i.c. endpoint titration experiments and were as follows; 22L = 2.5×109, ME7 = 4.0×108, RML-06 = 3.2×108, RML-81 = 4.8×108, 79A = 1.6×108 (units = 50% infectious dose (ID50)/gm of brain). Brain homogenates were diluted for inoculation in phosphate buffered balanced saline (PBBS) pH 7.2, supplemented with 2% fetal bovine serum (Hyclone, Logan, UT).
Observations were made daily to assess clinical signs of scrapie disease, which included ataxia, altered gait, wasting, kyphosis, hind limb weakness, aimless wandering, somnolence, immobility and leg clasping reflex. Mice with clinical signs were euthanized and brain tissue was analyzed for PrPres by immunoblotting. Mice with both clinical signs and PrPres by immunoblot were defined as diseased, and the day of euthanasia was recorded as the incubation period in the data presented. Uninoculated control transgenic mice were followed up to 700 days of age and showed no clinical signs of scrapie differing from normal signs of senescence. Several uninoculated mice were analyzed by histopathology after euthanasia at 465 days of age, and brain tissue had no detectable grey matter vacuolation or abnormal PrPres deposition.
Statistical analysis of data in the coexpression experiment was done by a one way ANOVA with Dunnett's multiple comparison test using GraphPad Prism software.
Tissue samples from mice were analyzed for PrPres and PrPsen by immunoblot. Brain homogenates (20%w/v) were prepared in 10 mM Tris-HCl [pH 7.4] using a mini-beadbeater (Biospec products, Bartlesville, OK). All homogenates were sonicated for 1 min using a Vibracell cup-horn sonicator (Sonics, Newtown, NJ) as previously described [40]. To test for PrPres, samples were proteinase K treated as follows; 20 µl of a 20% (w/v) tissue homogenate was adjusted to 100 mM Tris HCl (pH 8.3), 1% Triton X-100,1% sodium deoxycholate and 50 µg/ml proteinase K (PK), in a total volume of 31 µl. Tubes were mixed and incubated for 45 minutes at 37°C. The reaction was stopped by adding 2 µl of 100 mM Pefabloc (Roche Diagnostics, Indianapolis, Indiana) and placed on ice for 5 min. Samples tested for PrPsen were treated with proteinase inhibitors: 10 µM leupeptin, 1 µM pepstatin A, and 7 µg/ml aprotinin. PrPsen samples were not treated with PK. An equal volume of 2X Laemmli sample buffer (Biorad, Hercules, CA) was added to both PrPsen and PrPres samples, and then tubes were boiled 5 minutes. Samples were frozen at −20°C until the day of analysis, when samples were thawed, reboiled for 5 minutes and then electrophoresed on a 16% Tris-Glycine SDS-PAGE gel (Invitrogen, Carlsbad,CA) and blotted to PVDF membranes (Biorad) using a 7 minute transfer, program 3 (P3) on an iBlot (Invitrogen) device. Immunoblots were blocked in a solution of 2.5% milk (Biorad) in 0.1 M Tris, 1.5M NaCl and 0.05% Tween20, for 1 hour. Then blots were probed with D13 anti-PrP monoclonal antibody [44], (InPro, San Francisco, CA) at a final concentration of 0.2 µg/ml, followed by a peroxidase-conjugated anti-human IgG secondary antibody (Sigma, St. Louis, MO) at a final dilution of 1∶5000 in the blocking buffer described above. Bands were detected using enhanced chemiluminescence substrate (ECL) as directed by manufacturer (GE Healthcare, Pittsburgh, PA).
For histopathological analysis mice were deeply anesthetized and euthanized by cervical dislocation. Tissues were removed and placed in 3.7% phosphate-buffered formalin for 3 to 5 days before dehydration and embedding in paraffin. Serial 5 µm sections were cut using a standard Leica microtome, placed on positively charged glass slides and dried overnight at 43°C. Slides were then deparaffinized using standard procedures. Slides were stained with hematoxylin and eosin and analyzed for pathological changes. Immunohistochemical detection of PrPres using DAB chromogen (DAB Map kit; Ventana Medical Systems, Tucson, AZ.) was done as follows: antigen retrieval and staining were performed using the Ventana automated Discovery XT stainer. PrP antigens were exposed by incubation in CC1 buffer (Ventana) containing Tris-Borate-EDTA, pH 8.0 for 20 minutes at 95°C. Staining for PrP was done using human anti-mouse PrP monoclonal antibody D13 at a dilution of 1∶500 at 37°C for 2 hours, followed by a biotinylated anti-human IgG at 1∶500 (Jackson ImmunoResearch, West Grove, PA.), and avidin-horseradish peroxidase with DAB as chromogen. Slides were examined and photomicrographs were taken observed using an Olympus BX51 microscope and Microsuite FIVE software.
For preparation of normal brain homogenates containing PrPsen to be used as the substrate for PMCA reactions healthy transgenic mice expressing either PrP-170S (Tg330) or PrP-170N (Tga20) [41] were deeply anesthetized and then perfused with PBS containing 5 mM EDTA. Brains were removed and homogenized in a beadbeater at a concentration of 20% (w/v) in PMCA conversion buffer (PBS containing 4 mM EDTA, 1% Triton X-100 and complete mini-protease inhibitor cocktail (Roche), sterilized by filtration with a 0.2 µm filter) then cooled on ice for 10–15 min. Homogenates were diluted with PMCA conversion buffer to 10% (w/v) then clarified by a brief 1500 X g spin [45]. Supernatants were stored at −80° C in 1 ml aliquots until used in PMCA reactions.
Brain homogenates used as seeds for the PMCA reactions came from clinically sick 22L or RML-infected C57BL/6 mice. Brains were homogenized at a 20% (w/v) concentration using 0.1 M Tris pH 7.4, then diluted in the same buffer to a final concentration of 10% (w/v) prior to storage at −80°C. Scrapie-positive brain homogenates were used to “seed” normal brain homogenate in the following manner. A 10% scrapie-positive brain homogenate (seed) was added to normal brain homogenate (substrate) at desired dilutions and then these master mixes were aliquoted into multiple 0.2ml reaction tubes (GeneMate, ISCBioexpress, Kaysville, Utah). One of these tubes was frozen as an unsonicated control and the remaining tubes were repeatedly sonicated and incubated as described below. Tubes were positioned in a plastic tube rack PMCA adapter (Misonix, Farmingdale, NY) and placed on the rim of a microplate horn of a Misonix Model 3000 microsonicator so that the 50 µl samples were immersed in the sonicator bath. The microplate horn was covered with a plastic lid to minimize evaporation from the water bath. The sonicator was located inside an incubator set to 37°C and was programmed to perform cycles, each consisting of a 40 second pulse of sonication set at 60% maximum followed by a 30 min incubation. Forty-eight cycles (i.e. 24 h) constituted one round of PMCA.
PMCA reaction tubes were removed from the sonicator and vortexed and 10 µl of the 50 µl volume was sampled and mixed with 10 µl PK at 100 µg/ml for a final PK concentration of 50 µg/ml. Samples were maintained at 37°C in a water bath for one hour. The PK digestions were halted with 2 µl of 100 mM Pefabloc (Roche Diagnostics) and placed on ice for 5 min. An equal volume of 2X Laemmli sample buffer (Biorad, Hercules, CA) was added to the PMCA samples, and then tubes were boiled 5 minutes. Samples were frozen at −20°C until the day of analysis, when samples were thawed, and reboiled for 5 minutes. PMCA products were visualized by SDS page gel electrophoresis and immunoblotting as described above for brain tissues, with the following exceptions. Gels were blotted to HyBond ECL nitrocellulose membranes (GE Healthcare Life Sciences) instead of PVDF membrane. Immunoblots were blocked in Near-Infrared Fluorescence Western Blotting Blocking Buffer (Rockland Immunochemicals Inc., Gilbertsville, PA) and PBS mixed in equal parts. Primary antibody was D13 1∶100 diluted supernatant derived from CHO cells expressing the D13 antibody construct [44]. These cells were obtained from R. Anthony Williamson, The Scripps Research Institute, La Jolla, CA. The secondary antibody was IRDye800CW-conjugated goat-anti-mouse IgG (LiCor, Lincoln, NE) diluted at 1∶10,000. Both antibodies were diluted in the blocking buffer described above with the addition of 0.2% Tween 20. Finally, bands were detected using an Odyssey near-infrared fluorescence scanner (LiCor). Groups were compared using a Mann-Whitney test with GraphPad Prism software.
To study the effect of the human PrP polymorphism, N171S, on susceptibility to prion disease we generated transgenic mice expressing mouse PrP-170S, the mouse homolog of the human PrP-171S (Table 1). Five founder lines were produced on the C57BL/6 background, and these were crossed serially to C57BL/10 mice homozygous for the Edinburgh version of the PrP null gene (Prnp-/-) [42]. Three lines of the transgenic mice (Tg290, Tg330, and Tg340) with the Prnp-/- gene, and hemizygous for the PrP-170S transgene, were obtained and used for further study.
Because PrPsen expression is known to influence scrapie incubation period [41], [46], we determined PrPsen levels in brain by immunoblot. Both Tg330 and Tg340 mice expressed PrPsen at levels 2 to 3-fold higher than were seen in non-transgenic control C57BL/10 mice (Prnp+/+) which express PrP-170N (Figure 1). Tg290 expressed approximately 10-fold lower PrP levels than did control mice (data not shown). All three strains of PrP-170S mice were next tested for susceptibility to scrapie infection.
Four strains of scrapie were used to test the influence of the N170S polymorphism on susceptibility to TSE disease. In Tg330 and Tg340 mice, the 22L and ME7 scrapie strains produced 100% incidence of typical fatal prion disease with clinical signs similar to control C57BL/10 mice (Table 2). Shorter incubation periods were observed in the Tg330 and Tg340 mice than in C57BL/10 mice, probably due to higher PrPsen expression levels in the transgenic mice. The diagnosis of scrapie was confirmed by the detection of PrPres in brain by immunoblotting. Brain PrPres levels in transgenic mice were slightly less than those in control mice, possibly due to shorter incubation periods (Figure 2A). In Tg290 mice inoculation with scrapie strain 22L produced disease in 6 of 8 mice at an average incubation period of 615 dpi (data not shown). This lower incidence of disease and slower tempo appeared to be related to the low PrPsen expression in Tg290 mice, and this line was not studied further.
Unexpected results were observed upon inoculation of RML and 79A scrapie strains. Whereas control C57BL/10 mice were uniformly susceptible to these two strains, transgenic Tg 330 and Tg340 mice were quite resistant. Only two mice inoculated with RML and two mice inoculated with 79A had clinical disease and PrPres detectable by immunoblot (Figure 2), and all four occurred at late times ranging from 439-545dpi (Table 2).
Brains were also examined microscopically for pathology and presence of PrPres. In Tg mice infected with strains ME7 or 22L, at the time of clinical disease vacuolation and PrPres deposition were widespread in many brain regions (Figure 3A,B,C) similar to non-transgenic control mice (not shown). In contrast, after infection with strains 79A or RML, Tg mice with clinical signs or PrPres detectable by immunoblot showed localized vacuolation and PrPres deposition (Figures 3E and 3H) mostly limited to the thalamus, hippocampus and pons. In addition, ten RML or 79A-infected Tg mice, who were negative for PrPres by immunoblot, had subclinical infection as shown by brain PrPres deposits mainly localized to the vestibular nuclei in the pons or to the anterodorsal region of the thalamus detected at 545-603dpi (Figure 3I,J,K,L). In these same areas gray matter vacuolation was either absent or minimal (Figure 3F). Detection of these subclinical mice provided evidence of a higher incidence of infection than was shown by standard clinical observation and analysis of brain PrPres by immunoblotting. Perhaps these mice should be considered preclinical as they might have developed clinical disease if allowed to survive for a longer time. These results suggested that infection of Tg PrP-170S mice by RML or 79A scrapie occurred at slow and reduced levels, and this appeared to account for the rare presence of clinical signs in this experiment.
Our in vivo experiments showed a dramatic difference in the tempo and levels of PrPres generation in brains of Tg330 and Tg340 mice after infection with scrapie strains RML and 79A compared to strains 22L and ME7. Possibly a delay in PrP conversion and subsequent accumulation of PrPres might account for our results. Therefore to test PrP conversion under in vitro conditions we used the PMCA cell-free system [47]. As a source of PrPsen, brain homogenates derived from either Tga20 mice expressing PrP-170N or Tg330 mice expressing PrP-170S were used, and these reactions were initiated by seeding with PrPres from PrP-170N mice infected with scrapie strains 22L or RML.
Surprisingly, RML PrPres seed generated PrPres conversion with both PrP-170N and PrP-170S brain homogenates. Quantification of PrPres produced in the PMCA revealed slightly greater amounts of product in the PrP-170N reactions, but these differences were not statistically significant (Figure 4A, 4B and 4C). However, the positive conversion of PrP-170S by RML PrPres indicated that there was not a basic inability of RML scrapie PrPres to convert PrP-170S. This was in contrast to the slow conversion in vivo requiring over 400 days when Tg330 and Tg340 mice were inoculated with RML scrapie (Figures 2 and 3).
In similar PMCA reactions 22L PrPres seed also gave higher PrPres generation using PrP-170N compared to PrP-170S substrate. These differences were statistically different (Figure 4D, 4E, 4F), and might in part be due to the higher PrP expression in substrates from tga20 versus Tg330 mice. In summary, comparison of PrP-170N and PrP-170S mice by seeding with PrPres from RML or 22L gave no evidence that in vitro conversion of PrP-170S by RML PrPres was abnormal as detected by PMCA.
In numerous previous studies expression of two different PrP alleles from the same or different species has been shown to reduce the level of prion infection as well as the incidence and tempo of disease [22], [48], [49]. The resistance of the Tg330 and Tg340 mice to RML and 79A scrapie prompted us to test whether the presence of both PrP-170S and PrP-170N in the same mouse would result in lower disease incidence or increased incubation times.
To investigate this question, 22L or RML scrapie was inoculated into mice expressing two PrP alleles, one allele of the PrP-170S transgene and one allele of the mouse Prnp gene expressing PrP-170N. Mice expressing only one allele of PrP-170N (Prnp+/−) were inoculated as controls. With the 22L strain, mice expressing both 170S and 170N PrP had significantly faster incubation times (106–115dpi) than mice expressing only 170N (259 dpi) (Table 3). Therefore interference between these PrP variants did not seem to occur. The decrease in incubation period seen when both PrP alleles were expressed could be explained if the PrP-170S variant could contribute to a more rapid incubation time, as would be predicted by the susceptibility of the original Tg330 and Tg340 mice to 22L infection.
In the case of strain RML, the expression of both alleles was associated with a decrease in incubation time compared to expression of PrP-170N alone, but these differences were not statistically significant (Table 3). Therefore, the more highly expressed PrP-170S protein appeared to contribute very little towards accelerating the disease tempo in these mice, and also there was no evidence for interference. This outcome might be predicted from the very low susceptibility of the PrP-170S transgenic mice to RML scrapie infection (Table 2). In summary, the PrP-170S variant appeared to be “neutral” during the infection by RML scrapie, showing no interference with the PrP-170N in the co-expression experiment.
In the present experiments the influence of a normal human prion protein gene allelic variation, N171S, on prion disease susceptibility was studied in a mouse system using transgenic mice expressing the mouse homolog, N170S. Susceptibility to prion disease induced by mouse scrapie strains, 22L and ME7, was identical in control mice expressing PrP-170N versus transgenic mice expressing PrP-170S. In contrast, after inoculation with scrapie strains, RML or 79A, mice expressing PrP-170S were markedly resistant compared to control mice expressing PrP-170N. Of the 43 transgenic mice inoculated with these two scrapie strains, disease occurred only after 439 days and this was seen in only 4 mice. An additional 10 mice had subclinical infection after 545–603 days, as shown by detection of PrPres by IHC, however the remaining 29 mice had no evidence of clinical signs or PrPres accumulation in brain after observation up to 603 days. These results demonstrated that mice expressing PrP-170S were highly resistant to infection by strains RML and 79A, but this resistance was strain-specific since there was no resistance to two other strains (ME7 and 22L).
One explanation for the scrapie strain-specific differences seen in our experiments may lie in the origins of the scrapie strains used (Figure 5). Strain ME7 was the result of a passage of natural scrapie in Suffolk sheep directly to mice and therefore was not related to other scrapie strains[50]. On the other hand, 22L, RML and 79A share a common origin in that they were all derived from the Moredun Institute's sheep scrapie brain pool 1(SSBP/1)[51]. However, there were major differences in the passage history of each of these 3 strains, and strains 79A and RML were more closely related to each other than to strain 22L [51], [52] (Figure 5). This may explain the similar resistance pattern of strains 79A and RML in PrP-170S mice.
Although the biochemical explanation for scrapie strains in general remains a mystery at this time, it has been hypothesized that PrPres from each strain adopts a slightly different conformation which is conferred on successive PrPsen molecules as they are converted to PrPres [53]. Accordingly, our current data might be explained by the conformational selection model of prion strains and species barriers [54], [55] which was first developed to explain prion strain differences involving the Sup35 protein in yeast [56], [57], [58]. This model suggests that only a subset of all possible PrPres conformations is compatible with any individual PrP primary structure. Thus incompatibility between the infecting prion strain and the host PrP would result in a transmission barrier [54]. For example, mouse PrP-170S might easily assume the conformations required by strains ME7 or 22L, but might be less adept at assuming the conformations associated with strains RML or 79A. In practice most mutations studied lead to altered incubation periods [30], [31], but when strong species barriers exist, either no transmission or low level subclinical cross-species transmission has been observed [59], [60], [61].
Another possible explanation for our results is that in vivo conversion of PrP-170S by RML or 79A strains might generate PrPres with a lower than usual stability. This might be due to either high dissociation of PrPres into smaller oligomers or increased susceptibility of PrPres to catabolic degradation. However, we were not able to detect any evidence for the presence of PrPres with increased susceptibility to proteinase K in PrP-170S transgenic mice infected by either RML or 79A scrapie strains (data not shown).
Effects of PrP amino acid variations on prion disease species and strain transmission barriers in some cases appear to correlate with differences in PrP conversion. For example, sheep homozygous for PrP with V136R154Q 171 or A136R154Q 171 show opposite patterns of susceptibility to scrapie strains SSBP1 and CH1641 [62], [63], and in vitro generation of PrPres in PMCA reactions agreed with the in vivo resistance observed [64]. However, when we tested PrP conversion in vitro using PMCA, both 22L and RML PrPres were able to seed the generation of PrPres derived from PrP-170S (Figure 4). This conversion of PrP-170S by RML scrapie was in contrast to the ineffective production of disease and slow and low PrPres generation in PrP-170S mice infected with RML scrapie. Biological and biochemical differences between conditions in brain tissue of live mice and PMCA test tube reactions might account for the discrepancies between our in vivo and in vitro results. Similar discrepancies between PrPres generation by PMCA in vitro and clinical susceptibility have been noted previously, and in some cases strong in vivo transmission barriers between species have been easily overcome by using PMCA with minor alterations in conditions [65]. Thus PrP conversion by PMCA would appear to be less selective than in vivo infection by TSE agents.
Interestingly, the N171S polymorphism in humans (homologous to the N170S change in our Tg330 and Tg340 mice) occurs near other PrP residues implicated in influencing PrP structure and folding as well as susceptibility to prion diseases in animals. For example, alterations in PrP folding in vitro have been noted after mutations homologous to human residues 168 [66] and 170 [67]. At the structural level, PrP mutation at human residue 170 (S170N) appears to create a stabilized loop structure located near residues 165-175 [68], and this change may influence susceptibility to CWD and other prion agents [37], [38]. However, others have reported no effect of S170N on species-specific PrP conversion in vitro in a mouse-hamster system [69], [70], [71]. In contrast, resistance of rabbits to prion disease appears to be associated with a serine at human PrP residue 174 [72]. Similarly, the sheep polymorphism at residue 171 (human residue 168) is important in the resistance of sheep to classical scrapie strains [6], [73], [74]. Possibly the N171S polymorphism examined in the present study might be able to modulate prion disease because of its location near the PrP loop structure and other nearby influential residues.
In humans, familial prion diseases have been associated with PrP mutations in the near vicinity of the PrP loop structure, i.e. Q160X [75], Y163X [76], D178N [77] and V180I [78]. The N171S polymorphism is not by itself associated with familial prion disease [26]. However, a previous study identified a family with an unusual psychiatric disorder associated with PrP N171S [79]. More recently, in an African-American family with unusual psychiatric signs and sleep abnormalities preceding onset of familial CJD, disease was linked to expression of a PrP molecule containing both PrP N171S and D178N mutations [80]. Interestingly this is the first family with African ancestry where the D178N mutation has been detected, as the 12 previously reported families were of European or Japanese descent [78], [80]. This family might be an example where the N171S polymorphism altered the clinical disease signs when expressed in combination with a known pathogenic PrP mutation. It remains unclear whether this effect is mediated by a direct influence of these mutations on PrP misfolding or whether indirect effects involving other non-PrP molecules might also play a role. Nevertheless, the fact that the N171S polymorphism is present in healthy populations of humans [81], suggests that N171S is likely non-pathogenic by itself and that there may even be a selective advantage for maintaining its presence in human genomes.
|
10.1371/journal.ppat.1003067 | Influenza Human Monoclonal Antibody 1F1 Interacts with Three Major Antigenic Sites and Residues Mediating Human Receptor Specificity in H1N1 Viruses | Most monoclonal antibodies (mAbs) to the influenza A virus hemagglutinin (HA) head domain exhibit very limited breadth of inhibitory activity due to antigenic drift in field strains. However, mAb 1F1, isolated from a 1918 influenza pandemic survivor, inhibits select human H1 viruses (1918, 1943, 1947, and 1977 isolates). The crystal structure of 1F1 in complex with the 1918 HA shows that 1F1 contacts residues that are classically defined as belonging to three distinct antigenic sites, Sa, Sb and Ca2. The 1F1 heavy chain also reaches into the receptor binding site (RBS) and interacts with residues that contact sialoglycan receptors and determine HA receptor specificity. The 1F1 epitope is remarkably similar to the previously described murine HC63 H3 epitope, despite significant sequence differences between H1 and H3 HAs. Both antibodies potently inhibit receptor binding, but only HC63 can block the pH-induced conformational changes in HA that drive membrane fusion. Contacts within the RBS suggested that 1F1 may be sensitive to changes that alter HA receptor binding activity. Affinity assays confirmed that sequence changes that switch the HA to avian receptor specificity affect binding of 1F1 and a mAb possessing a closely related heavy chain, 1I20. To characterize 1F1 cross-reactivity, additional escape mutant selection and site-directed mutagenesis were performed. Residues 190 and 227 in the 1F1 epitope were found to be critical for 1F1 reactivity towards 1918, 1943 and 1977 HAs, as well as for 1I20 reactivity towards the 1918 HA. Therefore, 1F1 heavy-chain interactions with conserved RBS residues likely contribute to its ability to inhibit divergent HAs.
| Influenza infection kills thousands of people every year and causes major pandemics every few decades. The most lethal outbreak of influenza known was the 1918 H1N1 influenza pandemic that killed an estimated 20 to 100 million people. The 1918 virus was likely introduced into the human population from birds. We previously described five human neutralizing antibodies from survivors of the 1918 pandemic that bind the hemagglutinin (HA) surface antigen. Here, we define the binding sites of antibodies 1F1 and 1I20 on the 1918 HA and demonstrate that these overlap with the glycan receptor binding site. The glycan specificity differs between human and avian viruses for the linkages of the sialylated sugar receptors [human (α2–6) or avian (α2–3)]. 1F1 and 1I20 binds viruses that contain HA residues that mediate preference for α2–6 sialylated sugars. Three other control antibodies were not affected by preferences for the linkages of the sialylated sugar receptors because they bind elsewhere. Since the receptor-binding site is relatively conserved, this may explain the cross-reactivity of 1F1 and the enhanced binding of 1F1 and 1I20 to HAs with human receptor specificity.
| The hemagglutinin (HA) protein of influenza viruses binds to sialic acid receptors on host cells and is the major target of neutralizing antibodies. Amino-acid changes in the immunodominant HA antigenic sites that arise in response to immune selective pressure (antigenic drift) enable seasonal influenza A viruses to cause repeated epidemics and necessitate continuous reevaluation of the composition of influenza vaccines. Characterization of antibodies that display the ability to cross-neutralize divergent viruses may suggest strategies to elicit more broadly protective immunity. The broadest cross-reactive influenza mAbs described to date recognize conserved regions of the HA stem [1], [2], [3], [4], [5], [6] as compared to the HA head region, which is much more variable. Nevertheless, a few cross-reactive antibodies to the HA head have also been found [7], [8], [9], [10], [11], [12], [13], [14]. S139 is a murine monoclonal antibody against antigenic site B [7], but also reaches into the receptor binding site [13]. Recently, human monoclonal antibodies of the VH1-69 lineage against the receptor-binding pocket have been described by Ohshima et al. [8] and our group [11]. Whittle et al. described the H1N1 antibody CH65 [9] which is complementary in its H1N1 activity to our H1N1 antibody 5J8 [10]. Antibody C05 also binds to the receptor binding site of multiple influenza A subtypes using mainly its CDR H3 loop [14]. Recently, a cross-reactive antibody to influenza B CR8033 was shown to bind to the head and overlap with the receptor-binding pocket [12]. Extensive epitope mapping with large panels of murine mAbs previously identified five major antigenic sites on the HA head domain of H1N1 viruses, and these have been termed Sa, Sb (residues 186–198), Ca1, Ca2 (residues 140–145, 224–225) and Cb [15], [16], [17]. These highly variable surface-exposed regions are located in the membrane-distal end of the HA trimer near the HA RBS.
In a previous study [18], we described five naturally occurring human mAbs that potently inhibit the 1918 H1N1 pandemic influenza virus. The antibodies were cloned from the B cells of individuals born prior to 1918, and were isolated prior to the 2009 H1 pandemic; the mAbs were designated 1F1, 1I20, 2B12, 2D1, and 4D20 [18]. Two of these mAbs, 1F1 and 1I20, independently selected escape mutants of the 1918-like influenza A/Swine/Iowa/15/30 (H1N1) virus with the same proline to histidine escape mutation at HA1 position 186, a residue adjacent to the Sb antigenic site [18]. Sequence analysis with the online antibody database tool IMGT [19] revealed the antibody heavy chain genes encoding mAbs 1F1 and 1I20 used similar V, D, and J gene segments (VH3–30, D3–22, JH5 for 1F1; VH3–30, D3–10, JH5 for 1I20; Table S1), but different light chain types (λ for 1F1; κ for 1I20; Table S1). Similar concentrations of 1F1 and 1I20 inhibit the 1918 virus, and administration of either antibody protected mice from lethal 1918 virus challenge [18]. One notable difference between these two antibodies, however, is that mAb 1F1 not only inhibits the 1918 H1N1 virus, but also 1943 as well as select 1947 and 1977 human H1N1 viruses [18]. 1I20 does not inhibit these antigenically drifted post-1918 human viruses.
We used a combination of X-ray crystallography, site-directed mutagenesis, selection of antibody escape mutant viruses and biochemical assays to define the epitopes of 1F1 and the related 1I20 antibodies. The structure of 1F1 in complex with the 1918 HA demonstrated that the 1F1 heavy chain reaches into the HA receptor binding pocket and makes contacts with residues that interact with the sialoglycan receptor. Additional contacts are made with residues from the Sa, Sb and Ca2 antigenic sites. Hemagglutination-inhibition and binding assays corroborated the crystal structure data and indicated that the 1I20 mAb likely binds HA in a similar manner as 1F1.
To better understand the mechanism of 1F1 neutralization and its ability to cross-react with H1N1 viruses separated by decades of virus evolution, we determined the crystal structure of Fab 1F1 in complex with SC1918 HA using diffraction data that extend to 3.3 Å, but we report a nominal resolution of 3.55 Å due to anisotropy (see Table S2). We also determined crystal structures of the 1F1 Fab and “avianized” 1918 HA (SC1918 D190E D225G; AV1918) components separately at high resolution (1.45 Å and 1.80 Å resolution, respectively). The new AV1918 structure was refined at significantly higher resolution than closely related structures reported previously (e.g., SC1918, PDB code 1RUZ) [20], [21]. The availability of these high-resolution models for molecular replacement and as restraints in structure refinement significantly improved the final quality of the lower resolution Fab-HA complex structure. Data collection and refinement statistics are reported in Table S2. The overall structure of AV1918 HA is very similar to previously reported structures for SC1918 HA. The two amino-acid substitutions that differentiate Av1918 from SC1918 (D190E and D225G) had no significant effect on the overall architecture of the receptor binding site, aside from the altered side-chain substitutions.
The crystal structure of the 1F1-SC1918 complex contains two complete HA trimers in the asymmetric unit. Each HA protomer (6 in all for the two trimers) is bound by a single copy of the 1F1 Fab in the expected stoichiometry of 3 Fabs per HA trimer. Of the 6 Fabs in the asymmetric unit, only two are fully ordered. In the remaining four copies, only the variable domains are well defined, as observed in other Fab and Ig-domain containing structures [1], [22], [23]. Together these disordered domains constitute ∼17% of the total expected protein mass in the asymmetric unit. As the variable and constant domains are joined via two flexible hinges in the “elbow” region and these four sets of 1F1 constant domains fail to make significant crystal contacts in the relatively open crystal lattice, these constant regions likely adopt an ensemble of conformations in the crystal and are not interpretable in our electron density maps. Indeed, the Fab elbow angles vary over 30° between the 1F1 Fabs resolved in the free Fab structure (215° and 217°) and in the SC1918 complex (185° and 190°).
Fab 1F1 binds an epitope at the apex of the HA spike, in the HA1 “head” region (Figure 1, overview). The 1F1 epitope contains several residues that typically contact the sialoglycan receptor, including 135, 153, 183, 190, 194, 222, and 225 (Figure 2, RBS with glycan, contact residues labeled). Of particular note are the H1 HA receptor specificity-determining residues 190 and 225 [20], [21], [24], [25], [26], [27]. Many of these contacts are mediated by CDR-H3, which inserts its tip into the RBS. 1F1 also contacts a number of more variable residues outside the RBS, including 133A, 145, 156, 159, 186, 187, 189, 192, 193, 196, 227, and 228 (Figure 2). In total, 1F1 binding to the HA buries a total of 1440 Å2 of protein surface at the interface. Of this, approximately 72% is buried by the heavy chain and 28% by the light chain, similar to many other antibodies to proteins.
Remarkably, the epitope of 1F1 on H1 HA is similar to the HC63 epitope in H3 HA (PDB code 1KEN) [28], [29], despite highly divergent sequences on both sides of the interface for both the antibodies and the HA targets (Figure 3, 1F1 vs. HC63). Furthermore, the overall orientation of the VH domains from HC63 and 1F1 are very similar, resulting in their CDR-H1, -H2, and -H3 contacting similar surfaces on the H3 and H1 HAs, respectively. In particular, HC63 also inserts the tip of CDR-H3 into the receptor binding site, albeit somewhat less deeply than 1F1 due to a shorter length CDR-H3 (11 residues for HC63 versus 17 for 1F1). However, a slight rotation (∼20°) of the VH domain around its interface with the HA results in significantly different interactions between the light chains of these two antibodies. In contrast to HC63, where its CDR-L1 and -L2 are also centered on the receptor binding site, the 1F1 light chain is displaced by more than 10 Å, moving the tip of CDR-L2 out of the receptor binding site where it binds the outer surface of the 190-helix. Despite these differences, the overall similarity between the HC63 and 1F1 interactions is intriguing, and is suggestive perhaps of a limited number of preferred antibody binding modes, even across subtypes.
Given that the 1F1 epitope is comprised of several residues that typically contact the sialoglycan receptor, 1F1, the related 1I20, and three antibodies, mAbs 2B12, 2D1, and 4D20, that recognize other epitopes, were tested for binding using a label-free biosensor (Table 1) to recombinant, soluble HA proteins possessing SC1918 HA (α2,6SA specificity), NY1918 HA (dual α2,6SA and α2,3SA specificity), or Av1918 HA (α2,3SA specificity). These HAs differ only at positions 190 and 225 [24]. Binding of mAbs 2B12, 2D1, and 4D20, was unaffected by a D190E change, a D225G change converting the South Carolina to the New York sequence, or a double mutant that is Av1918 HA (Table 1). The D190E mutation has reduced affinity for mAb 1F1 by about 250-fold and for 1I20 by about 1,900-fold (Table 1). The D225G mutation led to a reduction of affinity of ∼360-fold for 1F1 and eight-fold for 1I20. Binding of mAb 1I20 to Av1918 was not detected, and a substantially reduced affinity of only 1.0×10−6 M was found for mAb 1F1 (Table 1). These data are consistent with the structural data implicating sialoglycan receptor-contacting residues within the 1F1 epitope. 1I20 mAb bound to a similar epitope consistent with the fact that it shares common genetic elements with 1F1 [30], although both antibodies are derived from different clonal ancestors because of different junctions, different light chains, and a different pedigree of somatic mutations.
Although influenza A/South Carolina/1/18, A/Weiss/43, A/Fort Monmouth/1/47, and A/USSR/92/77 (referred to below as the 1918, 1943, 1947 FM, and 1977 viruses, respectively) possess divergent HAs, including divergent Sb sites (Figure 4), mAb 1F1 inhibits hemagglutination by each of these viruses (Figure 5A) [18]. To map HA residues critical for HAI activity against the 1943 and 1977 viruses, 1F1 escape mutants for 1943 and 1977 viruses were selected. Sequencing of escape mutant HA genes of 1943 virus escape mutants reproducibly identified an A227T mutation in HA1. Sequences derived from 1977 escape mutant viruses possessed a D190N change in HA1, although some escape mutant strains appeared to be mixed populations that also contained an S186F change. To confirm that individual changes at positions 190 and 227 are sufficient to confer escape from inhibition, we introduced the changes into cDNAs encoding the HAs of 1943 or 1977 H1N1 viruses and produced virus-like particles (VLPs). As expected, 1F1 lost HAI activity towards the A227T 1943 HA (Figure 5B) and the D190N mutant 1977 HA (Figure 5C).
Prior studies to select for 1F1 and 1I20 escape mutants in a virus closely related to the 1918 virus, influenza A/swine/Iowa/30 (H1N1), repeatedly resulted in changes at residue 186 [18]. To determine whether the residues identified by escape mutant selection for the 1943 and 1977 viruses could also be implicated in 1F1 HAI activity towards the 1918 HA, residues 190 and 227 were mutated in the context of the 1918 HA. When an A227T mutation was introduced into a 1918 virus HA, it did not affect 1F1 HAI activity. However, alignment of the HAs of 20th century human H1N1 isolates identified several other amino acids that have appeared at position 227 (Figures 4, 6A). When these were introduced into the 1918 HA, two mutations, A227H and A227P, substantially impacted 1F1 HAI activity, demonstrating a role for residue 227 in inhibition of 1918 HA by both 1F1 and 1I20 (Figure 6A). When a D190N mutation was introduced into a 1918 HA, 1F1 HAI activity was partially abrogated (Figure 6B). These data support a role for residues 190 and 227 in 1F1 inhibition of the 1918 virus as well as in the 1943 and 1977 viruses.
The three HA residues involved in the four escape mutations selected by 1F1 (P186H, S186F, D190N, and A227T) all map to the HA-1F1 interface in the crystal structure. The P186H and S186F mutations introduce larger side chains at a position buried in the interface and likely lead to a steric clash between the HA and CDR-H3. It is less clear how the D190N and A227T mutations led to virus escape, as they did not completely abolish 1F1 HAI activity when introduced in the 1918 HA as compared to 1977 HA. Given the somewhat conservative nature of these substitutions, they may be expected to have a small effect on the binding of 1F1 for HA, which is insufficient to reduce antibody binding below the threshold necessary for escape. However, in the context of HAs bound by 1F1 with lower affinity, such as 1943 and 1977, these more subtle mutations may reduce 1F1 binding beyond what is required for effective neutralization.
Relatively few mAbs that bind the influenza virus HA head have been demonstrated to neutralize significantly divergent strains [7], [8], [9], [10], [11]. A greater understanding of the basis for such broadly cross-reactive antibodies may suggest novel vaccine or therapeutic approaches to influenza virus infection. This study provides a detailed characterization of one such antibody, 1F1, and highlights common, emerging properties of cross-reactive anti-head antibodies. Notably, such antibodies appear to react with the relatively conserved receptor binding domain, reach into the receptor binding pocket and/or are sensitive to changes in HA receptor binding specificity [7], [8], [9], [10], [11], [12], [13], [14]. The availability of 1I20, a mAb that shares a heavy chain closely related to that of 1F1, but did not neutralize the 1943, 1947 or 1977 viruses tested, provides additional insight. Because 1I20 maps to the same epitope as 1F1, it is very likely that 1I20 will also reach into the receptor binding pocket. This finding suggests that not all mAbs that make contact with the RBS will be broadly cross-reactive and that additional determinants of cross-reactivity exist as the antibody footprint typically includes surrounding hypervariable loops as well.
We previously characterized 1F1 and 1I20 as Sb site antibodies based on the P186H escape mutation for a residue that is immediately adjacent to the Sb site (residues 187–197 [16]). However, this antigenic site definition is based on BALB/c mouse hybridoma antibodies against influenza A/Puerto Rico/8/1934 (H1N1) [15], [16], [17]. Those mapping studies, while extensive, cannot necessarily be considered complete especially as applied to the human antibody response [15], [17]. Also, co-crystal structures of H1 HA antibodies binding the globular head that would validate the conventional definition for H1 HA are limited. We have shown previously that the epitope of the H1 HA Sa site antibody 2D1 extends to residues beyond the conventionally defined antigenic site to sites Sb and Ca1 [31]. Here, the crystal structure demonstrates that 1F1 interacts with residues within Sa, Sb and Ca2 and also reaches into the HA receptor binding pocket. This epitope is strikingly similar to that described for HC63 [29], an H3-specific mouse mAb that also exhibited HAI activity against multiple divergent H3N2 viruses [28]. Only a limited number of H1N1 antibodies have been crystallized in complex with their respective HAs. Their epitopes seem to be comprised of multiple antigenic sites, rather than just a single distinct antigenic site. It might now be time to get away from the original classification of epitope sites. Still, this conventional definition remains useful for ease of topographic orientation on the HA head. HC63 appears to neutralize virus through two distinct mechanisms: The direct occlusion of the receptor binding site by VH interferes with receptor binding and virus attachment [29]. Interestingly, HC63 inhibits the pH-induced conformational changes associated with membrane fusion [29], presumably by cross-linking or otherwise interfering with the separation of the heads, which has been suggested to be required to allow efficient reorganization of HA2 to its fusion-active conformation [32]. In contrast, while 1F1 potently inhibits receptor binding, it was unable to block the pH-induced conformational changes in HA that drive membrane fusion (data not shown). While the overall binding modes of 1F1 and HC63 are similar, the HC63 footprint extends across the interface between two adjacent HA1 subunits [29], while that of 1F1 is wholly contained within a single HA1 domain and cannot cross-link the subunits of the trimer as a monovalent Fab.
The extension of the 1F1/1I20 epitope towards the RBS and the modest conservation of the residues in this pocket may explain in part why 1F1 shows cross-reactivity towards later 1943, 1947, and 1977 viruses. The related mAb 1I20 does not inhibit these viruses, likely as a result of numerous substitutions in contact positions in CDR-H1 and -H3 (Table S1). In contrast, mAbs 2B12, 2D1, and 4D20 are not affected by changes in human versus avian receptor specificity since their epitopes do not involve those residues.
The HA receptor specificity of influenza A virus strains is a determinant of virus transmissibility and virulence. Therefore, methods to rapidly and easily determine HA receptor specificity would be of interest for influenza virus surveillance and research purposes. 1F1 and other antibodies with broad reactivity and which are sensitive to changes in receptor specificity could serve as such reagents. The conserved receptor-binding pocket may also be an attractive target for universal or improved influenza vaccine design as a complement to targeting the hemagglutinin stem. Structures like those of the 1F1-HA complex may serve as templates for such vaccine constructs.
For the experiments described below, H1N1 HA influenza sequences are used based on the following strains (abbreviation in brackets): A/South Carolina/1/18 (1918 wt), A/Weiss/43 (1943 wt), A/Fort Monmouth/1/47 (1947 FM), A/USA/L3/47 (1947 L3) [33] (GenBank accession number GI: 343409202), A/USSR/92/77 (1977 wt), and A/New Caledonia/20/99 (1999 wt) virus. The stated positions of all HA residues designated in this manuscript are based on the amino-acid numbering conventions used for H3 [34].
The antibody proteins were expressed recombinantly in mammalian cells as described [31]. Briefly, we cloned the matched heavy or light chain gene by RT-PCR (mAb 1F1 heavy/λ, mAb 2B12 heavy/λ, mAb 2D1 λ, mAb 4D20 λ) using In-Fusion enzyme (Clontech/Takara Bio) into opened pEE6.4 or pEE12.4 vectors (Lonza Group Ltd), respectively. These vectors were modified to contain mouse κ leader sequences. cDNA of the remaining antibody chains was synthesized (GeneArt) based on the published nucleotide sequences and cloned into the expression vectors. 1F1 Fab was produced by limited proteolysis of 1F1 IgG by endoprotease Lys-C. Digests containing ∼0.25 µg Lys-C per 1 mg 1F1 IgG in 25 mM Tris, 1 mM EDTA, pH 8.5 were incubated at 37°C for 4 hours, then stopped by the addition of TLCK and leupeptin to a final concentration of ∼1 mM and 0.4 mM, respectively. Alternatively, Fabs were expressed recombinantly by introducing a stop codon into the heavy chain gene immediately after the codon for the cysteine of the hinge disulfide. DH5α cells were transformed with plasmids for EndoFree Maxi preparation (Qiagen). Purified DNA was co-transfected transiently into HEK 293F cells (Invitrogen) using PolyFect reagent (Qiagen) in disposable shaker flasks. The supernatant was harvested on day seven and purified through a gravity column with CaptureSelect Fab λ resin (BAC B.V., GP Naarden, The Netherlands) in D-PBS for Fabs 1F1, 2B12, 2D1, and 4D20 or purified on an ÅKTA FPLC instrument (GE Healthcare Life Sciences) using HiTrap Protein G columns (GE; all other proteins) and concentrated with 15 mL centrifugal filter units with 30 kD molecular weight cut-off (Millipore, Billerica, MA). The purity of all expressed proteins was assessed using reducing, denaturing SDS-PAGE gel electrophoresis (Invitrogen).
The SC1918 and Av1918 HAs were expressed using the baculovirus system and purified essentially as previously described [1]. Following initial capture of 1F1 Fab from Lys-C digest of IgG (Protein G affinity chromatography) or from cell culture supernatant from recombinant expression (a-lambda), Fab was further purified by cation exchange chromatography (MonoS, GE Healthcare) in sodium acetate buffer, pH 5.0 and a linear NaCl gradient from 0–1 M. Fractions containing Fab were buffer exchanged into 10 mM Tris, pH 8.0, 150 mM NaCl and subjected to gel filtration (Superdex 200, GE Healthcare). Purified SC1918 HA was mixed with a stoichiometric excess of recombinant 1F1 Fab and the complex was isolated from free Fab by gel filtration.
Initial crystallization screening for 1F1 Fab (14 mg/mL), Av1918 HA (28 mg/mL), and the 1F1-SC1918 complex (9.6 mg/mL) was conducted using the robotic CrystalMation system (Rigaku) at the Joint Center for Structural Genomics (JCSG; www.jcsg.org). Diffraction quality crystals were subsequently grown in sitting drops by vapor diffusion (0.5 µL protein solution +0.5 µL well solution with 1 mL reservoir for 1F1 Fab and 1F1-SC1918 complex; 100 nL protein solution +100 nL well solution with 200 µL reservoir for Av1918 HA). Crystals used for data collection were grown at 20°C from 20% PEG 4000, 200 mM dibasic sodium phosphate (1F1 Fab); 4°C from 100 mM Tris pH 8.0, 40% MPD (2-methyl-2,4-pentanediol (Av1918); or 20°C from 8.5% PEG 6000, 100 mM Tris pH 7.9 (1F1-SC1918 complex). Crystals were cryprotected in the mother liquor (Av1918), with well solution supplemented with 15% glycerol (1F1 Fab), or with 25% ethylene glycol (1F1-SC1918 complex) and flash cooled in liquid nitrogen.
Diffraction data were collected on the General Medicine/Cancer Institutes Collaborative Access Team (GM/CA-CAT) beamline 23ID-D at the Advanced Photon Source at Argonne National Laboratory (Av1918) and on beamline 11-1 (1F1 Fab and 1F1-SC1918 Complex) at the Stanford Synchrotron Radiation Lightsource (SSRL). The data were indexed integrated and scaled using HKL2000 (HKL Research), and merged using Xprep (Bruker). The structures were solved by molecular replacement using Phaser [35]. Structures from PDB codes 2FB4 and 1RZF (1F1 Fab, variable domains and constant domains, respectively); 1RUZ (Av1918), or the 1F1 Fab and Av1918 coordinates reported here (1F1-SC1918 complex), were used as search models and total of 2 Fabs (1F1 Fab), 1 HA trimer (3 protomers) (Av1918), or two HA trimers (six protomers), 2 Fabs, and 4 sets of VH/Vλ domains (1F1-SC1918 complex) were ordered in the asymmetric unit. In the 1F1-SC1918 complex, attempts to place the remaining 4 sets of CH1/Cλ domains were unsuccessful, and very little density was observed for the missing protein components after refinement, suggesting that these domains are disordered in the crystal. As VH and Vλ are joined to CH1 and Cλ by flexible linkers, the relative orientations of the variable and constant domains are not rigidly defined and, in the absence of stabilizing crystal contacts, the constant domains can likely adopt an ensemble of conformations in the relatively open crystal lattice. Rigid body refinement, simulated annealing and restrained refinement (including TLS refinement) were carried out in Phenix [36]. Riding hydrogens were used during refinement. Between rounds of refinement, the model was built and adjusted using Coot [37]. Waters were built automatically into the 1F1 Fab and Av1918 HA models using the “ordered_solvent” modeling function in Phenix [36]. Refinement statistics can be found in Table S2. A depiction of the representative electron density at the 1F1-HA interface can be found in Figure S3.
The coordinates and structure factors for 1F1 Fab, Av1918 HA, and the 1F1-SC1918 complex have been deposited in the Protein Data Bank (PDB) with accession numbers 4GXV, 4GXX, and 4GXU, respectively.
Expression plasmids encoding the parental or mutated 1918, 1943 or 1977 HA proteins were co-expressed with an N1 neuraminidase to produce VLPs in 293T cells, as described previously [18], [38]. Briefly, VLP were generated by co-transfection of 293T cells with 1 µg each of expression plasmids for HA and NA. Two days post-transfection, supernatants were collected. HAI assays were performed as described [39]. Briefly, serially diluted antibodies were pre-incubated with 8 hemagglutinating units of virus or VLP per well. Chicken red blood cells were added to a final concentration of 0.5% and the plate was incubated on ice for 30–60 min.
Antibody escape mutant influenza A/Weiss/43 and A/USSR/92/77 viruses were selected [15], [40]. Briefly, escape mutant viruses were selected by treatment of virus with excess antibody, followed by recovery of inhibition resistant viruses in 10-day-old embryonated chicken eggs. RNA was extracted from virus-infected allantoic fluid, then cDNA was generated by RT-PCR, directly cloned, sequenced, and aligned to previously determined wt virus HA gene sequences.
Binding affinity of recombinant 1918 Fabs to recombinant trimerized His-tagged HA protein containing the sequence of 1918 wt or its avianized variant strains was measured using anti-Penta-HIS tips on the Octet QK platform (FortéBio, Menlo Park, CA). The soluble HA protein was expressed and purified as described [41]. Data were calculated using Origin 7.5 SR6 software (OriginLab Corp., Northampton, MA) based on automated curve fittings prompted by the Octet 4 software (ForteBio) using a 1∶1 binding model. All measurements are listed as the average of two duplicate measurements. The Fabs were diluted to a concentration of 60 µg/mL (1F1, 1I20, 2B12) or 30 µg/mL (2D1, 4D20). Curve fittings and experimental errors can be found in the supporting information (Figure S2, Table S3).
|
10.1371/journal.pbio.1000154 | The Neuropeptide PDF Acts Directly on Evening Pacemaker Neurons to Regulate Multiple Features of Circadian Behavior | Discrete clusters of circadian clock neurons temporally organize daily behaviors such as sleep and wake. In Drosophila, a network of just 150 neurons drives two peaks of timed activity in the morning and evening. A subset of these neurons expresses the neuropeptide pigment dispersing factor (PDF), which is important for promoting morning behavior as well as maintaining robust free-running rhythmicity in constant conditions. Yet, how PDF acts on downstream circuits to mediate rhythmic behavior is unknown. Using circuit-directed rescue of PDF receptor mutants, we show that PDF targeting of just ∼30 non-PDF evening circadian neurons is sufficient to drive morning behavior. This function is not accompanied by large changes in core molecular oscillators in light-dark, indicating that PDF RECEPTOR likely regulates the output of these cells under these conditions. We find that PDF also acts on this focused set of non-PDF neurons to regulate both evening activity phase and period length, consistent with modest resetting effects on core oscillators. PDF likely acts on more distributed pacemaker neuron targets, including the PDF neurons themselves, to regulate rhythmic strength. Here we reveal defining features of the circuit-diagram for PDF peptide function in circadian behavior, revealing the direct neuronal targets of PDF as well as its behavioral functions at those sites. These studies define a key direct output circuit sufficient for multiple PDF dependent behaviors.
| Animals depend on being awake at the right time of day to find food and mates and fend off predators. Circadian pacemaker neurons in the brain play a crucial role in timing of specific behaviors to the appropriate times of day. These neurons are further specialized to those primarily responsible for morning and evening behavior. We have used the fruit fly Drosophila as a simple model system to elucidate the neural circuits important for timed daily behavior. In flies, a small group of clock neurons devoted to morning behavior express a neuropeptide, PIGMENT DISPERSING FACTOR (PDF). Until now it was unclear what the direct neural targets of this peptide are and how its actions at those targets mediate timed behavior. Here we find that the so-called morning clock neurons communicate directly to other clock neurons, those responsible for evening behavior. This communication sustains high amplitude morning activity and sets the phase of evening activity as well as the period of activity rhythms in constant conditions. These studies reveal the circuit diagram for PDF function in circadian behavior.
| Circadian clocks act in many organisms to promote daily rhythms of behavior and physiology. In Drosophila, clock function under conditions of light-dark entrainment (12-h light∶12-h dark; LD) is evident as increases in locomotor activity in advance of lights-on (morning anticipation) and lights-off (evening anticipation). These rhythms are driven by well-conserved transcriptional feedback loops in which the basic helix-loop-helix transcription factor heterodimer, CLOCK/CYCLE, activates components such as period (per), timeless (tim), and clockwork orange (cwo) that feedback and regulate CLOCK/CYCLE binding to its cognate DNA targets [1]–[4]. These feedback loops generate daily gene expression rhythms.
Approximately 150 pacemaker neurons in the adult Drosophila brain are implicated in the regulation of circadian locomotor behavior. These neurons can be roughly divided into the PIGMENT DISPERSING FACTOR (PDF)-expressing small and large LNv (sLNv, lLNv), a single non-PDF sLNv, the dorsal lateral neurons (LNd), and three groups of dorsal neurons (DN1, DN2, and DN3) [5]. Ablation of PDF+ neurons results in substantial reduction in morning anticipation [6],[7]. A functional clock in the small subset of PDF+ neurons is sufficient to drive morning behavior, and these cells have thus been dubbed “morning” (M) cells [8]. The large LNv have been observed to promote arousal especially during the light period [9]–[11]. A subset of ∼30 circadian pacemaker neurons, including the non-PDF sLNv, LNd, and/or a small subset of DN1s and DN3s [7],[8],[12],[13], are essential for evening anticipatory behavior, and are thus dubbed “evening” (E) cells. Mammalian circadian clocks may also have a similar morning and evening organization [14],[15].
Drosophila also maintains robust locomotor activity rhythms during constant-dark conditions (DD), reflecting the endogenous function of its circadian clock. The PDF-expressing LNv play a critical role in sustaining free-running rhythms, as ablation of the PDF+ LNv leads to decreased DD rhythmicity [6]. Moreover, tissue-specific rescue experiments indicate that the circadian clock component PERIOD (PER) [7] and the circadian output ion channel NARROW ABDOMEN (NA) [16] are each required in the PDF+ LNv to promote robust, sustained DD rhythmicity. The function of PDF neurons is instructive, as selectively altering the period of these cells drives changes in period length in several non-PDF neurons and sets the circadian period of locomotor activity [17]. It is not known if the ability of PDF neurons to influence non-PDF pacemaker neurons reflects a direct cellular connection.
The PDF neuropeptide is implicated as the principal transmitter of the LNv group, as flies lacking Pdf function exhibit phenotypes similar to ablation of the PDF+ LNv [6]. In LD, these phenotypes include reduced morning behavior and advanced evening behavior. During DD, null Pdf01 mutants exhibit progressive dampening of locomotor rhythmicity and a slightly shortened period. A receptor for Drosophila PDF has been identified (PDFR, aka han, groom-of-pdf, CG13758), and loss of this receptor leads to circadian phenotypes essentially identical to Pdf01 mutants [18]–[20].
The DD behavioral phenotypes of Pdf01 mutants are accompanied by alterations in the molecular clock. PER oscillations in the DN1 of Pdf01 mutants rapidly damp during DD, indicating a role for PDF in sustaining molecular rhythms [21]. In contrast, the LNd of Pdf01 mutants exhibit persistent rhythms, but with an advance in the phase of PER oscillations, consistent with the observed short behavioral period of these flies [22]. Additionally, desynchronized PER nuclear localization rhythms are observed in the sLNv of Pdf01 mutants, but only after many days of DD [22]. These data suggest that PDF may also reset or synchronize these molecular clocks. However, molecular alterations have not been observed in Pdf01 mutants in LD [23], suggesting that PDF may be acting downstream of the molecular clock under these conditions.
While the molecular consequences of manipulating PDF/PDF RECEPTOR (PDFR) function have been well described, it was not previously known which of these effects reflected the direct actions of PDF on the affected cells or whether they were mediated by cellular intermediates. In addition, it was not known which of these direct cellular targets was mediating the multiple effects of PDF on behavior, especially under LD conditions. Here we demonstrate that PDFR expression limited to the ∼30 non-PDF evening cells can not only alter the timing of evening behavior, but also drive the amplitude of morning behavior. Our data indicate that the effect of PDFR expression on morning behavior does not likely occur through the core clock, but instead through the regulation of neuronal output. We also demonstrate a role for PDFR in non-PDF cells to reset evening phase and regulate period length, consistent with core clock resetting. Finally, we find that PDFR likely functions within a more distributed group of pacemaker neurons, including the PDF+ LNv, to promote sustained DD rhythmicity. This study defines the major direct targets for PDF in vivo and their functions in circadian behavior.
To define the neuroanatomical targets of PDF action in circadian behavior, we performed tissue-specific rescue of a Pdfr mutant using the GAL4-UAS system. For these experiments, we utilized a strong loss-of-function mutant allele of Pdfr, han5304. Like null Pdf01 mutants, Pdfr han5304 flies display strongly reduced morning anticipation and phase-advanced evening anticipation in LD, as well as a reduced morning peak at the onset of DD [6],[18]. Previous studies had suggested that PDFR functions in circadian neurons largely based on partial rescue using a single perGAL4 driver [18]. perGAL4 drivers, in addition to demonstrating expression in all major circadian pacemaker groups, also drive widespread expression in nominally noncircadian brain areas, including the central complex, antennal lobe, and lateral horn [24], raising questions as to the precise site of PDFR function. To address this issue, we utilized clockGAL4 [25], which drives broad expression among all major circadian neuronal groups [16] but relatively limited noncircadian expression, including the pars intercerebralis (PI) and cells surrounding circadian neurons [16]. Using this driver, we find that PDFR expression in Pdfr mutants rescues morning anticipation and the proper timing of LD evening behavior (Figures 1A–1C and S1; p<0.05). Given the relatively limited noncircadian expression of clockGAL4, these results suggest a major function for PDFR in circadian neurons.
We next assessed PDFR function specifically in the pacemaker neuron subsets known to control morning and evening behavior. We performed rescue using a GAL4 driver containing the promoter and first intron of the cryptochrome gene (cryGAL4-13) [26]. cryGAL4-13 drives expression in both PDF-expressing morning cells and ∼30 non-PDF evening cells (LNv, LNd, small subset of DN1 and DN3), while promoting little or no expression in other circadian pacemaker neurons (e.g., most DN1, all DN2, and most DN3) or outside the circadian system [7],[13],[16]. cryGAL4-13 driven expression of UAS-Pdfr restores the timing of evening behavior (p<0.0001) and also promotes significant restoration of morning behavior during LD and the first day of DD (DD1; Figures 1D, 2A–2C; Table 1; p<0.001). We examined and quantified morning behavior during DD1 as the lights-on response in LD can mask some of the clock-driven morning behavior. We then further restricted UAS-Pdfr expression specifically to either the evening cells or morning cells. Expression was restricted to evening cells by blocking GAL4 induction selectively in PDF+ cells using GAL80 (PdfGAL80; cryGAL4-13), while morning cell-specific expression was driven using PdfGAL4. Expressing UAS-Pdfr only in non-PDF evening cells rescues both the timing of evening behavior and the magnitude of morning anticipation (Figures 1E and 2D; Table 1; p<0.0001). In contrast, UAS-Pdfr expression restricted to morning cells does not have comparable effects on morning or evening behaviors (Figures 1F and 2E; Table 1), as previously reported [18]. These findings suggest that the PDF+ LNv can communicate directly to the non-PDF “evening” cells through PDFR to promote morning behavior.
We next examined whether the behavioral contribution of evening cells to morning behavior might be driven by changes in the circadian clock. The etiology of circadian phenotypes in flies with disrupted PDF signaling has largely focused on the role of PDF in synchronizing and/or resetting circadian clocks. These studies have largely identified changes in molecular oscillations of core clock components, such as PER, which reflect core clock timing, under constant darkness conditions. It has been proposed that light can compensate for the loss of PDF/PDFR as no large changes in the core clock have been described in Pdf01 mutants in LD [23]. However, these experiments were performed with only two time points.
Given our interest in determining the molecular basis of morning and evening behavioral phenotypes in LD, we performed PER immunolabeling in wild-type (UAS-Pdfr/+) and Pdfr mutant (Pdfr han5304; UAS-Pdfr/+) flies during LD using four time points. Previous studies have shown that PER expression restricted to PDF neurons is sufficient to rescue morning anticipation of per01 mutants, suggesting that PDF actions do not require clock function in non-PDF neurons for morning behavior [8]. We asked whether changes in the LNv clock could account for loss of morning behavior in Pdfr mutants. However, no significant differences in PER oscillations were observed in the PDF+ sLNv important for morning behavior (Figure 3A–3C). While small changes were observed in some pacemaker neuron clusters (lLNv, LNd, and DN3), high amplitude oscillations were observed in all pacemaker neurons groups in Pdfr mutants, in contrast to the highly significant reduction in morning behavior (Figure 2). In addition, clock function in the lLNv is not required for morning behavior [7],[8],[27], suggesting (but not excluding) the possibility that these small molecular changes do not underlie changes in morning behavior. The LNd and a subset of DN3 have been implicated in regulating evening behavior [7],[13], and subtle changes in the LNd and/or DN3 may be responsible for the ∼2-h phase advance in evening anticipation (Figure 1B). A higher temporal resolution will be necessary to definitively demonstrate a molecular phase shift in these cell clusters. Nonetheless, relatively small phase changes are unlikely to explain large amplitude changes in morning behavior. In this case, the dramatic effects of PDFR on morning behavior largely reflect its function in circadian output.
In addition to defects in morning and evening behaviors, Pdf01 and Pdfr mutants exhibit decreased rhythmic power and shortened period length in DD [6],[18],[19]. To determine whether the anatomical requirements for PDFR function in free-running rhythmicity match those for morning and evening behaviors, we assessed DD rhythms in PDFR rescue flies. Expression of PDFR using a broad circadian driver clockGAL4 promotes significant rescue of DD rhythmicity, as reflected by rhythmic power (p<0.0001) and period length (p<0.0001; Table 2). Period length of clockGAL4 rescue flies is slightly short (23.3+/−0.1 h), yet comparable to clockGAL4 driven overexpression of PDFR in a wild-type background (23.4+/−0.1 h; Table 2), likely due to a modest (∼30 min) overexpression effect (Table 2). Nonetheless, clockGAL4 rescue of period is statistically significant and supports a role for PDFR in circadian neurons to promote normal DD period and rhythmicity.
To assess PDFR DD function in specific circadian neuron subsets, we analyzed DD rescue using GAL4 drivers with limited circadian expression. Expression of PDFR in morning and evening cells using cryGAL4-13 rescues DD period length and partially restores DD rhythmicity (Table 2; p<0.0001), suggesting broader or stronger DN expression provided in clockGAL4 may be needed for robust rhythmicity. However, further restriction of PDFR expression to evening cells using PdfGAL80 fully blocks the cryGAL4-13 rescue of rhythmic power (p = 0.6; Table 2). While we cannot rule out residual GAL4 activity, these data are consistent with full GAL80 repression of GAL4 activity in the LNv. These data uncouple LD and DD rescue and suggest a role for PDFR within the PDF+ LNv to promote DD rhythms. Yet consistent with previous findings [18], PDFR expression restricted to PDF+ neurons (PdfGAL4) has no significant effect on either free-running rhythmicity or period length in DD, indicating that PDFR function in these cells is not sufficient for normal DD rhythms (Table 2). To confirm a role for PDFR in the PDF+ neurons, we expressed PdfGAL80 in the context of pan-neuronal elavGAL4-mediated rescue. elavGAL4 expression results in strong rescue of all LD and DD phenotypes (Figure 1G, Table 2; p<0.0001). As with cryGAL4-13, blocking elavGAL4 driven PDFR selectively in the PDF+ LNv using PdfGAL80 results in a substantial reduction in rhythmic power (Table 2; p<0.0001). Taken together, these data suggest that PDF/PDFR communication within the LNv plays an important role in sustaining robust DD rhythmicity. In addition, our rescue also suggests that other cells also contribute to DD rhythmicity.
Notably, period length is also significantly rescued with all GAL4 drivers tested except PdfGAL4 (Table 2; p<0.0001). Yet unlike rhythmic power, period length rescue is unchanged when PDFR expression is blocked in PDF+ neurons via PdfGAL80. In fact, PDFR expression restricted only to evening cells (PdfGAL80; cryGAL4-13) retains significant rescue of period length, as evident from group activity profiles (Figure 4) and individual fly analyses (Table 2). This remains true even if only strongly rhythmic flies (Power≥40) are considered (unpublished data; p<0.0001). Thus, direct PDF communication among PDF-expressing neurons, as well as with other target neurons, is important for sustaining DD rhythms. In contrast, the PDF+ LNv communicate directly to non-PDF evening cells to set period length, indicating functional and anatomical specialization of PDF signaling.
Taken together, our functional neuroanatomical approaches highlight PDFR function in circadian pacemaker neurons. However, reports of the PDFR expression pattern are conflicting. Two initial reports utilized independently generated antisera to assess PDFR expression in the Drosophila brain. One reported expression limited mainly to circadian neurons [18], while the other observed broad expression that included only few circadian neurons [20]. We previously reported pdfr expression using in situ hybridization and noted expression in potential dorsal neurons and the PI [19]. A more recent report indicates that the reported immunofluorescence patterns may not represent specific PDFR signal [28], calling into question the true PDFR expression pattern. We have made several additional attempts to generate specific antisera to PDFR but have yet to identify reproducible and robust signals (unpublished data).
To examine Pdfr expression, we instead used a P-element exchange strategy to insert a P{GAL4} element ∼40 bp upstream of the presumptive Pdfr transcription start site (PdfrGAL4; see Materials and Methods) [29]. A targeted GAL4 insertion into a locus of interest has been a valuable approach to report endogenous gene expression patterns [30]. If the GAL4 insertion falls under the control of enhancers that normally drive Pdfr expression, we predict that PdfrGAL4 will reflect endogenous Pdfr expression. In this case, PdfrGAL4 if combined with UAS-Pdfr should be able to rescue Pdfr mutant phenotypes. The original insert used to generate PdfrGAL4 displayed a modest circadian rhythmicity phenotype [20] and a ∼50% reduction in transcript levels (unpublished data). Consistent with these data, we find that PdfrGAL4/Pdfrhan5304 flies display poor DD rhythmicity (Table 2). Importantly, this reduced rhythmicity is strongly rescued by PdfrGAL4 driven expression of UAS-Pdfr (Table 2; p<0.0001), suggesting that PdfrGAL4 is a faithful reporter of Pdfr expression.
We then examined the driven expression pattern for PdfrGAL4. Upon crossing PdfrGAL4 to UAS-nuclear green fluorescent protein (GFP) (nGFP), we observe broad GFP expression in the adult Drosophila brain, including circadian neuron regions, PI, optic lobe, and ellipsoid body (Figure 5A), the latter possibly consistent with noncircadian functions of PDF in arousal and geotaxis [9]–[11],[20],[31]. Given our rescue data, we more closely examined expression within circadian pacemaker neurons. To directly assess circadian expression, we labeled PdfrGAL4 UAS-nGFP brains with PER antisera. We observe prominent GFP expression in the sLNv, all LNd, and several DN1 (Figure 5B and 5C). The PI and DN expression is consistent with our published in situ expression pattern [19]. Weak expression is observed in the lLNv. We also consistently observe expression in two DN3s, and we sometimes observe expression in one of the two DN2s (unpublished data). These expression data nicely complement our functional neuroanatomy data. Expression in the sLNv is consistent with a role for PDFR in these cells to sustain free-running rhythmicity (see PdfGAL80), as the sLNv are known to be especially important for DD rhythmicity [8],[17]. Pdfr expression in the LNd, DN1, and DN3 subset is consistent with our data demonstrating an important role of the non-PDF evening cells in morning and evening behavior and DD period length.
Here we define the direct targets of PDF using circuit-specific rescue and find that the direct action of PDF on just ∼30 neurons, the so-called evening pacemaker neurons, mediates PDF dependent effects on morning, evening, and free-running behaviors. We corroborate our functional rescue data with a novel GAL4 enhancer trap reporting endogenous pdfr expression. We also provide strong evidence that PDF, in addition to its well-described effects on the core clock mechanism, also likely affects the output of pacemaker neurons providing novel mechanistic insight into PDFR function. These studies define a major direct conduit for in vivo PDF signaling in circadian behavior.
A number of reports have examined the molecular consequences of manipulating PDF neuron function. Altering the core clock, output, or projections of PDF neurons alters the molecular clock in non-PDF circadian neurons and evening behavior under short days or in constant darkness [17],[19],[22],[23],[32]–[36]. However, these studies leave open a number of key questions important for elucidating the PDF circuit diagram. Not surprisingly, functional changes in PDF neurons can be propagated widely through the nervous system, not only to the direct cellular targets of that group of neurons (primary target neurons), but to the targets of those targets (secondary), and so on (tertiary). Thus, the direct and indirect effects of PDF could not be distinguished in these papers. In addition, these studies do not identify the behavioral functions of PDFR (or PDF) at these different cellular targets particularly on LD morning and evening behavior. Some of these studies also rely on analysis of mutant flies with significant developmental abnormalities [34],[36].
Measurements of PDF activation in ex vivo brains have also been used to infer direct cellular targets [28]. Bath application of PDF to cultured brains up-regulated cAMP levels. However, these assays required ∼1 min to observe significant activation. Given the slow response time course relative to the faster rate of synaptic transmission, PDF effects on a primary target neuron could be propagated through circuitry to secondary neurons to increase cAMP, on a similar minute time course. Thus, one cannot exclude the possibility that some of the observed responses may be indirect. In addition, effects might even reflect direct responses to nonphysiological levels of PDF. Moreover, this study does not address the behavioral functions of PDF at those sites. By using the direct molecular target of PDF, the PDF receptor, to rescue Pdfr mutant phenotypes, we functionally define these direct neuronal targets in vivo. We demonstrate that the expression of PDFR in a highly restricted group of neurons (∼30 neurons) is sufficient to rescue morning behavior, evening phase, and circadian period, thus defining a major direct output circuit for multiple PDF-dependent behaviors (Figure 6).
How does PDF function at these neuronal targets? Previous studies have identified molecular clock changes especially under constant darkness conditions indicating that PDF acts to reset molecular oscillators. Consistent with this model, we observed that PDFR expression in the E cells can rescue circadian period and evening activity phase. Moreover, we identified subtle molecular changes in E cells in LD that are consistent with a small phase shift in molecular oscillations. Thus, at least some PDF-dependent behaviors can be attributed to its function in resetting clocks.
While there are PDF effects on core clocks, our data also suggest an additional output function particularly in regulating morning behavior (Figure 6). Both Pdf and Pdfr mutants have been shown to have strong effects on the amplitude of morning behavior. Our studies similarly demonstrate major changes in the amplitude of morning behavior despite robust oscillations in both the sLNv (which are sufficient for morning behavior) as well as other circadian cell groups including the E cells. The published data further support this model. The finding that PDF can acutely affect neuronal firing rate in other insects [37] strongly suggested that such clock-independent output functions were possible, if not likely, as core clock changes and their subsequent translation into neuronal firing changes would take place over a longer time frame. It has previously been shown that rescue of clocks exclusively in the PDF neurons in the arrhythmic per01 mutant rescues morning behavior [8]. If morning behavior works by PDF targeting of the E cells, then PDF must act on the output of the E cells in these flies, as there is no clock in the E cells. In addition, manipulation of sodium channel activity shifts the phase of PDF rhythms and morning behavior but these are not accompanied by shifts in molecular oscillators in the sLNv, LNd, or DN1, consistent with an output function [33]. Taken together, we believe our data coupled to the published literature support the notion that PDF can affect the output of E cells in addition to its phase resetting effects.
Previous data have suggested that PDF activates MAPK phosphorylation in the dorsal brain just prior to the increase in morning activity [38], suggesting that PDF release may promote morning activity. Consistent with this hypothesis, recent data suggest a role for the PDF expressing large LNv in driving locomotor activity and arousal [9]–[11],[39]. These effects may be mediated through the sLNv, which in turn project to the dorsal brain [10],[11]. PDF release in the morning may also reset oscillators in the E cells (Figure 6).
Our identification of a role for so-called “E” cells in M behavior, also fits well with prior data suggesting that E cells can control M behavior and highlights additional complexity of the M-E model. Manipulating the clock in E cells can shift morning behavioral phase under long photoperiods [40], whereas rescue of the arrhythmic per0 mutant in non-PDF neurons can rescue morning behavior [7]. However, these results were interpreted to indicate that E cell clocks signal through M cell clocks to drive morning behavior. Indeed, these authors proposed that M cells signal through unknown circuits to drive morning behavior [7]. Here we demonstrate that the E cells themselves are direct targets of the M cells to drive morning behavior. Given our data, E cells may signal to other pacemaker neurons or even nonpacemaker neurons rather than to M cells to drive morning behavior.
How then does one reconcile the apparent observation that clock function is sufficient in E cells to drive morning behavior with the observation that E cells are not necessary for M behavior [7]? One possibility is that redundant pathways control morning behavior. Thus, PDF communication to E cells is sufficient, but may not be necessary, to drive morning behavior. Nonetheless, these data demonstrate that the function of M and E cells is more intertwined than previously thought, necessitating a revision of the simplest versions of the M-E model.
As E cells constitute a focused yet heterogeneous group of cells [7],[13],[35],[41],[42], it will be of interest to determine whether distinct subsets of them are responsible for E and M behavior. E cells consist of the non-PDF small LNv, two DN3s, the LNd, which can be further subdivided by their expression of Neuropeptide F (NPF) [43], and a subset of DN1s, two of which persist from larval development and the remainder that express the transcription factor GLASS [21]. We have attempted to rescue Pdfr mutant phenotypes using PdfGAL4 and npfGAL4 combined, but we fail to observe significant rescue of any LD or DD phenotypes (unpublished data), suggesting a role for PDFR in E cells other than the NPF-expressing LNd. The DD period is likely driven from some or all LNd, as DN1 rhythms of Pdf01 mutants rapidly damp in DD while LNd rhythm persist with a short period for several days in DD [22], comparable to the period of DD locomotor rhythmicity in Pdf01 or Pdfr mutants. Moreover, since the residual DD rhythms in Pdf01 and Pdfr mutants occur in the evening, we propose that the LNd may contribute to the phase advanced LD evening behavior in these flies. Nonetheless, disco mutant flies that lack intact LNs but retain DNs also retain evening anticipation; this suggests redundant LN and DN pathways for evening behavior [44]. GLASS+ DN1s are missing in glass mutant flies and these flies display an intact evening peak but an altered morning peak, in that this peak is poorly entrained and variable in phase [45]. This suggests that the GLASS+ DN1 may be important for morning behavior. Additional functional cell-specific reagents will be necessary to assess the relative contribution of the PDF-sLNv, LNd, DN1, and DN3 in PDF-dependent circadian behaviors.
While our data suggest that the E cells are an important conduit for PDF action in the brain especially for circadian period, phase, and morning behavior, we also find that multiple targets are likely important for regulating rhythmic strength (Figure 6). In E cell only rescue, we do not observe significant rescue of rhythmic strength, indicating that other cells are relevant. Knockdown of pan-neuronal rescue in PDF neurons substantially reduces rhythmic strength (Table 2). On the other hand, PdfGAL4-mediated rescue does not rescue DD rhythmicity. Thus, PDFR function in PDF neurons is necessary but not sufficient for DD rhythmic strength. Based on our expression analyses of PdfrGAL4 (Figure 5) and PDF responsiveness by PDF application [28], these target cells are likely the PDF+ small LNv. We have observed a similar function for the LNv in regulating rhythmic strength in tissue-specific rescue of na mutants [16]. Desynchronized molecular rhythms in these cells may contribute to the reduction in rhythmic strength observed in Pdf01 mutants [22].
Importantly, PDF neurons are not the only targets of PDF relevant to sustaining DD rhythms. Expression in broader sets of neurons including E cells (cryGAL4-13), most circadian pacemaker neurons (clockGAL4), and all neurons (elavGAL4) results in progressively increasing levels of rhythmicity (Table 2). In addition, PdfGAL80 knockdown of pan-neuronal rescue does not suppress rhythmicity to mutant levels, further highlighting the role of both PDF neurons and non-PDF neurons in DD rhythmicity.
The rescue data and PdfrGAL4 pattern presented here are also largely consistent with a report on PDF-responsiveness in the adult Drosophila brain [28]. Shafer et al. [28] observe PDF responsiveness in each of the circadian neuron groups (PDF+ sLNv, non-PDF sLNv, lLNv, LNd, DN1, DN2, DN3), albeit only weak responsiveness in a subset of lLNv assayed. The LN responsiveness matches PdfrGAL4 quite well, as we observe PdfrGAL4/UAS-GFP expression in all sLNv, all LNd, and weakly in a subset of lLNv (Figure 5B). Among the DN clusters, we observe PdfrGAL4/UAS-nGFP in approximately half of the DN1 (Figure 5C), reproducibly in two DN3, and occasionally in one of the two DN2 (unpublished data). Whereas Shafer et al. report PDF responsiveness in most DN cells assayed, these experiments were performed using a cryGAL4-39/UAS-Epac-cyclicAMP reporter. cryGAL4-39 expression has been reported to include only a subset of DN1s and DN3s, and (in some reports) DN2s, comparable to the DN pattern we describe for PdfrGAL4 [21],[28],[42]. Moreover, as noted above, these PDF-response measurements could reflect some degree of indirect responsiveness.
Despite the likely complexity of PDF function in circadian behavior, the data presented here define a major direct output pathway for PDF-dependent circadian behaviors. These studies highlight both the function in resetting core clocks as well as communicating timing information downstream of these core oscillators. It will be of interest to further refine the targets in the circadian system as well as define the molecular and cellular mechanisms by which PDF acts on those neural circuits to regulate circadian behavior.
For rescue experiments, either Pdfrhan5304 [18], Pdfrhan5304;; UAS-Pdfr [20], or Pdfrhan5304; elavGAL4 [46] virgin females were crossed to y w, GAL4/GAL80, or UAS males. For overexpression experiments, UAS-Pdfr (line 10) flies were crossed to either y w (control) or specific GAL4/GAL80 strains. For PdfrGAL4 rescue experiments, female progeny were used for behavioral assays. For all other behavior, male progeny were assayed.
Locomotor activity levels were monitored using Trikinetics Activity Monitors for 5 d of LD followed by 7 d of DD at 25°C. For LD analyses (Figure 1), activity levels from each fly were normalized and averaged within genotypes over 4 d, as described previously [47]. For DD analyses (Figures 2 and 4), activity levels were normalized and averaged over the last 2 d of LD followed by 7 d of DD. To calculate time of evening anticipation in LD (Table 1), we determined the largest 2-h increase in normalized average activity for each fly over the last 7 h of the light phase. The time designation refers to the end point of the maximal activity increase, as averaged among individual flies in each genotype.
To quantitatively analyze morning behavior, we examined the first day of DD, as the lights-on peak in LD can mask the increase in morning behavior. To calculate DD Day 1 Morning Index (Table 1), normalized activity levels were averaged over three consecutive 30-min time points. For each genotype, maximum average activity of the group was determined for any two consecutive 30-min time points over the 6 h surrounding CT 0 (ZT 21- CT3). Minimum average activity was then determined for all time points before and after the observed maximum activity, up to 7 h before or after CT 0 (ZT 17- CT 7). Morning index value was obtained by subtracting the average of these minimum values from the maximum activity value.
For DD rhythmicity (Table 2), chi-squared periodogram analyses were performed using Clocklab (Actimetrics). Rhythmic flies were defined as those in which the chi-squared power was ≥10 above the significance line. Period calculations also considered all flies with rhythmic power ≥10, with the exception of one outlier removed as indicated. All p-values reported were calculated using Student's two-tailed t-tests.
Male Pdfrhan5304;UAS-Pdfr/+ and UAS-Pdfr/+ flies were entrained for 3–5 d at 25°C and anesthetized with CO2. The flies were dissected in 3.7% formaldehyde diluted in PBS at ZT1, ZT7, ZT12, and ZT18. After fixing for 30 min at room temperature, the brains were rinsed three times in PBS and incubated in PBT (PBS with 0.1% Triton) for 10 min at room temperature. The brains were then incubated with 5% goat serum diluted in PBT for 30 min at room temperature, followed by overnight incubation of 1∶4,000 rabbit anti-PER diluted in PBT containing 5% goat serum at 4°C. After several PBT rinses, the brains were incubated with 1∶500 goat-anti-rabbit AlexaFluor 594 (Amersham) in PBT overnight at 4°C. Final rinses in PBT and PBS were followed by mounting in 80% glycerol diluted in PBS. All slides were coded as to sample identity and remained so until the numerical analysis stage. PER-stained specimens were photographed with 60× oil lens on a Nikon Eclipse 800 laser scanning confocal microscope. For a given experiment the microscope, laser, and filter settings were held constant, and all specimens were photographed in the same microscopy session. PER immunostaining was quantified from digitally projected Z stacks using ImageJ (NIH). PER-stained soma were outlined to obtain average pixel intensity. On each projection image an unstained area was quantified to be used for background subtraction. All background-subtracted intensity measurements within a condition (time and genotype) were averaged. To combine experiments, background subtracted measurements were scaled to ZT1 of Pdfrhan5304;UAS-Pdfr/+ in that experiment. Statistical analysis was conducted in SPSS and Excel using ANOVA.
Targeted transposition was used to replace P{EY11181}, a P-element insertion approximately 40 bp upstream of the Pdfr transcription start site, with P{GawB}, a P-element containing GAL4. To perform targeted transposition, P{EY11181}, P{GawB} CyO flies were crossed to P-element transposase [29]. Strains in which P{GawB} mobilized to the X chromosome were identified by eye color and then analyzed by genomic PCR, to determine whether the GAL4 element had inserted into the pdfr upstream region. One strain (PdfrGAL4-19) was identified using this method, and the insertion position of the GAL4 element was confirmed using inverse PCR (Model Systems Genomics, Duke University). For expression analyses, PdfrGAL4 flies were crossed to UAS-nuclearGFP (UAS-nGFP). Female progeny were entrained, dissected, and labeled with anti-PER protein as previously described [16]. Images were obtained using laser scanning confocal microscopy (Nikon C1).
|
10.1371/journal.pmed.1002620 | Reducing the burden of dizziness in middle-aged and older people: A multifactorial, tailored, single-blind randomized controlled trial | Dizziness is common among older people and is associated with a cascade of debilitating symptoms, such as reduced quality of life, depression, and falls. The multifactorial aetiology of dizziness is a major barrier to establishing a clear diagnosis and offering effective therapeutic interventions. Only a few multidisciplinary interventions of dizziness have been conducted to date, all of a pilot nature and none tailoring the intervention to the specific causes of dizziness. Here, we aimed to test the hypothesis that a multidisciplinary dizziness assessment followed by a tailored multifaceted intervention would reduce dizziness handicap and self-reported dizziness as well as enhance balance and gait in people aged 50 years and over with dizziness symptoms.
We conducted a 6-month, single-blind, parallel-group randomized controlled trial in community-living people aged 50 years and over who reported dizziness in the past year. We excluded individuals currently receiving treatment for their dizziness, those with degenerative neurological conditions including cognitive impairment, those unable to walk 20 meters, and those identified at baseline assessment with conditions that required urgent treatment. Our team of geriatrician, vestibular neuroscientist, psychologist, exercise physiologist, study coordinator, and baseline assessor held case conferences fortnightly to discuss and recommend appropriate therapy (or therapies) for each participant, based on their multidisciplinary baseline assessments. A total of 305 men and women aged 50 to 92 years (mean [SD] age: 67.8 [8.3] years; 62% women) were randomly assigned to either usual care (control; n = 151) or to a tailored, multifaceted intervention (n = 154) comprising one or more of the following: a physiotherapist-led vestibular rehabilitation programme (35% [n = 54]), an 8-week internet-based cognitive-behavioural therapy (CBT) (19% [n = 29]), a 6-month Otago home-based exercise programme (24% [n = 37]), and/or medical management (40% [n = 62]). We were unable to identify a cause of dizziness in 71 participants (23% of total sample). Primary outcome measures comprised dizziness burden measured with the Dizziness Handicap Inventory (DHI) score, frequency of dizziness episodes recorded with monthly calendars over the 6-month follow-up, choice-stepping reaction time (CSRT), and gait variability. Data from 274 participants (90%; 137 per group) were included in the intention-to-treat analysis. At trial completion, the DHI scores in the intervention group (pre and post mean [SD]: 25.9 [19.2] and 20.4 [17.7], respectively) were significantly reduced compared with the control group (pre and post mean [SD]: 23.0 [15.8] and 21.8 [16.4]), when controlling for baseline scores (mean [95% CI] difference between groups [baseline adjusted]: −3.7 [−6.2 to −1.2]; p = 0.003). There were no significant between-group differences in dizziness episodes (relative risk [RR] [95% CI]: 0.87 [0.65 to 1.17]; p = 0.360), CSRT performance (mean [95% CI] difference between groups [baseline adjusted]: −15 [−40 to 10]; p = 0.246), and step-time variability during gait (mean [95% CI] difference between groups [baseline adjusted]: −0.001 [−0.002 to 0.001]; p = 0.497). No serious intervention-related adverse events occurred. Study limitations included the low initial dizziness severity of the participants and the only fair uptake of the falls clinic (medical management) and the CBT interventions.
A multifactorial tailored approach for treating dizziness was effective in reducing dizziness handicap in community-living people aged 50 years and older. No difference was seen on the other primary outcomes. Our findings therefore support the implementation of individualized, multifaceted evidence-based therapies to reduce self-perceived disability associated with dizziness in middle-aged and older people.
Australian New Zealand Clinical Trials Registry ACTRN12612000379819.
| Dizziness is a frequent complaint amongst middle-aged and older people.
Dizziness can lead to poor health outcomes, including reduced quality of life, depression, and falls.
The increasingly multifactorial nature of dizziness with advancing age makes it difficult to objectively establish a diagnosis and offer effective interventions.
Although few multidisciplinary studies have been conducted, none to date have examined whether tailoring the intervention to the specific cause(s) of dizziness can improve dizziness in middle-aged and older community-dwellers.
We assessed 305 community-dwellers aged 50 years and above who reported dizziness in the past year on a range of questionnaires and tests of sensory and physical function, mental health, quality of life, and cardiovascular function, at baseline and 6 months.
We randomly assigned the participants to usual care (control; n = 151) or to a 6-month tailored, multifaceted intervention (n = 154) comprising one or more of the following: vestibular rehabilitation (35% [n = 54]), cognitive-behavioural therapy (CBT) for anxiety and depression (19% [n = 29]), a home-based exercise programme to train balance and lower-limb strength (24% [n = 37]), and/or medical management (40% [n = 62]).
We found that the multifactorial tailored intervention was effective in improving dizziness-related quality of life but did not affect balance, gait, or the frequency of dizziness episodes over 6 months.
Our pragmatic study findings suggest that prescribing middle-aged and older people with dizziness evidence-based therapies directly targeting their deficits is likely to reduce their dizziness handicap but might not improve their physical function.
Low-dizziness handicap on entry to the study and the inclusion of participants not currently receiving a dizziness treatment might partly explain the lack of intervention effect on balance and gait, as well as explain why we were unable to identify a cause of dizziness in 71 participants (23% of total sample).
We suggest that community-based dizziness clinics use existing healthcare services to implement tailored and multifaceted dizziness interventions based on multidisciplinary assessments.
| Dizziness is common in older people: prevalence rates in the community range between 10% and 30% [1–4] and increase with age [1,2,5]. Dizziness is often associated with a marked increase in self-reported functional disability [1,6], depressive symptoms [3,4], decreased participation in social activities, poor self-reported health, and reduced falls efficacy [4]. In addition, the risk of experiencing multiple falls is significantly heightened among older people who report dizziness in the past [7], and this is likely to translate into an increased number of fall-related injuries. In fact, cross-sectional analysis from the 2008 United States National Health Interview Survey reveals that, among people who had reported a fall in the past 12 months, those who had reported dizziness or balance problems in that time were at 1.5-fold–increased odds of injury from the fall compared with their healthy peers (46% versus 36%) [8], and this association remained significant after controlling for age and sex [9]. Dizziness is a subjective complaint that is commonly referred to by patients as vertigo, light-headedness, imbalance, or a ‘floating sensation’; terms that have traditionally been associated with vestibular, cardiovascular, balance, and psychological disorders, respectively. An additional complication is the variation in the natural course of dizziness that is very dependent on etiology, i.e., differential patterns in the progression from mild to severe forms, the addition of secondary symptoms such as anxiety, and potential spontaneous regression.
There is good evidence from Cochrane reviews, systematic reviews, and large randomized controlled trials for effective therapies addressing aspects of each disorder: vestibular rehabilitation for peripheral vestibular disorders (including repositioning manoeuvres for benign paroxysmal positional vertigo [BPPV]) [10], syncope assessment and management [11,12], balance and strength training for older people [13,14], and cognitive-behavioural therapy (CBT) for patients with psychogenic dizziness [15]. Despite the availability of such effective treatments, a significant number of cases of dizziness remain unresolved—20% to 45%—because the cause of dizziness could not be identified [16,17].
Older people often have symptomatology suggestive of more than one subtype of dizziness [18–21]. Yet multidisciplinary interventions addressing this symptom have been scarce. Most published trials to date have focused on a single therapy: vestibular rehabilitation (e.g., [22–24]), balance training [25], CBT [26], or pharmacological therapy [27]. To our knowledge, only a few trials targeting vertigo/dizziness published to date were multidisciplinary, combining vestibular rehabilitation and CBT [28–30]; balance training and vestibular rehabilitation [31]; and education, vestibular rehabilitation, and CBT [24,32]. Three of these studies [28,30,32], including a large randomized controlled trial [28], pointed to significant improvement in dizziness handicap post intervention. Yet they either involved a highly selected sample of patients (e.g., primary care patients with labyrinthine deficits [28])—precluding a potential tailoring of the intervention—or were of a pilot nature, involving short follow-ups and fewer than 45 participants [30,32].
In summary, despite significant evidence of the multifactorial nature of dizziness with advancing age and the availability of a range of evidence-based therapies addressing multiple aspects of dizziness, no study to date has investigated the effectiveness of a multidisciplinary intervention to improve dizziness in the general older population. Therefore, within a randomized controlled trial design, we designed a comprehensive, multidisciplinary battery of vestibular, cardiovascular, neuromuscular, balance, and psychological assessments to improve the likelihood of obtaining a diagnosis for the symptom of dizziness in middle-aged and older people. The primary aim was to compare the effect of a tailored multifaceted intervention versus usual care on frequency of dizziness episodes, dizziness handicap, balance, and gait in people aged 50 years and over with self-reported dizziness. Secondary aims were to determine the effects of the programme on physiological falls risk, cardiovascular, and psychological measures associated with dizziness.
We conducted a 6-month single-blind, parallel-group randomized controlled trial in a single research centre (Neuroscience Research Australia, NeuRA) in Sydney, Australia. The trial was approved by the Human Research Ethics Committee of the University of New South Wales (HC 12152). All eligible participants provided written informed consent prior to participating in the study. Details of the protocol and study design have been reported previously [33] and can be found in the online supplemental material (S2 Text) and on the following webpage: https://bmcgeriatr.biomedcentral.com/articles/10.1186/s12877-017-0450-3. The study protocol is registered in the Australia New Zealand Clinical Trials Registry ACTRN12612000379819. A data safety monitoring committee involving a NeuRA senior scientist not involved in the study and the resident biostatistician oversaw the study.
We recruited community-living middle-aged and older adults through advertisements; flyers in community facility, hospital, and university noticeboards; articles in newspapers and newsletters for older people; the NeuRA website, newsletter, and mailing list; and by mailbox drops within the local community. After an initial phone screening, eligible individuals were invited to participate if they (i) were aged 50 years and over, (ii) had experienced at least one significant episode of dizziness in the past 12 months, (iii) lived independently in the community, and (iv) were able to understand the English language. Individuals were excluded if they (i) had a degenerative neurological condition, (ii) were currently receiving treatment for their dizziness, (iii) had a cognitive impairment (General Practitioner assessment of Cognition [GPCOG] score <5) [34], and/or (iv) were unable to walk 20 metres without difficulty, with or without the use of a walking aid. Participants identified at baseline assessment with conditions that required urgent treatment such as suspected stroke, transient ischemic attack, or other undiagnosed neurological or acute cardiovascular condition, severe depressive, or severe anxiety symptoms (Patient Health Questionnaire nine-item [PHQ-9] [35]: total score ≥20; or Generalized Anxiety Disorder seven-item [GAD-7] [36]: total score of ≥20, respectively; both with or without expression of suicidal thoughts) were also excluded from the study and were referred following consent for appropriate treatment.
Participants were randomly assigned to the intervention or control groups following baseline assessment and case conference, using a computer-generated random number schedule with permuted blocks (sizes: 2, 4, and 6). Allocation was concealed by using central randomization performed by NeuRA personnel not otherwise involved in the study. Outcome assessors, including those monitoring the dizziness episodes, were masked to study group allocation. Due to the nature of the intervention, it was not possible to blind the staff administering interventions or the participants. Participants were instructed not to inform the assessors of their intervention status.
All eligible participants attended NeuRA for a 3-hour baseline assessment undertaken by trained research staff. The assessments included diagnostic tests for descriptive purposes (e.g., clinical tests of vestibular function) and for allocating intervention participants to treatment arms as well as baseline measures for the primary and secondary outcomes. Participants returned to NeuRA for reassessment of outcomes at the end of the trial (6 months post randomization).
Multidisciplinary case conferences were held fortnightly throughout recruitment. At these meetings, consensus recommendations and baseline assessment results were used to prioritize and guide appropriate therapies for each participant [33]. Panel members included a geriatrician, a vestibular neuroscientist, a psychologist, an exercise physiologist, the study coordinator, and a baseline assessor. The intervention plan was guided by published normative data for the sensorimotor, balance, and psychological tests as well as the presence of abnormal results in our vestibular and cardiovascular tests. Due to the multifactorial aetiology of dizziness in older people, many participants required multiple interventions that were implemented in a staged manner within the 6-month follow-up period. The therapies selected to address the different aspects of dizziness—vestibular rehabilitation, Otago home-based exercise programme, CBT, and medical management—have been shown to be effective in addressing aspects of each disorder [10–15], as it was our intention to include evidence-based therapies available within health services in our multifaceted intervention. The interventions are outlined below.
The primary outcome measures captured 4 crucial aspects of the trial: dizziness episodes, dizziness handicap, balance, and walking stability. The total number of dizziness episodes during the 6-month follow-up were monitored with monthly dizziness diaries (S3 Text) and follow-up telephone calls as required. The Dizziness Handicap Inventory (DHI) score was computed from a 25-item scale that assesses an individual’s perception of handicap due to dizziness and encompasses emotional, functional, and physical burden [42]. Increased DHI scores have been significantly associated with low mental and physical health-related quality of life, as well as with increased emotional distress [43]. Scores of 0 to 30 represent mild symptoms, 31 to 60 moderate symptoms, and 61 to 100 severe symptoms [42]. Balance was assessed using the choice-stepping reaction time (CSRT), a validated measure of fall risk that also incorporates strength and reaction time [44]. Based on published data, normal mean times for CSRT should range between 750 to 1,200 milliseconds [44]. Walking stability was assessed by the mean step-time variability (coefficient of variation of step time; seconds), recorded as participants performed 3 walking trials at self-selected speed along on a 4-m-long electronic mat placed in the middle of an 8-m-long walkway [45].
Secondary outcome measures were used to elucidate how the interventions might have assisted in ameliorating dizziness symptoms. These measures included assessments of the following: orthostatic hypotension (positive tilt table test) [46], composite physiological fall risk (physiological profile assessment) [47], dynamic balance (coordinated stability) [48], fear of falling (Iconographical Falls Efficacy Scale) [49], anxiety (GAD-7 scale; GAD-7 score >7 indicates clinically significant anxiety symptoms) [36], depression (PHQ-9 scale; PHQ-9 score >9 indicates clinically significant depressive symptoms) [35], and neuroticism (scale from the NEO Five-Factor Inventory) [50]. Clinical tests of vestibular function including the Dix-Hallpike test of BPPV [10] and the head-shake test for the presence of vestibular asymmetry [51] were also conducted. These vestibular tests were not secondary outcome measures (according to our clinical trial registration and published protocol) but were used for diagnostic purposes and for allocating participants into treatment arms. Therefore, we have not presented the results of these tests in the Results section.
Adverse events (e.g., a fall during an exercise session) were monitored with monthly calendars and telephone calls as required. Uptake and adherence to all interventions was documented with therapist records and/or participant diaries. Deviations from the original protocol (clinical registration) were the inclusion of participants aged 50 to 64 years to increase generalizability to a wider demographic, as well as removal of the vertigo symptom scale as a secondary outcome measure because it was clear early in the study that this scale was not useful in documenting types of dizziness and was poorly completed by participants. Cost-effectiveness of the intervention will be addressed in a subsequent paper.
A power analysis determined that 300 participants (150 per group) were required to provide 80% power to detect a statistically significant 20% between-group difference in the primary outcome measures. For these calculations, we assumed an alpha of 0.05 and a dropout rate of 15%. We assumed the following control group means (SDs) based on values from previous studies of older people: CSRT: 1,322 (331) milliseconds; step-time variability: 0.02 (0.01) seconds; and DHI score: 37 (2) [44,45,52,53]. These data represent clinically significant differences based on relevant randomized controlled trials and cohort studies [22,23,28,52,54,55].
All analyses were completed with an intention-to-treat approach (we analysed all available data in the groups to which participants were allocated), and analysis of the primary outcomes was conducted masked to group allocation (S4 Text). Due to the Poisson-like distribution of the dizziness episodes, these data were contrasted between groups with negative binomial regression adjusting for length of follow-up and the outcome presented as a relative risk (RR) (95% CI). Non-normally distributed continuous variables were presented as median (interquartile range) in addition to mean (SD). Between-group comparisons of retest performance for the continuously scored primary and secondary outcome measures were made using generalized linear models controlling for baseline performance. A priori–specified analyses focusing on participants eligible for the vestibular rehabilitation, exercise, CBT, and/or medical management interventions were also conducted for relevant outcome measures. Multiple imputations of the missing primary outcome data were conducted followed by sensitivity analyses; the results are provided in the supplemental files (S5 Text). Values of p < 0.05 were considered statistically significant. Holm-Bonferroni adjustments for multiple comparisons were also performed for the primary outcome measures. Analyses were conducted using the SPSS Version 24.0 software package (IBM, Armonk, NY).
A total of 424 individuals were assessed for eligibility to participate in the study. Predominantly because they did not satisfy the inclusion criteria or declined to participate (Fig 1), 119 participants were deemed ineligible. Between 15 October 2012 and 2 March 2015, 305 people were recruited, consented to participate, and were then randomly allocated to the control (n = 151) or intervention (n = 154) groups. The included participants (62% women; n = 190) were aged 50 to 92 years (mean [SD]: 67.8 [8.3]). All had intact cognitive function according to their performance on the GPCOG tool of cognitive impairment. Participants in the 2 groups were well-matched with regard to baseline characteristics (Table 1).
Baseline assessments combined with dizziness history identified at least one potential cause of dizziness for most participants (77%; n = 234 of total n = 305: 1 cause, n = 115 [38%]; 2 causes, n = 92 [30%]; 3 to 4 causes, n = 27 [9%]). A total of 126 participants (41%) were identified with vestibular disorders (suspected BPPV [n = 71], other peripheral vestibular problem [n = 48], vestibular migraines [n = 7]), 54 (18%) with clinically significant symptoms of anxiety and/or depression, 79 (26%) with lower-limb weakness and poor balance, and 18 (6%) with multiple comorbidities. In addition, 107 participants (35%) had medical or medication issues that could precipitate dizziness warranting treatment or advice from a general practitioner. In 71 participants (23%), no cause of dizziness could be identified from the assessments, questionnaires, and interviews.
Two participants fainted at baseline assessment (S1 Table) and were medically managed appropriately. No serious adverse events were reported over the course of the multifaceted intervention.
For the 154 participants randomly assigned to the intervention group, 21% were assigned no intervention, 39% assigned a single intervention, 32% assigned 2 interventions, and 8% assigned 3 or 4 interventions (Fig 2).
There was a 79% (n = 45) uptake into the vestibular rehabilitation for those assigned to that therapy (n = 57)—5 participants could not be contacted by the vestibular physiotherapist, 2 declined participation because they lived too far from the assessment site, 2 others declined participation because they were not feeling dizzy anymore, 2 had medical issues, and 1 was not interested anymore. All of those who took up the vestibular rehabilitation programme did attend all their sessions with the vestibular physiotherapist. Uptake for the Otago home exercise programme in those assigned to that intervention (n = 39) was 85% (n = 33)—4 participants discontinued the programme for medical reasons, 1 was erroneously not enrolled in the programme, and 1 moved to a care facility. Regarding adherence to the Otago home exercise programme, 25 of 33 assigned participants (76%) completed 75% or more of the recommended number of exercise sessions (32 of 33 completed 50% or more of the recommended number of sessions). Uptake for the CBT programme was 31% among intervention participants whose anxiety and depression scores met the criteria for inclusion in this therapy (n = 31)—6 declined (1 was too busy, 1 had medical reasons, 1 had just moved to a care facility, 2 were no longer interested, and 1 was unknown), 1 could not be contacted, and 15 expressed interest and enrolled in the course but never completed the modules. All the participants who took up the CBT completed the programme. Finally, among participants requiring medical management, only 2 of 9 intervention participants eligible for a falls clinic appointment took up the recommendation. Reasons for failing to attend the falls clinic intervention included living out of area and declining to attend (1), failing to obtain a referral from their general practitioner (4), concurrent referral by general practitioner to a neurology clinic (1), or already attending a community-health programme (1).
At the completion of the 6-month trial and when controlling for baseline performance, DHI scores in the intervention group were significantly reduced compared with the control group (mean [95% CI] difference between groups [baseline adjusted]: −3.7 [−6.2 to −1.2]; p = 0.003). There were no statistically significant between-group differences in dizziness episode frequency during the follow-up period, when controlling for length of follow-up (RR [95% CI]: 0.87 [0.65 to 1.17]; p = 0.360). There were also no statistically significant between-group differences in CSRT performance (mean [95% CI] difference between groups [baseline adjusted]: −15 [−40 to 10]; p = 0.246) and step-time variability during gait (mean [95% CI] difference between groups [baseline adjusted]: −0.001 [−0.002 to 0.001]; p = 0.497) (Table 2).
The intention-to-treat analysis revealed no significant improvements in composite physiological fall risk and dynamic balance measures, anxiety, depression, neuroticism, fear of falling, or orthostatic hypotension (Table 3).
There were indications for intervention-specific improvements at the 6-month reassessment (S2–S5 Tables). Compared with the control group and controlling for baseline scores, DHI scores were significantly lower in the intervention versus the control group following vestibular rehabilitation (mean [95% CI] difference between groups [baseline adjusted]: −6.3 [−10.2 to −2.3]; p = 0.002), CBT (mean [95% CI] difference between groups [baseline adjusted]: −9.2 [−16.7 to −1.8]; p = 0.015), and medical management (mean [95% CI] difference between groups [baseline adjusted]: −4.6 [−8.4 to −0.8], p = 0.019). At completion of the trial and controlling for baseline performance, compared with their control counterparts, participants in the vestibular rehabilitation group also significantly improved their CSRT performance (mean [95% CI] difference between groups [baseline adjusted]: −43 [−84 to −2]; p = 0.040), and those in the exercise intervention group significantly reduced their composite physiological fall risk (mean [95% CI] difference between groups [baseline adjusted]: −0.37 [−0.68 to −0.06]; p = 0.018). At the 6-month follow-up, the CBT recipients significantly improved GAD-7 scores of anxiety (mean [95% CI] difference between groups [baseline adjusted]: −3.6 [−6.6 to −0.7]; p = 0.015).
This study investigated the effects of a multifactorial tailored intervention on dizziness handicap, frequency of dizziness episodes, and physical function in a sample of middle-aged and older people suffering from dizziness. Overall, our intervention significantly improved dizziness handicap. Given that around 12% of the community aged over 50 years report dizziness [5], using comprehensive objective assessment and individualized evidence-based interventions has the potential to offer a more effective and efficient approach to this common problem.
The need for multifactorial assessment and intervention has been apparent for some time. Nearly 2 decades ago, Sloane and Dallara advocated for the development and implementation of strategies to more effectively reduce symptoms and dizziness disability. They recommended focusing on improving functional ability to increase the quality of life when the cause of dizziness is uncertain and/or when subsequent treatment does not decrease dizziness symptoms [56]. More recently, Maarsingh and colleagues recommended a simultaneous diagnosis- and prognosis-oriented approach for dizzy older people in primary care, whereby therapies can be prescribed based on deficits identified at assessment [57,58]. However, the majority of dizziness interventions have focused on a single cause of dizziness–most often vestibular impairment [22–24,27–29,31,59,60]—with very few studies combining 2 therapies [29,30,32] despite accumulating evidence to show that dizziness becomes increasingly multifactorial in older age [18–21]. Our comprehensive baseline assessment identified 2 or more potential causes of dizziness in 39% of participants, clearly supporting the need for tailored interventions comprising evidence-based therapies as a crucial aspect of an effective dizziness management plan.
The adjusted effect size of the intervention on dizziness handicap was small (0.25) and lower than anticipated in the power calculation (0.37), and potential reasons for this are discussed in the limitations. However, given our sample’s low mean DHI score at baseline (mean [SD]: 25.4 [18.3]), a between-group mean difference of 3.7 points in DHI score is between the 10% score change suggested by Treavalen [61] as clinically relevant for this scale and the 11-point minimally important change computed by Tamber and colleagues [62]. This score difference is also similar to that reported in recent large randomized controlled trials of vestibular rehabilitation (mean between-groups difference in DHI range: 4.3 to 6.15 points) (sample sizes, n range: 170–337) [22,23,28]. The intervention had no significant effects on the remaining primary outcomes: frequency of dizziness episodes, CSRT, and gait variability. In retrospect, these measures may have been suboptimal choices for the interventions tested. The dizziness frequency measure proved to be fluctuating and highly variable, with participants suffering dizziness episodes of markedly different severity and ranging from no episodes over the follow-up period to omnipresent. It was therefore (a) unlikely to have been a uniform measure and (b) difficult to effect change.
The intervention-specific analyses revealed that the individual interventions were effective in managing specific aspects of dizziness, as hypothesized: fall risk (composite physiological function) for the Otago exercise programme [13,63], psychogenic dizziness-related anxiety for the CBT programme [15], and balance (CSRT) for vestibular rehabilitation [23,60]. The faster CSRTs in the group receiving vestibular rehabilitation is particularly encouraging because CSRT is a composite measure of balance, strength, and reaction time and a proxy-measure of fall risk. The improved performance in this dynamic balance test may have resulted from the incorporation of visual motion and head/eye coordination exercises along with exercises to enhance balance confidence.
The fortnightly multidisciplinary meeting was crucial to the success of the intervention because it allowed us to examine the participants’ results from an interdisciplinary perspective, which has been a major limitation of previous research in the field. We recommend that smaller clinics adopt a multidisciplinary approach to the assessment of their patients even if they do not have the specialist on site. Such approach can be done by selecting key measures from each domain.
Strengths of the study included the broad sample of middle-aged and older people reporting dizziness, the multidisciplinary input and objective criteria for assigning participants to interventions, [33] the prospective ascertainment of dizziness episodes, the timely implementation of the interventions, and the pragmatic study design that utilised available healthcare services. The approach therefore represents a model for multidisciplinary assessment and intervention that could be incorporated into clinical practice.
We also acknowledge certain limitations. First, the single-centre nature of the study together with the reliance on self-selected volunteers as participants may limit the generalizability of the study. Second, our intervention comprised one or more individualized evidence-based components. As such, we cannot determine the effect of specific interventions, and it may be that certain interventions were key and others less efficacious. Third, in 23% of the sample, no cause of dizziness could be identified from the assessments, questionnaires, and interviews. Maarsingh and colleagues reported unclear causes of dizziness in only 8% of 417 older people comprehensively assessed in a primary care setting; however, their participants suffered from persistent dizziness [19] and accordingly reported a higher mean (SD) DHI of 36.3 (19.9). In contrast, eligibility for our study only required having suffered from at least one episode of dizziness in the past year. In addition, people receiving treatment for their dizziness at the time of baseline assessment were excluded from participating in our study; these individuals might otherwise have reported more severe dizziness and/or dizziness handicap. Low initial dizziness severity (mean [SD]: 25.4 [18.3]) leaves less room for improvement and could potentially contribute to the difficulty in establishing a diagnosis of dizziness in some of the participants with unclear causes; participants with no obvious cause of dizziness on baseline assessments (n = 71) had significantly lower mean DHI scores (mean [SD]: 15.8 [10.1]) than the rest of the sample (mean [SD]: 28.2 [19.1]; p < 0.001). Other possible reasons include but are not limited to the following: the need for the condition to be present at baseline testing (e.g., misplaced otoconia in the vestibular system producing BPPV), a previous dizziness episode triggered by severe circumstances (e.g., participant thrown to sea floor by a dumping wave) unlikely to reoccur or be reproducible in the laboratory, and a rare cause requiring specific testing (e.g., carotid massage).
Fourth, uptake of the interventions was only good (exercise, vestibular rehabilitation) to fair (CBT, Falls Clinic). Evident factors leading to non-adoption of the interventions included the receipt of healthcare through other care providers (specialist clinics and community health programmes), the requirement of a general practitioner referral letter for a Falls Clinic appointment, delays in organising the interventions, and dizziness symptoms being considered insufficiently disabling by participants to warrant action. Some participants referred to CBT declined this intervention, reporting that they expected a therapy for dizziness and not one for anxiety and depression. Comparative data regarding uptake and adherence to CBT within dizziness interventions are lacking. In fact, to our knowledge, only few randomized controlled trials to reduce dizziness have included CBT as an intervention. These trials included small sample sizes (n ≤ 41) and delivered CBT either on its own or integrated within a vestibular rehabilitation programme [26,30,32]. The face-to-face delivery method might have ensured good adherence and might explain why uptake and adherence were not reported in any of these trials. Future trials might consider the following strategies to improve CBT uptake: providing education on anxiety and depression and the good evidence base for CBT to effectively reduce depression and anxiety, additional follow-up (via telephone calls) from the research investigators and/or the general practitioner, together with exploration of potential barriers (prior beliefs and expectations) once the participant has signed up for the CBT course. However, in those who took up the interventions, adherence to the intervention was good (exercise, vestibular rehabilitation) to excellent (CBT), possibly owing to efforts to minimise potential barriers to participation in the interventions (i.e., appointments arranged with vestibular physiotherapist, provision of home-based exercise programme, and CBT offered in an internet-based or booklet form).
Finally, it is likely that participation in this study involving a comprehensive assessment may have reassured a number of participants, particularly those who were forwarded a report indicating no interventions were indicated, and subsequently reduced perceptions of dizziness and handicap. This is consistent with the finding that negative beliefs about the consequences of dizziness can be modified by vestibular rehabilitation therapy [64], for example, and the large body of research showing the powerful role of reassurance in healthcare [65].
Our randomized controlled trial provides evidence that a multifactorial, tailored pragmatic approach, involving evidence-based therapies, is effective in improving dizziness handicap in a sample of community-living people aged 50 years and over self-reporting dizziness. Our findings therefore support the implementation of a multifactorial assessment combined with tailored interventions to reduce self-perceived disability associated with dizziness in middle-aged and older community-living people.
|
10.1371/journal.pgen.1000329 | Why Is the Correlation between Gene Importance and Gene Evolutionary Rate So Weak? | One of the few commonly believed principles of molecular evolution is that functionally more important genes (or DNA sequences) evolve more slowly than less important ones. This principle is widely used by molecular biologists in daily practice. However, recent genomic analysis of a diverse array of organisms found only weak, negative correlations between the evolutionary rate of a gene and its functional importance, typically measured under a single benign lab condition. A frequently suggested cause of the above finding is that gene importance determined in the lab differs from that in an organism's natural environment. Here, we test this hypothesis in yeast using gene importance values experimentally determined in 418 lab conditions or computationally predicted for 10,000 nutritional conditions. In no single condition or combination of conditions did we find a much stronger negative correlation, which is explainable by our subsequent finding that always-essential (enzyme) genes do not evolve significantly more slowly than sometimes-essential or always-nonessential ones. Furthermore, we verified that functional density, approximated by the fraction of amino acid sites within protein domains, is uncorrelated with gene importance. Thus, neither the lab-nature mismatch nor a potentially biased among-gene distribution of functional density explains the observed weakness of the correlation between gene importance and evolutionary rate. We conclude that the weakness is factual, rather than artifactual. In addition to being weakened by population genetic reasons, the correlation is likely to have been further weakened by the presence of multiple nontrivial rate determinants that are independent from gene importance. These findings notwithstanding, we show that the principle of slower evolution of more important genes does have some predictive power when genes with vastly different evolutionary rates are compared, explaining why the principle can be practically useful despite the weakness of the correlation.
| The fact that functionally more important genes or DNA sequences evolve more slowly than less important ones is commonly believed and frequently used by molecular biologists. However, previous genome-wide studies of a diverse array of organisms found only weak, negative correlations between the importance of a gene and its evolutionary rate. We show, here, that the weakness of the correlation is not because gene importance measured in lab conditions deviates from that in an organism's natural environments. Neither is it due to a potentially biased among-gene distribution of functional density. We suggest that the weakness of the correlation is factual, rather than artifactual. These findings notwithstanding, we show that the principle of slower evolution of more important genes does have some predictive power when genes with vastly different evolutionary rates are compared, explaining why the principle can be practically useful for tasks such as identifying functional non-coding sequences despite the weakness of the correlation.
| When referring to any DNA sequence, a popular textbook of cell and molecular biology [1] states that “if it's conserved, it must be important” and calls this “one of the foremost principles of molecular evolution” (p. 416). Here, the word “conserved” means that the sequence has a low rate of evolution such that its orthologs from distantly related species are detectable and alignable. The word “important” means that the sequence has relevance to the wellbeing and fitness of the organism bearing the sequence. The above principle is often used in a comparative context, asserting that functionally more important DNA sequences evolve more slowly. Despite the fact that thousands of biologists accept this principle and use it daily in identifying functionally important DNA sequences, its validity had not been systematically examined until a few years ago when gene importance could be measured at the genomic scale [2]–[10]. Unexpectedly, however, genomic studies of bacteria, fungi, and mammals showed that although the evolutionary rate of a gene is significantly negatively correlated with its importance, the latter only explains a few percent of the total variance of the former [3],[4],[10],[11]. The striking contrast between the wide acceptance and apparent utility of the principle and the weakness of the correlation revealed from genomic analysis of a diverse array of organisms is perplexing.
The perceived theoretical basis of this simple principle is the neutral theory of molecular evolution, which asserts that most nucleotide substitutions during the evolution of a gene are due to random fixations of neutral mutations [12]–[14]. Based on this theory, Kimura and Ohta first predicted that functionally more important genes should evolve slower than less important ones because the former have a lower rate of neutral mutation than the latter [15], although their use of “functional importance” appears to mean “functional constraint on the gene” rather than “importance to the fitness of the organism”. A few years later, Wilson et al. separated the two meanings and decomposed the substitution rate of a gene (k) into two factors: the probability (P) that a random mutation will be compatible with the function of the gene and the probability (Q) that an organism can survive and reproduce normally without the gene (i.e., gene dispensability) [16]. Under the simple assumption that a mutation either completely abolishes the function of a gene (with a probability of α = 1−P) or does not affect it at all (with a probability of 1−α), we can write the substitution rate of a gene as the sum of the rate of fixation of neutral mutations and that of null mutations. Here, α can also be interpreted as functional density, the effective fraction of sites in a gene (or protein) that are required for its function. Let u be the total mutation rate, β = 1−Q be the probability that an organism cannot survive or reproduce without the gene (i.e., gene importance or the coefficient of selection against null mutations), N be the organism's population size, and Ne be the effective population size. For diploid organisms, we have(1)where is the probability of fixation of a new null mutation with fitness 0<Q<1, under genic selection (i.e, the selection against the null allele is β in homozygotes and β/2 in heterozygotes) [12]. Because f<1/(2N), k is a monotonically decreasing function of α. It is obvious that k is also a monotonically decreasing function of β, because the stronger the selection against null mutations, the lower f and k are. However, note that the above formula also indicates that in large populations, f and hence k should be relatively insensitive to β except when β is extremely small (i.e., on the order of 1/Ne). In other words, under the simplistic model assumed here, a strong negative correlation between gene importance and evolutionary rate is not expected [6] (see also Text S1 and Figure S1). However, under a more realistic model with the presence of slightly and moderately deleterious mutations, a much stronger correlation between gene importance and evolutionary rate becomes theoretically possible [7]. The strength of the correlation depends on the distribution of the deleterious functional effects of random mutations (Text S1 and Figure S1). Because the true distribution is currently unknown, theories cannot predict precisely the strength of the correlation between gene importance and evolutionary rate. These considerations notwithstanding, the apparent utility of the principle in daily practice and its lack of empirical support from genomewide studies require an explanation.
There are two simple, yet untested, hypotheses that potentially explain the weakness of the observed correlation between gene importance and evolutionary rate. First, the importance of a gene to an organism is now commonly measured by the fitness reduction caused by the deletion of the gene from the genome in a benign lab condition; deleting an important gene reduces the fitness of the organism more than deleting a less important one. But, because lab conditions differ significantly from the natural environments of organisms, gene importance determined in lab may be quite different from that in nature [6],[17]. For example, in rich media, ∼80% of yeast genes are not essential for growth [18]. However, metabolic network analysis and experimental studies showed that most of these dispensable genes are important for growth under other conditions [18],[19], some of which may resemble the natural environments of the species better than rich media. Hence, it is plausible that the weakness of the correlation between gene importance and evolutionary rate is due to inaccuracy in measuring genes' natural importance, which we refer to as the lab-nature mismatch hypothesis. But, measuring gene importance in a species' natural environment is difficult because many species such as the yeast Saccharomyces cerevisiae are found in diverse environments that are poorly characterized [20]. Moreover, even if we know the present-day natural environments of a species, they may not reflect the environments where the species lived in the past. These historical environments are crucial because the gene evolutionary rate that is being correlated to gene importance is determined by comparison between species. Nonetheless, if gene importance is measured in many different conditions, we can examine whether the correlation between gene importance and evolutionary rate is much stronger in some conditions than in the benign lab condition, which could at least demonstrate the plausibility of the lab-nature mismatch hypothesis. Here we test this hypothesis in yeast using gene importance measures from both experimental data and computational predictions. The experimental data came from a set of recently published fitness measurements of yeast single-gene-deletion strains under 418 lab stress conditions [19]. We complemented this dataset with in silico predictions of importance for metabolic enzyme genes under 104 nutritional conditions, achieved by flux balance analysis (FBA) of reconstructed metabolic networks [21],[22].
Another potential factor influencing the correlation between gene importance (β) and evolutionary rate (k) is functional density (α) in Equation 1. If α and β are negatively correlated (i.e., more important genes have lower functional density), the correlation between k and β will be weakened. Although there is no reason to believe that α and β are negatively correlated, it is worth verifying using actual data. For a given protein, α may be approximately measured by the fraction of sites in functional domains, which can be computationally predicted.
In this work, we show that neither of the above two hypotheses is correct in yeast. Rather, the weakness of the correlation between gene importance and evolutionary rate is likely to be factual rather than artifactual. We show, however, that the principle of slower evolution of more important genes does have some predictive power when genes with vastly different evolutionary rates are compared, explaining why the principle can be practically useful despite the weakness of the correlation.
The most frequently used yeast gene importance data came from the measures of relative growth rates of 5936 single-gene-deletion yeast strains in the nutritionally rich YPD medium [23]. Recently, the same type of measure was taken for all YPD-viable single-gene-deletion yeast strains under 418 diverse laboratory conditions, of which ∼75% are chemical drug treatments and the rest are environmental stress conditions such as different pHs and temperatures [19]. These two datasets of gene importance are used in our analysis.
Evolutionary rates of S. cerevisiae genes are estimated by comparing these genes to their orthologs in related species. Because the functional importance of a gene may change during evolution [10],[24], it is better to use a closely related species for rate estimation. However, when the species are too close, the number of nucleotide substitutions per gene may be insufficient for precise estimation of evolutionary rates. A previous study found the strongest correlation between gene importance and evolutionary rate when S. cerevisiae is compared with S. bayanus [10]. We thus use this species pair and obtain 3999 genes with identifiable orthologs. Our results remain qualitatively unchanged when several other yeast species were compared with S. cerevisiae (data not shown). We use the number of nonsynonymous substitutions per nonsynonymous site (dN) between orthologs to measure the rate of gene evolution (k in Equation 1). Because the mutation rate (u in Equation 1) may vary among genes, we also use the ratio between dN and the number of synonymous substitutions per synonymous site (dS) as a measure of k/u in Equation 1.
When gene importance is measured under the nutritionally rich YPD medium, the Spearman's rank correlation coefficient between gene importance (i.e., amount of fitness reduction caused by gene deletion) and dN is ρ = −0.2189 (P<10−43; Figure 1A). Our examination of 418 other lab conditions found the strongest correlation to be ρ = −0.2379 (P<10−51; Figure 1A). Thus, none of the 418 conditions provides a substantially stronger correlation than what is observed with YPD. Similar results were obtained for the correlation between gene importance and dN/dS (Figure 1B).
Krylov et al. suggested another measure of gene evolutionary rate known as the propensity for gene loss (PGL), which is the number of times that a gene is lost during the evolution of a group of species [11]. Although PGL and dN are correlated with each other [11], they measure the rate of gene evolution from different angles. The correlation between PGL and gene importance is expected to be weaker than that between dN and gene importance, because mutations that impair gene function only slightly do not matter to gene loss. We estimated PGL for each S. cerevisiae gene by counting the number of gene loss events on the known phylogeny of 12 fungal species (see Materials and Methods). Consistent with our expectation, the correlation between gene importance and PGL is weaker than that between gene importance and dN (or dN/dS) for both YPD and the other 418 lab conditions (Figure 1C). Regardless, the examination of the 418 lab conditions does not substantially improve the strength of the correlation between gene importance and PGL.
Because the 418 experimentally examined conditions contain mostly artificial chemical treatments and hence may not cover the diverse natural environments of the yeast, we decide to complement the experimental data with computationally predicted gene importance values for 546 metabolic enzyme genes under 104 conditions generated by random combinations of different nutrients following a sampling strategy that mimics the potential nutritional environments of the wild yeast (see Materials and Methods). We then used two different experimentally validated computational methods to predict the fitness reduction caused by the deletion of each enzyme gene. These methods rely on the reconstructed high-quality yeast metabolic network [25], which contains 632 biochemical reactions associated with 546 enzyme genes after the removal of dead-end reactions [26]. The first method we used is flux balance analysis (FBA). Under the assumption of steady state of every cellular metabolite, FBA maximizes the rate of biomass production under the stoichiometric constraints of all metabolic reactions [22]. Simulation of different nutritional conditions is achieved by setting the boundaries of uptake reaction fluxes and simulation of gene deletion is achieved by constraining the flux of corresponding enzymatic reaction to zero (see Materials and Methods). In our analysis, we consider the FBA-optimized rate of biomass production as the wild-type Darwinian fitness of the cell under the condition specified. The relative fitness of a cell lacking a gene is the FBA-optimized rate of biomass production of the cell, divided by that of the wild-type cell. Previous studies demonstrated that FBA makes excellent qualitative predictions of yeast gene essentiality under typical experimental conditions [18],[25]. A recent study further showed consistent performances of FBA across many different conditions [27]. Following a previous study [28], we approximated the YPD condition in the FBA model and predicted the fitness values of single-gene-deletion yeast strains. We found that the FBA-predicted fitness values correlate well with the experimentally determined fitness values under YPD (Pearson's r = 0.562, P<10−41). We were not able to verify FBA for the other 418 lab conditions because these conditions are difficult to specify in FBA.
Our extensive analysis of 104 simulated conditions identified the strongest correlation between FBA-predicted gene importance and dN to be ρ = −0.2186 (P = 10−6; Figure 1D) for 546 enzyme genes. Although this correlation is 34% stronger than that estimated using experimentally determined gene importance under YPD (ρ = −0.1636, P = 6×10−4; Figure 1D) for the same set of genes, the fraction of variance in dN that is explainable by gene importance is still as low as (−0.2186)2 = 4.8%. Similar results are obtained when either dN/dS (Figure 1E) or PGL (Figure 1F) is used as a measure of gene evolutionary rate. One interesting observation is that the standard deviation of ρ from the 104 simulated conditions (0.042, 0.037, and 0.037 in Figure 1D, E, and F, respectively) is much greater than that for the 418 experimental conditions (0.013, 0.009, and 0.008 in Figure 1A, B, and C, respectively). Part of this difference is due to the use of essentially all genes in Figure 1A–C but only enzyme genes in Figure 1D–F. However, even when only enzyme genes are considered, the standard deviation of ρ is still smaller for lab conditions (dN: 0.024; dN/dS: 0.021; PGL: 0.020) than for the 104 simulated conditions, suggesting that the simulated conditions represent a more diverse set of conditions than the experimental conditions.
FBA assumes that a cell can readjust its metabolic fluxes to achieve the highest possible biomass production immediately after the deletion of any gene, which is probably unrealistic. Segre and colleagues proposed a modified method known as the minimization of metabolic adjustment (MOMA) [29]. Instead of maximizing biomass production upon gene deletion, MOMA minimizes the changes of fluxes from those of the wild-type cell. Empirical examples suggested that MOMA outperforms FBA in predicting gene essentiality and metabolic fluxes [29]. We found that MOMA-predicted fitness values of single-gene-deletion strains are slightly better than FBA-predicted values in correlating with the experimentally determined fitness values in YPD (Pearson's r = 0.571, P<10−43). However, none of the 104 simulated conditions provide a better correlation between MOMA-predicted gene importance and evolutionary rate than the correlation found using experimentally measured gene importance in YPD (Figure 1G–H).
Although we examined 104 simulated conditions, it is possible that they still do not cover the natural conditions of yeast. We simulated 105 additional conditions and found that the distribution of the correlation coefficient ρ (Figure S2) is virtually identical with that from the initial 104 conditions. Because the distribution of ρ is approximately normal, statistically speaking, it is extremely unlikely to obtain a much stronger correlation by examining even 106 conditions. Due to the large amount of computational time required for examining large numbers of conditions and the similarity of the results from 104 and 105 conditions, we used the gene importance values predicted from the 104 conditions in subsequent analysis.
Because under no single condition, either experimentally examined or computationally simulated, did we find a strong correlation between gene importance and evolutionary rate, and because yeast may have had experienced diverse natural conditions during its evolution, we ask whether we can find combinations of single conditions for which the correlation between gene importance and evolutionary rate is much stronger than that under any single condition. We consider a simple scenario in which gene importance values under different conditions are weighted and linearly combined to form an average gene importance value across all the conditions considered. These weighting coefficients potentially represent the (unknown) relative durations of the conditions where the yeast has lived. We identify these coefficients by mathematically maximizing the correlation between the weighted average gene importance and evolutionary rate. We further constrain the weighting coefficients to be non-negative because negative coefficients are biologically meaningless. Employing the least squared method in statistics, we can transform this maximization task into a quadratic programming problem. The mathematical representation of the problem is(2)where ki is the evolutionary rate of gene i and fi is the weighted average importance of gene i in all conditions, calculated by averaging gene importance under each condition (fij) using non-negative weighting coefficients of the condition (cj). We solved the quadratic programming problem using the commercial optimization package CPLEX and then calculated the correlation between the weighted average importance of a gene and its evolutionary rate. Note, however, that the above estimation of c guarantees the identification of the strongest Pearson's linear correlation between fi and ki, but not Spearman's rank correlation. We know of no method that guarantees the identification of the strongest rank correlation between fi and ki.
Our results showed that the improvement of the correlation by combining individual conditions is trivial (Table 1). For example, for the 418 experimental conditions, the strongest Pearson's correlation between the weighted average gene importance and dN is r = −0.2187 (P<10−43), only 5% stronger than the strongest correlation found among all single conditions (r = −0.2082, P<10−39). Similar results were observed for the other measures of gene evolutionary rate and for combinations of the 104 simulated conditions (Table 1). These results indicate that even weighted average of gene importance across multiple conditions is not strongly correlated with gene evolutionary rate.
Why doesn't the consideration of so many experimental and simulated conditions and combinations of conditions improve the correlation between gene importance and evolutionary rate? Using FBA, one can classify enzyme genes into three categories according to their importance across multiple conditions: always-essential, sometimes-essential, and always-nonessential. Deleting an always-essential gene causes lethality in all conditions; deleting a sometimes-essential gene causes lethality in some but not all conditions; deleting an always-nonessential gene does not cause lethality in any condition, although it may reduce the fitness of the organism to a non-zero level. Because always-essential genes are as important as or more important than the other two classes of genes in any condition, it is clear that in order to achieve a strong correlation between gene importance and evolutionary rate in any condition or combination of conditions, the evolutionary rate of always-essential genes must be lower than those of the other two classes of genes. Here the enzyme genes are classified into the above three groups based on the essentiality predicted in the 104 simulated conditions. Although the average dN of always-essential genes is lower than that of sometimes-essential genes and that of always-nonessential genes, the differences are small and not statistically significant (Figure 2A). The same is true for dN/dS (Figure 2B) and PGL (Figure 2C). These results strongly suggest that no single condition or combination of conditions will show a strong correlation between gene importance and evolutionary rate even when more conditions are examined. Thus, if the conditions under which yeast evolved belong to the 418 experimentally examined conditions or are amenable to the current FBA, the lab-nature mismatch hypothesis must be rejected.
Equation 1 shows that if functional density (α) and gene importance (β) are independent from each other, evolutionary rate of a gene (k) should decrease with the increase of β. The observed weakness of the correlation between gene importance and evolutionary rate prompts us to examine the presumption of independence between α and β, because the correlation between gene importance and evolutionary rate could have been weakened if there is a negative correlation between α and β. By definition, α is the proportion of mutations that destroy the function of a gene, which may be experimentally determined by large-scale site-directed mutagenesis coupled with gene functional assay, a formidable task even for a few genes. In theory, one can use the average number of allowable alternative states across all amino acid sites of a protein to estimate 1−α. But such a measure is currently difficult to acquire at the genomic scale, because it requires the alignments of orthologs from many (i.e., ≫20) divergent species to assure that all potentially allowed amino acids have had chance to appear at any given site. Use of many divergent species greatly increases misidentification of paralogs as orthologs and the risk of comparing functionally-different orthologous proteins, leading to potential overestimation of 1−α. A further complication is that the evolution of a site is often dependent on other sites, meaning that an amino acid is allowed at a site only when another site has a particular amino acid [30],31. Thus, the number of allowed amino acids at a site is not a unique number, but rather depends on the genetic background of the same gene or even other genes. Given these difficulties, we decide to use the proportion of amino acid sites within computationally predicted functional domains of a protein to estimate α approximately, because α is expected to be much greater within functional domains than outside domains. This estimation of α is based on the assumption that all sites within functional domains are important to the function of the protein whereas all sites outside domains are unimportant. Although this assumption does not hold in reality, it should not affect our results as long as it does not systematically bias our estimation of α among genes of different β.
Computational algorithms for predicting protein functional domains are based on proteins of known structures and/or amino acid sequences with high evolutionary conservation [32]. There are many available algorithms for protein domain prediction and they are based on different assumptions. Here we employ two widely used methods. The first is the ProSite prediction algorithm [33], which is based on known conserved functional motif sequences. ProSite predictions are relatively conservative and should contain few false positives, as on average only 10% of amino acid sites in a protein are predicted by ProSite to be within functional domains. The second method we used is InterProScan [34], which integrates 13 well known domain prediction algorithms and databases to look for domains. Because InterProScan uses multiple algorithms, its predictions are more comprehensive. To avoid false positive predictions, we consider only those sites that are identified by at least two algorithms of InterProScan as functional domain sites. Under this criterion, on average 47% of protein sites are identified as functional domain sites.
To examine whether the proportion of sites within predicted domains indeed provide information about functional density, we conducted three tests. First, based on the domains predicted by ProSite, we found that sites within domains evolve more slowly than those outside domains in 89% of the yeast genes. The corresponding number is 77% when the domains are predicted by InterProScan. These percentages are significantly greater than the random expectation of 50 percent (P<10−100, χ2 test). Second, the mean dN within domains is 40% and 54% that outside domains in ProSite and InterProScan analysis, respectively, both being significantly different from the random expectation of 100% (P<10−50, paired t-test). Finally, we examined if there is a negative correlation between the proportion of sites within domains and the evolutionary rate of the gene, and found the correlation to be ρ = −0.24 (P<10−50) and −0.56 (P<10−50), respectively, in ProSite and InterProScan analysis. Taken together, the proportion of sites within predicted domains indeed provide information about functional density and thus may be used as a proxy for α.
Because our results do not support the lab-nature mismatch hypothesis, we here consider only experimentally measured gene importance under YPD (β). We found very weak positive correlation between α estimated by ProSite and β (ρ = 0.049, P = 0.0002) (Figure 3A). If InterProScan predictions are used, there is a stronger positive correlation between α and β (ρ = 0.15, P<10−30), suggesting that important genes tend to have a higher fraction of functional sites (Figure 3B). We also repeated the analysis under more stringent criteria of InterProScan where a site is considered as a functional domain site only when it is recognized by at least 3 to 6 algorithms. The observed correlation between α and β remains significant (ρ = 0.08–0.12, P<0.0001).
However, the above analysis has a confounding factor. Because sequence conservation information is used in predicting functional domains and because important genes tend to be more conserved in sequence (though the correlation is weak), the above observed level of positive correlation between α and β may in part or in total due to the artifact of the analysis. Indeed, we found that after the control of dN, the partial correlation between α and β becomes ρ = 0.0190 (P = 0.240) for the ProSite analysis and ρ = −0.0110 (P = 0.497) for InterProScan analysis (≥two algorithms). This result suggests no genuine correlation between α and β. Thus, the weakness of the correlation between gene importance and evolutionary rate is unlikely the result of a potential negative correlation between gene importance and functional density.
Our analysis rejected two frequently proposed explanations of the weakness of the observed correlation between gene importance and evolutionary rate, raising the question of why the correlation is so weak. As mentioned in Introduction, depending on the distribution of the fitness effect of deleterious mutations, the expected correlation may not be strong (Figure S1 and Text S1). In addition, there may be other reasons. Bivariate analysis of yeast data revealed a strong negative correlation between gene expression level and evolutionary rate [35], which led to the recent proposal of the translational robustness hypothesis, asserting that selection against toxicity of misfolded proteins generated by translational errors is the single most important factor governing the rate of protein sequence evolution [36],[37]. This hypothesis explains several factors known to correlate with the rate of protein sequence evolution (e.g., gene expression level and codon usage bias). However, many other rate determinants are known in yeast, including the number of protein interaction partners and gene length, although their impacts on the evolutionary rate are generally much smaller than that of gene expression level [38]. Principal component regression analysis and partial correlation analysis have suggested independent and significant contributions of all these factors [39],[40], although it is not always clear how these factors determine the rate of gene evolution independently from the influence of gene importance [41]. In bacteria and mammals, independent contributions from multiple factors to gene evolutionary rate are also known [4],[5]. Theoretically speaking, the single most important rate determinant is the fraction of mutations that are unacceptable to the gene (α), but this fraction is affected by many biological factors. The fact that the rate of gene evolution is jointly determined by multiple independent factors, some of which are stronger determinants than gene importance, is likely an additional reason why the rate is only weakly correlated with gene importance. To simplify the explanation, let us assume that the rate of gene evolution (k) is determined linearly by n independent factors (A1 to An) as , where ε represents the statistical error that cannot be explained by the n factors and ai's are coefficients. Pearson's correlation coefficient between k and factor Ai is(3)where Var stands for variance and Cov stands for covariance. Because one rate determinant, gene expression, already accounts for >25% of the variance of k [36],[37] and several other factors also make independent and nontrivial contributions [39],[40], the correlation between gene importance and evolutionary rate is much weakened, compared to that when gene importance is the sole contributor.
Taken together, we showed empirically that the correlation between gene importance and gene evolutionary rate is weak and showed that this weakness may not be inconsistent with theoretical predictions. In fact, if we randomly pick two yeast genes, the probability that the slower evolving of the two is the more important one is only 54% (based on 100,000 pairs of randomly sampled genes under YPD) (Figure 4A). That is, the prediction based on one of the foremost principles of molecular evolution has a success rate of only 54%, not much greater than that of a pure guess (50%). When the two genes being compared have a larger difference in evolutionary rate, the prediction about their relative importance becomes more accurate, as expected (Figure 4A). For example, we ranked all yeast genes by their evolutionary rates and found that when two genes are separated in rank by over 95% of all genes, the probability that the slower evolving one is more important than the other is 81% (Figure 4A). Essential genes are functionally most important. When the gene importance data from YPD is considered, we found that 55% of the top 5% most conserved genes are essential, whereas only 20% of the remaining 95% of yeast genes are essential (Figure 4B). Similar results are found using the gene importance data from the other 418 lab conditions (Figure 4B). Note that the above demonstrated predictability may not be entirely due to the causal relationship between gene importance and evolutionary rate, because other confounding factors such as gene expression level have not been controlled for. Regardless, our results show that although the correlation between gene importance and evolutionary rate is weak, the principle does have some predictive power when genes of extreme sequence conservation are considered.
There are several caveats in our analysis that warrant discussion. First, experimental measures of gene importance are not without errors. Repeated measures of gene importance under the same conditions showed a correlation as high as 0.92 for the YPD data [23] but a reduced mean correlation of 0.72 for the other 418 lab conditions [19], possibly due to less well controlled experimental procedures in the latter. Thus, the gene importance data we used could potentially explain a maximum of 0.722 = 52% of the variance of the evolutionary rate. But the strongest correlation actually observed was only r2 = 4.3% among the 418 individual conditions and 4.8% among combinations of the 418 conditions, both being substantially lower than the theoretical maximum. Similar arguments can be made for the analysis based on computationally predicted gene importance values.
Second, a limitation in using dN and dN/dS to measure the rate of gene evolution is that they can be used only for those S. cerevisiae genes that have orthologs in the species being compared with (i.e., S. bayanus). Our results would not represent a full picture if genes with and without orthologs have drastically different levels of gene importance. To examine this possibility, we compared their importance levels. Because we used reciprocal best hits in BLAST searches to define orthologs, a S. cerevisiae gene would not have its operational S. bayanus ortholog, if (i) the gene evolved extremely fast, (ii) the gene has been lost in S. bayanus, or (iii) the gene has been duplicated in S. cerevisiae such that its S. bayanus best hit happens to find its paralog to be the best hit. Thus, we separated S. cerevisiae genes into singletons and duplicates. We found no significant difference in gene importance between S. cerevisiae genes with and without S. bayanus orthologs, for either singletons (P = 0.11, Mann-Whitney U test; Table S1) or duplicates (P = 0.63, Table S1). Hence, the potential bias of studying only S. cerevisiae genes that have S. bayanus orthologs is negligible.
Third, we used three different measures of gene evolutionary rate: dN, dN/dS, and PGL. They all have pros and cons, aside from the above consideration. In principle, dN/dS would be the best measure, because it best measures k/u, which is determined by α and β only, according to Equation 1. Estimates of dN/dS, however, suffer from two problems. First, dS values may have been saturated because the average dS between S. cerevisiae and S. bayanus is as high as 1.24. Although using more closely related species could improve the estimation of dS, it would increase the estimation error of dN and that of dN/dS, due to a reduced number of nonsynonymous substitutions per gene. Second, codon usage bias, prevalent in highly expressed genes of yeast, could lead to underestimation of neutral substitution rates and thus overestimation of k/u. Because of the positive correlation between the importance of a gene and its expression level [10], codon usage bias causes greater overestimation of k/u for more important genes, weakening the negative correlation between k/u and gene importance. If there is little variation in mutation rate among genes, dN would be a better index of evolutionary rate for our purpose than dN/dS, because estimates of dN have smaller sampling errors than those of dN/dS. Our results show stronger correlations between gene importance and dN, compared to that between gene importance and dN/dS, suggesting that the disadvantages of using dN/dS outweigh its advantages. Propensity for gene loss (PGL) treats each gene as a unit and does not consider the number of substitutions per nucleotide or amino acid site. It is thus conceptually different from the evolutionary rate that Kimura and Ohta [15] and Wilson et al. [16] referred to. There are three reasons underlying our observation that gene importance correlates more poorly with PGL than with dN and dN/dS. First, because PGL is determined by the fixation of null mutations but not slightly deleterious mutations, it should be less influenced by gene importance, as explained in Introduction and Figure S1. Second, estimation of PGL requires genome sequences from a number of species related to the focal species of interest (S. cerevisiae). In the present case, PGL is estimated from 12 diverse fungi and thus may not accurately reflect the propensity of gene loss in S. cerevisiae, because the importance of a gene can change in evolution [10],[24]. Third, estimates of PGL potentially have large sampling errors, because the estimated number of losses per gene is quite small.
Fourth, to understand why no single condition or combination of single conditions provides gene importance values that correlate strongly with evolutionary rates, we classified enzyme genes into three groups (always-essential, sometimes-essential, and always-nonessential) and compared their respective evolutionary rates. Due to computational intensity, our classification was based on the FBA analysis of 104 simulated conditions, while in theory it should have been based on all possible conditions. This limitation potentially caused misclassification of some truly sometimes-essential genes as always-essential genes or always-nonessential genes and hence blurred the differences among the three groups. To rectify this problem, we used a strategy that guarantees the identification of all always-essential genes. The metabolic model of yeast allows us to know all nutrients that can be used by this metabolic model. If a gene is essential when all these nutrients are present, it must be essential when one or more of these nutrients are absent. We find that in fact the always-essential genes thus identified are identical to those identified from the 104 simulated conditions. There is, however, no systematic way to guarantee the exact separation of sometimes-essential and always-nonessential genes. We thus merged them and compared this combined group with always-essential genes. Again, we do not find the combined group to have significantly greater dN, dN/dS, or PGL than always-essential genes (Figure 2). Thus, our result is true not only for the 104 simulated conditions, but also for all possible combinations of nutrients usable by the yeast metabolic model. Our result differs from that of Papp et al. [18] where they found that enzyme genes active in more conditions have lower probabilities of presence in the genomes of 133 diverse species. At least five reasons may account for this difference. First, we counted PGL on a known phylogeny of related species using the parsimony method whereas these authors simply calculated the percentage of species that do not have the gene without considering the species phylogeny [18]. Second, most of the species they used are distantly related to yeast and their result is expected to be highly dependent on the choice of species. Third, we considered gene essentiality, a more relevant measure of gene importance than gene activity, because deleting an active gene may or may not have any fitness consequence, depending on alternative pathways in the metabolic network. Fourth, we used a more recent reconstruction of the yeast metabolic network, which is more complete and accurate than the one they used. Fifth and most importantly, because only nine conditions were examined, their result could simply be due to small sample size.
Fifth, Hirsh and Fraser suggested that the correlation between gene importance and evolutionary rate should exist only among genes with relatively low importance [7]. This is because, in Equation 1, f quickly declines to virtually 0 when β increases from 0 to 0.1 and any further increase in β has negligible effects on f and k, although Hirsh and Fraser came to this conclusion using a more complex model [7]. However, we found that the correlation for genes with β<0.1 is extremely weak (ρ = −0.05 for YPD and the strongest ρ = −0.04 among the 418 experimental conditions). We cannot test genes with even smaller β because the accuracy of the estimated β decreases and the number of useable genes decreases. The contradiction between Hirsh and Fraser's prediction and our empirical observation can be understood using Figure S1. Apparently, when there are many slightly and moderately deleterious mutations, use of all genes provides a stronger correlation than using only unimportant genes, because the expected evolutionary rates can still be different between a gene with β = 0.2 and a gene with β = 0.3 (Figure S1K). For example, in Figure S1L, using only genes with β<0.1 gives ρ = −0.36, whereas using all genes gives ρ = −0.83.
Sixth, the correlation between gene importance and evolutionary rate reported here may be in part caused by other co-varying factors. For three reasons, we did not control for confounding factors in our analysis. First, previous authors already determined that the correlation is statistically significant even after the control of confounding factors [3],[10]. Second, our goal here is to discern why the correlation is so weak even when part of it may come from confounding factors. Third, we study the difference in the magnitude of the correlation when various gene importance measures are used; confounding factors such as gene expression level would not affect this difference.
Despite the general belief and wide application of the principle that important genes evolve more slowly than less important ones, genomic analysis showed that the correlation between gene importance and evolutionary rate is quite weak. Our analysis does not support the hypothesis that the weakness of the observed correlation is due to the difference between gene importance in the lab and in nature. Furthermore, we found no evidence for the possibility that the correlation is weakened by the potential presence of a smaller fraction of functional sites in more important genes. We conclude that the weakness of the correlation is factual, rather than artifactual. This conclusion is not inconsistent with population genetic predictions, because the predictions vary depending on the prevalence and distribution of the fitness effect of deleterious mutations.
Our result cautions molecular biologists from predicting relative functional importance of genes directly from their relative levels of evolutionary conservation. Nevertheless, our finding that extremely conserved genes are highly likely to be functionally very important may explain the universal perception that the principle of slower evolution of more important genes (or DNA sequences) works well. For example, substantial amount of comparative genomic work aims at using the principle to identify functional non-coding sequences based on their extremely low rates of nucleotide substitution [42]–[45]. An ultra-conserved non-coding sequence is a segment of DNA of over 200 nucleotides with no variation among human, mouse, and rat. Pennacchio et al. found that such ultra-conserved sequences, when they are also conserved between mouse and fish, have a probability of 62% to be actual enhancers during mouse embryonic development [42]. Compared to the virtually zero probability with which a random segment of DNA in the mouse genome is an enhancer, the principle appears to work well. This success is not surprising, because only extremely conserved non-coding sequences are considered. Nevertheless, it should be noted that although a large fraction of extremely conserved non-coding sequences are functional, many functional sequences are not extremely conserved. In other words, the current application of the principle in detecting functional non-coding sequences has a high false-negative rate. Thus far, there has been no evidence that the correlation between sequence importance and evolutionary rate is stronger for non-coding regions than for coding regions. One reason for a potentially stronger correlation for non-coding regions is that several rate determinants in coding sequence evolution simply do not exist in non-coding sequence evolution (e.g., codon usage bias, amount of translation, gene length, and number of protein-interacting partners). In addition, the fraction of mutations that are slightly deleterious may be greater for non-coding regions than for coding regions, given the high modularity of regulatory sequences. In the future when relative importance of many functional non-coding sequences is measured, it will be interesting to examine whether non-coding sequences exhibit a greater correlation between importance and evolutionary rate.
The fitness values of homozygous-single-gene-deletion yeast strains in the YPD medium [23] were downloaded from http://www-deletion.stanford.edu/YDPM/YDPM_index.html. The corresponding data from the other 418 lab conditions [19] were obtained from http://chemogenomics.stanford.edu:16080/supplements/global/download.html. The microarray raw data were processed by the author-provided Perl scripts and were then normalized to the central mean to yield the relative fitness values of the deletion strains under each condition.
The metabolic network model of S. cerevisiae (iND 750) [25] used in this study was downloaded from the BiGG database (http://bigg.ucsd.edu) and parsed by the COBRA toolbox [46]. The network is composed of 1149 reactions, associated with 750 known genes. Some reactions do not have associated genes because the genes whose protein products catalyze these reactions have yet to be identified. The network model also provides information about stoichiometry, direction of reaction, and gene-reaction association. We followed an established protocol [26] to identify dead-end reactions, which are reactions that must have zero flux under a steady state. These reactions are involved in the generation of metabolites that are neither included in biomass nor transported outside the cell, and may reflect the incompleteness of the metabolic network model. After the removal of dead-end reactions, the yeast metabolic network used in our analysis contains 632 biochemical reactions with 546 associated enzyme genes.
Details of FBA have been described in the literature [21],[22]. Briefly, the flux of each reaction is determined by maximizing the rate of biomass production under the assumption of steady state and the constraints of stoichiometry. We used the optimization package CPLEX (www.ilog.com) to solve the linear programming problem. Gene deletion is modeled by constraining the flux of the corresponding reaction to zero.
MOMA has been previously described in detail [29]. Briefly, MOMA predicts the maximal biomass production rate upon deletion of a reaction by minimizing the differences in all metabolic fluxes between the deletion strain and the wild-type strain. All the constraints used in FBA are still enforced in MOMA. The quadratic programming problem is also solved by CPLEX. As in FBA, deletion of a gene is realized by constraining the flux of the corresponding reaction to zero.
The natural environments of yeast may change frequently. It is also likely that yeast usually faces nutritionally poor conditions but occasionally encounter rich conditions. To mimic their natural environments, we simulate random nutritional conditions in the following manner. For each condition, we generate a random number g from an exponential distribution with a mean of m = 0.1 for each of the 103 usable carbon-source nutrients. Here, g is the probability that the carbon-source nutrient is available. The actual presence or absence of each nutrient is then determined stochastically using g. We then add all required inorganic metabolites. Use of other m values (0.05 or 0.5) does not change our results. For each available nutrient, we fix the uptake rate at a random value between 0 and D = 20. The actual D value used is unimportant and does not alter our result. Only conditions that support the growth of the wild-type cell, as shown by FBA, are considered.
Singleton and duplicate genes of yeast S. cerevisiae are identified by BlastP searches of each gene against all other genes in the genome. A gene is considered as a duplicate if it hits at least one other gene in the genome with the criteria of an E-value = 10−10 and an alignable region >50% of the longer sequence. Otherwise, it is treated as a singleton.
Following [10], we used the maximum likelihood method to estimate synonymous (dS) and nonsynonymous (dN) substitution rates of yeast genes by comparing the orthologous genes of S. cerevisiae and S. bayanus, which were identified by reciprocal best BLAST hits. The PGL information was obtained from a previous study [47], which used the parsimony principle to estimate the number of gene losses on the phylogeny of 12 fungi (S. cerevisiae, S. bayanus, S. paradoxus, S. mikatae, Candida glabrata, Kluyveromyces lactis, Eremothecium gossypii, Debaryomyces hansenii, Yarrowia lipolytica, Neurospora crassa, Kluyveromyces waltii, and Schizosaccharomyces pombe).
We downloaded the latest release (Release 20.27) of protein domain scan algorithm ProSite [33] from ftp://ca.expasy.org/databases/prosite/, where an executable program and a compiled domain motif database were available. InterProScan [34] was downloaded from http://www.ebi.ac.uk/Tools/InterProScan/ with the current-release database, and was set up to run locally to identify protein domains.
|
10.1371/journal.ppat.1005024 | Vpu Exploits the Cross-Talk between BST2 and the ILT7 Receptor to Suppress Anti-HIV-1 Responses by Plasmacytoid Dendritic Cells | Plasmacytoid dendritic cells (pDCs) constitute a major source of type-I interferon (IFN-I) production during acute HIV infection. Their activation results primarily from TLR7-mediated sensing of HIV-infected cells. However, the interactions between HIV-infected T cells and pDCs that modulate this sensing process remain poorly understood. BST2/Tetherin is a restriction factor that inhibits HIV release by cross-linking virions onto infected cell surface. BST2 was also shown to engage the ILT7 pDC-specific inhibitory receptor and repress TLR7/9-mediated IFN-I production by activated pDCs. Here, we show that Vpu, the HIV-1 antagonist of BST2, suppresses TLR7-mediated IFN-I production by pDC through a mechanism that relies on the interaction of BST2 on HIV-producing cells with ILT7. Even though Vpu downregulates surface BST2 as a mean to counteract the restriction on HIV-1 release, we also find that the viral protein re-locates remaining BST2 molecules outside viral assembly sites where they are free to bind and activate ILT7 upon cell-to-cell contact. This study shows that through a targeted regulation of surface BST2, Vpu promotes HIV-1 release and limits pDC antiviral responses upon sensing of infected cells. This mechanism of innate immune evasion is likely to be important for an efficient early viral dissemination during acute infection.
| Plasmacytoid dendritic cells (pDCs) produce large quantities of type I interferon (IFN-I) upon stimulation by many viruses, including HIV. Their activation is very effective following cell contacts with HIV-1-infected CD4+ T cells. We investigated whether HIV-1 could regulate the antiviral responses of pDCs triggered upon sensing of infected cells. We show that HIV-1 suppresses the levels of IFN-I produced by pDCs through a process that requires expression of the Vpu accessory protein in virus-producing cells. A well-described role of Vpu is to promote efficient HIV-1 production by counteracting BST2, a host factor that entraps nascent viral particle at the cell surface. Apart from its antiviral activity, BST2 was reported to inhibit IFN-I production by pDCs through binding and activation of the ILT7 pDC-specific inhibitory receptor. Our results reveal that through a highly sophisticated targeted regulation of BST2 levels at the surface of infected cells, Vpu promotes HIV-1 release and limits IFN-I production by pDCs via the negative signaling exerted by the BST2-ILT7 pair. Overall, this study sheds light on a novel Vpu-BST2 interaction that allows HIV-1 to escape pDC antiviral responses. This modulation of pDC antiviral response by HIV Vpu may facilitate the initial viral expansion during acute infection.
| Plasmacytoid dendritic cells (pDCs) are a distinct subset of DCs that exhibit a unique ability to secrete high amounts of interferons and other cytokines in response to viruses. Even though they constitute less than 1% of the total cell content of peripheral blood in humans, they are considered a primary source of type-I IFN (IFN-I) for antiviral responses. Hence, pDCs represent the first line of defense against viral infections, and as such, serve as a vital link between innate and adaptive immunity. Detection of virus infection by pDCs is mediated through recognition of viral nucleic acids, including single-stranded RNA (ssRNA) and double-stranded DNA containing unmethylated CpG motifs by the Toll-like receptor 7 (TLR7) and 9 (TLR9) endosomal sensors. Activation of TLR7/9 induces signaling events that ultimately lead to the stimulation of IFN genes through IRF7 and pro-inflammatory cytokines genes via NF-κB [1].
The role of pDCs during HIV infection appears to be complex [2]. pDCs are activated in HIV and SIV infection and are rapidly depleted from blood, coinciding with their redistribution to lymph nodes and mucosal tissues [3] where they are largely responsible for the IFN-I being produced during acute infection [4]. In addition, pDCs may be chronically stimulated during HIV infection and a continuing source of IFN-I, a feature that seems central to the immune activation and the CD4+ T cell loss during pathogenic infection [2,5]. pDCs express the primary HIV receptor, CD4, as well as the main co-receptors, CXCR4 and CCR5, and as such support entry of X4 and R5 strains of HIV [6]. Upon sensing HIV-1, pDCs produce IFN-I and other cytokines, and undergo phenotypic activation [7–9]. Although high concentrations of purified HIV virions are capable of inducing IFN-I from pDCs, HIV-infected CD4+ T cells are much more effective at stimulating these IFN-producing cells (10-100-fold relative to cell-free virus) given their ability to establish cell contacts with them [6,8,10]. Thus, potent recognition of cell-associated HIV by pDCs may represent an important host strategy to overcome the poor detection of cell-free virions.
HIV-infected cells are sensed by pDCs through a process that involves endocytosis and fusion of virions that are transferred across cell contacts. Once fusion is completed, the ssRNA genome is believed to gain access to endosomes where recognition occurs in a large part through TLR7. This pathway of recognition of cell-associated HIV appears to differ from detection of cell-free HIV, in which there is a requirement for attachment and endocytosis but not for viral fusion [6,8,11]. Whether cell-free HIV-1 and cell-associated HIV-1 are recognized by separate pathways that converge on TLR7 remains unclear. Since efficient sensing involves cell contacts between pDCs and infected cells, it is conceivable that viral proteins and/or host factors expressed at the surface of HIV-producing CD4+ T cells might modulate pDC sensing or/and antiviral responses. Indeed, the HIV envelope glycoprotein gp120 was found to inhibit TLR9-mediated antiviral responses most likely by interacting with the pDC specific inhibitory receptor, BDCA-2 [12,13]. However, the interactions between HIV-infected T cells and pDCs that modulate TLR7-mediated pDC sensing and the resulting antiviral responses remain poorly understood.
BST2/Tetherin (also known as CD317) is a glycosylated type II integral membrane protein that is induced by IFN-I (for review, see [14]). The dimeric protein has a distinctive topology as it contains two membrane anchors. It consists of a short N-terminal cytosolic tail, a transmembrane (TM) domain, an elongated extracellular coiled-coil [15] and a glycophosphatidyl-inositol (GPI)-linked membrane anchor present at the C-terminus [16]. BST2 was found to strongly inhibit the release of HIV and other enveloped viral particles [17,18] and to impede cell-to-cell HIV transmission [19,20], although this notion has been challenged [21]. BST2 restricts HIV release through a direct “tethering” mechanism whereby the protein cross-links progeny virions to cell membranes by a process that requires both membrane anchors [22]. Interestingly, apart from its direct antiviral activity, BST2 has also been reported to act as the ligand of immunoglobulin-like transcript 7 (ILT7, also known as LILRA4 and CD85g), a receptor preferentially expressed at the surface of human pDCs [23]. ILT7 is an inhibitory receptor that controls inflammation by modulating the production of IFN-I and other pro-inflammatory cytokines by pDCs [24]. Indeed, binding of BST2 to ILT7 was found to negatively regulate activation of the two key pDC sensors, TLR7 and TLR9, hence controlling IFN-I secretion [25].
The prototypical HIV-1 molecular clone pNL4.3 and pandemic HIV-1 group M strains counteract the antiviral activity of BST2 via the viral protein U (Vpu) accessory protein. Vpu was found to enhance the release of progeny virions by downregulating BST2 from the cell surface of infected cells [17,18], although enhanced viral release in the absence of BST2 downregulation has been reported [26]. In the absence of Vpu, a large number of fully formed and mature HIV-1 particles are entrapped by BST2, thus forming large clusters of virions at the cell surface. These viral clusters could possibly occlude BST2 regulatory interactions as well as become potent targets for immune sensing by pDCs. We therefore hypothesized that BST2 antagonism by Vpu might modulate the ILT7 inhibitory signal that regulates TLR7/9-mediated production of IFN-I in activated pDC. In this study, we tested this hypothesis.
To examine the role of Vpu during innate sensing of infected cells by PBMCs, MT4 T cells were infected with GFP-marked wild type (WT) or Vpu-deficient (dU) HIV-1 viruses. As previously reported [17,18], Vpu triggered a significant but incomplete downregulation of surface BST2, resulting in an efficient release of HIV-1 particles (S1A–S1D Fig). In parallel, infected MT4 cell populations with a similar percentage of infected cells (GFP-positive cells) were co-cultured for 18–24 h with freshly isolated human PBMCs (Fig 1A). Consistent with previous results [6], IFN-I was very efficiently detected only after co-culture of infected T cells with PBMCs (Fig 1B). Interestingly, levels of IFN-I released in co-cultures were decreased in a Vpu-dependent manner (Fig 1B). Analysis of innate sensing of HIV-1-infected cells by PBMCs from different donors revealed that the presence of Vpu in infected donor cells led to an ~50% reduction in IFN-I production (Fig 1C). We confirmed our results using different Vpu variants as well as primary CD4+ T cells as virus-producing cells. Since pDCs express both HIV co-receptors, CXCR4 and CCR5, primary CD4+ T cells were infected with GFP-marked WT or dU NL4.3 encoding CXCR4 or CCR5-tropic envelopes (NL4.3 and NL4.3-Ada). It is important to note that NL4.3-Ada encodes a Vpu protein from the primary Ada isolate. As observed with MT4 infections, production of IFN-I triggered upon sensing of HIV-infected primary T cells by PBMCs was decreased by ~50% in the presence of Vpu (NL4.3) or Vpu (Ada) (Fig 1D and 1E). Furthermore, similar results were also obtained when NL4.3 virus encoding Vpu variants from transmitter/founder (T/F) strains Suma and CH077 were used. These Vpu variants decreased innate sensing as efficiently as NL4.3 Vpu (Fig 1F and 1G), a finding that was consistent with their comparable ability to downregulate and antagonize BST2 (S1E, S1F and S1G Fig). Collectively, these results suggest that Vpu-mediated BST2 antagonism suppresses the production of IFN-I triggered upon sensing of HIV-1 infected cells by PBMCs independently of co-receptor usage. Additionally, they further indicate that this property is a conserved feature of Vpu variants from primary and T/F virus strains.
Since pDCs are responsible for the vast majority of the IFN-I produced during HIV innate sensing, we then examined whether the differential regulation of IFN-I production mediated by Vpu could be observed in co-cultures with enriched pDCs. To this end, HIV-infected MT4 cells were co-cultured with PBMCs, pDC-depleted PBMCs or enriched populations of pDCs (Fig 2A). In the absence of pDCs, PBMCs lost the ability to sense infected MT4 cells (Fig 2B). In contrast, virus recognition was re-established in the presence of enriched pDC fractions (Fig 2B). Importantly, the differential regulation of IFN-I production mediated by Vpu was observed when enriched pDCs were co-cultured with infected MT4 cells (Fig 2C and 2D). These results indicate that Vpu modulates the antiviral response of pDCs upon sensing of HIV-1 infected cells.
Previous studies reported that pDCs recognize WT HIV-infected cells in a large part through TLR7 and by a process that involves Envelope (Env)-dependent transfer of virus across cell contacts [6,8]. Since detection of infected cells by pDCs was enhanced in the absence of Vpu, we examined whether sensing in that context still involved recognition of viral genomic RNA by TLR7. To this end, we first analyzed innate sensing of WT and dU HIV-infected cells in the presence or the absence of inhibitors of either viral fusion or reverse transcription (T-20 and 3TC, respectively). While treatment of HIV-producing cells with T-20 is expected to affect the Env-dependent fusion of transferred ssRNA-containing virions to pDCs, treatment of PBMCs with 3TC will specifically inhibit the conversion of viral RNA into proviral DNA in pDCs. Our results reveal that innate sensing of both WT and dU HIV-infected cells was dependent on viral fusion but not on reverse transcription (Fig 3A and 3B). These findings indicate that Env-dependent fusion of viral particles is required in both contexts and further suggest that viral genomic RNA is likely the primary contributor to WT and dU HIV sensing by pDCs, at least during the first 24 h of the co-culture. Similarly, we tested the effect of TLR9 (ODNTTAGGG, Invivogen) and TLR7/8/9 [27] antagonists (Fig 3C and 3D). These antagonists displayed minimal cytotoxic effects at the concentration that were used. Consistent with the fact that the TLR9 pathway is largely inactivated in pDCs presumably through an interaction of gp120 with BDCA-2 [12,13], we observed only a modest effect of the TLR9 antagonist on pDC response to WT and dU HIV-infected cells (Fig 3E and 3F). In contrast, inhibition of both endosomal TLRs present in pDCs, namely TLR7 and 9 [1] with the TLR7/8/9 antagonist resulted in an almost complete abrogation of IFN-I production in response to both WT and dU HIV-infected cells. These findings support the notion that innate sensing of both WT and dU HIV-1 infected cells relies in a large part on the activation of the TLR7 pathway.
Having shown that Vpu could dampen IFN-I production during innate sensing of infected cells, we next evaluated the contribution of BST2 towards this process. For this purpose, we generated BST2-depleted (MT4-shBST2) as well as control (MT4-shNT) MT4 cell populations. BST2 downregulation (Fig 4A) and enhanced virus particle release (S2A and S2B Fig) were observed in MT4-shNT cells infected with WT virus when compared to dU virus. In contrast, depletion of BST2 abolished the requirement of Vpu for efficient viral particle release (S2A and S2B Fig). Indeed, the absolute amounts of virus released after WT or dU infections were comparable to those released from WT-infected MT4-shNT cells (S2A Fig). These infected cells were then co-cultured with PBMCs. While the absence of BST2 did not affect the level of IFN-I produced upon sensing of dU HIV-infected cells, it completely abolished the reduction of IFN-I production mediated by Vpu. Indeed, upon depletion of BST2, WT and dU HIV-infected cells were sensed to a degree comparable to that observed with dU HIV-infected MT4-shNT control cells (Fig 4B and 4C). Similar results were observed with several independently depleted cell populations, thus excluding confounding effects resulting from cellular selection. These findings demonstrate that the control of IFN-I production mediated by Vpu is dependent on the presence of BST2 on infected donor cells. It is unlikely that the more efficient IFN-I production observed in co-cultures with Vpu-defective HIV-infected T cells is due to a more effective presentation of dU virions at the cell surface for transmission and/or sensing by pDCs. Indeed, in condition where BST2 was depleted, dU HIV-infected MT4 cells were sensed as efficiently as dU HIV-infected control cells (Fig 4B and 4C), despite the fact that virus release efficiency was clearly different in these two conditions (S2A and S2B Fig). Using these cell lines, we further showed that the X4/R5-tropic CH077 T/F virus was as efficient as the NL4.3 virus at suppressing innate sensing in the presence of BST2 (Fig 4D and 4E), thus providing additional evidence that T/F virus have the capacity to control IFN-I production by pDCs. To further confirm the role of BST2 in the Vpu-mediated control of innate sensing, we tested in our co-culture system a virus encoding a TM domain mutant of Vpu (Vpu A10-14-18L) that is drastically attenuated in its ability to bind and antagonize BST2 [28] (S2C and S2D Fig). This mutant virus was found to be phenotypically identical to dU virus during innate sensing (S2E and S2F Fig), further supporting the crucial role that Vpu-mediated BST2 antagonism plays in regulating pDC antiviral responses.
We next assessed whether Vpu-mediated reduction of IFN-I production was dependent on the presence of accessible BST2 molecules at the surface of infected donor cells. To this end, infected MT4 cells were pre-incubated with BST2-specific polyclonal rabbit antibodies (Rb BST2 Ab) directed against the extracellular domain of the molecule or rabbit pre-immune serum (Rb PI). Miyagi and colleagues previously reported that binding of similar anti-BST2 polyclonal Abs to BST2 did not trigger internalization of the protein but rather stabilized the restriction factor on the cell surface [29]. A saturation of potentially available cell-surface BST2 with the BST2-specific polyclonal Rb Ab was confirmed by the near absence of BST2 staining detected with an anti-BST2 mouse monoclonal antibody (mAbs 26F8) 18 h after addition of Rb BST2 Abs (Fig 5A). Hence, mock or infected MT4 cells were pre-incubated with BST2-specific polyclonal Rb Abs or the Rb PI control prior to the 18 h co-culture with PBMCs. Fig 5B and 5C reveals that treatment with Rb BST2 Abs abolished the differential innate sensing observed between WT and dU HIV infected cells. Noticeably, in both cases the levels of IFN-I were lower than those detected in dU HIV infected cells treated with control Abs, suggesting a potential slight interfering effect of the rabbit polyclonal anti-BST2 serum on pDCs. These results combined with those obtained upon depletion of BST2 suggest that control of IFN-I production by Vpu requires that BST2 be expressed at the surface of infected donor cells and available for potential interactions.
One possible model to explain the efficient IFN-I production by pDCs during sensing of dU HIV-infected cells could be that tethered viral clusters at the surface of infected cells would occlude BST2 from potential interaction with the ILT7 inhibitory receptor on pDCs. In contrast, in cells infected with WT HIV, Vpu would downregulate BST2 sufficiently to prevent restriction of virion release but maintain enough BST2 to engage and activate ILT7. Indeed, it was recently reported that besides inducing a downregulation of BST2, Vpu has the ability to displace the restriction factor from sites of viral assembly as a means to counteract BST2 restriction [30,31]. To assess the frequency of surface BST2 molecules that are not potentially engaged in restriction of progeny virions, we performed co-localization studies of BST2 and HIV Gag-p17, a marker of assembling HIV, on GFP-positive infected MT4 and primary CD4+ T cells. As previously reported [30], surface BST2 accumulated in patches that perfectly co-localized with p17 in the absence of Vpu (Figs 6A, top panels and S3E). BST2 molecules localizing outside of viral assembling sites (we refer to this pool as ˝free BST2˝) were rarely found (Fig 6B), suggesting that in dU HIV-infected cells the majority of surface BST2 appears engaged in restricting assembling viruses. Consistent with the flow cytometry data of S1C and S2D Fig and our previous observations [32], surface BST2 levels were significantly decreased in the presence of Vpu and this was linked to a reduction in p17 accumulation at the cell periphery (Figs 6A, 6C and S3E) in agreement with efficient virus particle release (S1A and S1B Fig). Noticeably, in this context, we could detect clusters of BST2 that were not co-localizing with p17 (marked with white open arrows) (Fig 6A and 6B), suggesting that Vpu displaces a residual pool of surface BST2 from virus assembly sites.
Two isoforms of human BST2 derived from alternative translation initiation from highly conserved methionine residues in the cytosolic domain were identified in both immortalized cell lines and primary cells [33]. Both isoforms are able to restrict virion release. The short isoform, which lacks 12 N-terminal residues, including conserved tyrosine and serine-threonine motifs present in the long isoform, was reported to be significantly more resistant to Vpu-mediated degradation, surface downregulation and antagonism as compared to the long isoform [33,34]. To assess the distribution of these BST2 isoforms relative to assembling virions in the presence or absence of Vpu, SupT1 T cells, which naturally do not express BST2 [17], were supplemented with either the long Vpu-sensitive BST2 isoform or the short BST2 isoform (S3A and S3B Fig). In the absence of Vpu, both isoforms were found to similarly co-localize with p17 (Fig 6A bottom panels), consistent with their comparable ability to restrict virus particle release (S3C and S3D Fig). Vpu induced a significant depletion of the long BST2 isoform at the cell surface, whereas it had only minor effects on short BST2 isoform surface levels (Figs 6A, 6C and S3B). Despite the lack of short isoform downregulation by Vpu, we did not observe a marked accumulation of p17 staining at the periphery of T cells (Fig 6A), further confirming that Vpu-mediated BST2 antagonism can occur in the absence of BST2 downregulation (S3B, S3C and S3D Fig). This was in sharp contrast to dU HIV-infected SupT1-shortBST2 cells which expressed similar levels of surface BST2 but displayed a noticeable accumulation of cell-associated p17 (Figs 6A, 6C, S3B, S3C and S3D). Interestingly, in the context of Vpu-expressing infected cells, significant amounts of short BST2 molecules (both in terms of frequency and intensity) could be detected outside viral assembly sites (Fig 6), indicating that a pool of short BST2 isoform molecules is potentially accessible for interactions in the presence of Vpu. Nevertheless, it is important to point out that a significant pool of free BST2 could also be detected in SupT1 cells expressing the long BST2 isoform in the presence of Vpu, albeit their levels (as indicated by integrated pixel density) and frequency outside viral assembly sites were lower as compared to the short isoform (Fig 6). Importantly, in none of the tested cell lines nor in primary CD4+ T cells, were we able to detect significant amounts of free BST2 in the absence of Vpu (Fig 6B), even though surface BST2 levels remained comparable to those of mock-infected cells (Fig 6A and 6C). These observations were confirmed using the gp120 Env protein as an additional marker of viral assembly sites, as previously described [35]. Co-localization studies of BST2 and Env in the presence or absence of Vpu further documented that Vpu induced a redistribution of BST2 outside assembly sites in primary CD4+ T cells and in SupT1 cells expressing the short BST2 isoform (S3F Fig). Overall, our data indicate that in the absence of Vpu, the pool of free BST2 at the surface of infected cells is limited. In contrast, in the presence of Vpu, there is a residual pool of surface BST2 that is excluded from viral budding sites and thus potentially accessible for interaction with ILT7 on pDCs. We expect this pool of BST2 to be predominantly composed of the short isoform.
The BST2 GPI anchor is required for BST2-mediated entrapment of virions and in its absence HIV particles are no longer retained at the surface of infected cells, as assessed by their efficient release [17]. The ability of the BST2-dGPI mutant to block the TLR7 pathway was tested. HEK293T were transfected with plasmids expressing either WT BST2 or a BST2-dGPI mutant and the levels of BST2 was evaluated by flow cytometry (Fig 7A). When PBMCs were co-cultured with BST2-expressing HEK293T cells in the presence of a TLR7 agonist, IFN-I production was reduced by ~66% as compared to control (Fig 7B). Importantly, HEK293T cells expressing the BST2-dGPI mutant efficiently repressed IFN-I production as well, highlighting the fact that the BST2-dGPI anchor is not required for inhibition of TLR7-mediated IFN-I production by BST2 (Fig 7B). Of note, no IFN-I was detected after co-culture of PBMCs with HEK293T cells in the absence of TLR7 agonist (Fig 7B). Sup T1 control cells (Empty) as well as SupT1 cells expressing BST2-dGPI at levels comparable to those detected in SupT1-WT BST2 cells were infected with WT or dU viruses prior to co-culture with PBMCs (S4A Fig). These cells exhibited the expected phenotype in terms of virus release (S4B and S4C Fig) and surface BST2 modulation by Vpu (S4A Fig). Consistent with the results obtained with BST2-depleted MT4 cells (Fig 4), BST2-deficient SupT1 cells infected with WT or dU virus triggered similar levels of IFN-I upon their co-culture with PBMCs (Fig 7C). In contrast, in co-cultures with BST2-expressing SupT1 cells, HIV-1 WT triggered significantly less IFN-I than its dU virus counterpart (Fig 7D and 7E). Interestingly, expression of BST2-dGPI at the surface of HIV-producing cells very efficiently repressed IFN-I production by PBMCs independently of the presence of Vpu. Indeed, in this context, the extent of repression of IFN-I production was comparable to that observed when PBMCs were co-cultured with WT HIV-infected SupT1-BST2 cells (Fig 7D and 7E). Taking together, these results suggest that trapping of progeny virions by BST2 prevents the restriction factor from eliciting an inhibition of IFN-I production by pDCs.
Because surface BST2 was a critical mediator of Vpu-mediated innate immune modulation, we next examined whether the BST2 binding partner, ILT7, expressed on pDCs could be involved in this process as well. The reported interaction between BST2 and ILT7 [25] was confirmed in vitro by surface plasmon resonance (KD = 2.33 μM) using recombinant GST-tagged ectodomain of BST2 (GST-BST2, a soluble form containing a region common to the short and the long BST2 isoforms) and baculovirus-expressed soluble ILT7 (Fig 8A and 8B). This interaction was further demonstrated in cellulo by flow cytometry and by proximity ligation assay using Fc-tagged ectodomain of BST2 (BST2-Fc) and ILT7-expressing HEK293T cells or a ILT7+ NFAT-GFP reporter cell line [24], respectively (Fig 8C, 8D and 8E). Interestingly, ILT7 activation could be triggered in the reporter cell line using plate-bound BST2-Fc or plate-bound anti-ILT7 Abs but not with soluble anti-ILT7 Abs (Fig 8F and 8G), most likely because of rapid internalization of the receptor-Ab complexes (Fig 8H). Activation of ILT7 could also be detected by co-culture with BST2-expressing cells (Fig 8I and 8J). Importantly, ILT7 activation in the presence of BST2-expressing HEK 293T cells could be blocked when either anti-ILT7 or anti-BST2 Abs were added to the co-culture (Fig 8K). These results indeed confirm that the interaction between these two proteins is capable of triggering ILT7 activation, a signaling event that leads to repression of IFN-I production in pDCs [24,25].
To examine whether the presence of Vpu in HIV-producing cells could affect ILT7 activation through BST2, we co-cultured BST2-expressing HEK 293T cells producing either WT or dU HIV with ILT7+ NFAT-GFP reporter cells. While ILT7 activation was very effective in co-culture with BST2-expressing cells, it was significantly reduced when these cells were co-expressing HIV (99% vs 74%) (Fig 9A). Noticeably, WT HIV-producing cells were found to significantly enhance ILT7 activation compared to HEK293T cells expressing dU HIV (74% vs 48%) (Fig 9A and 9B). While these latter cells express higher levels of surface BST2, it is expected that a large proportion of these molecules would be restricting progeny viruses (Fig 9C). These results suggest that Vpu-mediated BST2 antagonism allows the presence of a pool of surface BST2 molecules that are capable of engaging and activating ILT7 upon cell-to-cell contact.
To further examine the role of the ILT7-BST2 regulatory axis, we used a specific type of siRNA (see Materials and Methods) to deplete the endogenous ILT7 from enriched pDC cultures without inducing spontaneous IFN-I production nor cell cytoxicity (Fig 9D and 9E). siRNA-treated pDCs were then co-cultured with infected MT4 cells. While control pDCs were found to secrete reduced amounts of IFN-I upon co-culture with WT HIV-infected MT4 cells relative to co-cultures containing dU HIV-infected MT4 cells, depletion of ILT7 in pDCs abrogated such differences (Fig 9E and 9F). Moreover, addition of recombinant soluble ILT7 (soILT7-HA, Fig 9G) to co-cultures of PBMCs and infected MT4 cells similarly abolished the differential IFN-I production detected with WT and dU HIV-infected cells (Fig 9H and 9I). These results demonstrate that Vpu-mediated suppression of IFN-I production is dependent on the expression of ILT7 on pDCs.
The Vpu accessory protein has been shown to downregulate many host proteins involved directly or indirectly with anti-HIV immune responses, including among others CD4, NTB-A, CD1d and BST2 [36]. With respect to BST2, it was recently reported that this Vpu-regulated host factor could act not only as a potent inhibitor of HIV release but also as an innate sensor capable of inducing NFκB-dependent proinflammatory responses upon retroviral retention and activation of a Syk-dependent HemITAM in BST2 [37,38]. In the present study, we shed light on a novel Vpu-BST2 interaction, which allows HIV to suppress IFN-I production by pDCs via the negative signaling exerted by the ILT7-BST2 pair (Fig 10).
Using a system of co-culture between HIV-infected T cells and PBMCs or enriched populations of pDCs, we provide evidence that Vpu-mediated BST2 antagonism suppresses the production of IFN-I during innate sensing of HIV-infected T cells. This result initially obtained with a laboratory-adapted Vpu (NL4.3), was further confirmed using Vpu variants from a primary isolate (Ada) and from T/F strains (Fig 1). While this finding could suggest that the increased IFN-I response observed in the absence of Vpu may result from a more effective presentation of BST2-restricted dU HIV-1 virions at the cell surface for transmission or/and sensing by pDCs, a condition counteracted by Vpu, the data herein is not consistent with such a possibility. First, depletion of BST2 in MT4 T cells (Fig 4), a process that abolishes the restriction on HIV-1 release and its dependence on Vpu, did not reduce IFN-I production during innate sensing of dU HIV infected cells but rather enhanced pDC antiviral responses triggered upon sensing of WT HIV-infected cells. Noticeably, BST2 blocking experiments using anti-BST2 Abs (Fig 5) or soluble ILT7 (Fig 9) phenocopied this result. Second, a potential difference in the efficiency of virus transmission from infected T cell to pDCs cannot explain why depletion of ILT7 in pDCs eliminates the modulatory effect of Vpu on IFN-I production nor why a differential activation of ILT7 by WT or dU HIV-expressing cells can be observed with a reporter cell line that is not susceptible to HIV infection (Fig 9).
Our results are more consistent with a mechanism of innate immune evasion whereby Vpu regulates the levels of BST2 molecules on HIV-producing cells that are capable of engaging the ILT7 inhibitory receptor on pDCs (Fig 10). Our immuno-localization studies in HIV-infected T cells indeed reveal the presence of a residual pool of surface BST2 that is found outside viral assembly sites. In the absence of Vpu, such a pool could not be detected since a very large proportion of surface BST2 molecules were found to co-localize with assembling viruses (Fig 6). While the exact nature of this residual surface BST2 pool remains to be precisely defined, one possibility is that it could be generated by a displacement of BST2 molecules from sites of viral assembly by Vpu. Indeed, it was recently shown that in addition to causing BST2 sequestration to internal compartments and downregulation from the cell surface, Vpu had the ability to directly bind and displace BST2 from nascent virions at the plasma membrane [30,31]. This property would enable Vpu to exhibit residual BST2 antagonist activity in the absence of BST2 downregulation.
Two isoforms of human BST2 can be generated by alternative translation initiation [33]. These isoforms display distinct biological activities and have the ability to form homodimers and heterodimers. The short isoform, which lacks a dual-tyrosine-based sorting motif, phospho-tyrosine sites and potential ubiquitination acceptor residues present in the long isoform, is unable to induce NFκB activation and is significantly more resistant to Vpu-mediated downregulation and degradation [33,34]. Since both of these isoforms retain a functional ectodomain capable of interacting with ILT7, we examined whether they could have evolved to ensure both a potent restriction of invading viruses as well as an effective regulation of viral innate immune sensing and its resulting antiviral and proinflammatory responses. Thus, by preferentially targeting the long BST2 isoform for downregulation and degradation, HIV would not only overcome BST2 antiviral and signaling activities, but also maintain the negative signaling on IFN-I production exerted by the short BST2 isoform via ILT7. Our immuno-localization analysis reveals that both isoforms strongly co-localize with assembling virions in the absence of Vpu, and as such would be occluded from potential interactions with ILT7 on pDCs. A surprising finding from these studies was the effective capacity of Vpu to displace the short BST2 isoform from viral assembly sites and to overcome its tethering activity despite its resistance to Vpu-mediated downregulation (S3B, S3C and S3D Fig). This finding suggests that the sensitivity of short BST2 to Vpu antagonistic activity in infected T cells may be more significant than anticipated from previous data generated in transiently transfected HEK 293T cells [33]. In fact, both isoforms could be detected outside viral assembly sites in the presence of Vpu, suggesting that in both contexts a pool of surface BST2 molecules would potentially be accessible for interaction with ILT7. However, given their different surface levels and frequency of localization outside viral assembly sites, it is likely that the short BST2 isoform will be over-represented at the surface of HIV-1 WT-infected cells. Unexpectedly, despite these differences, we could not detect any differential repression of IFN-I production when pDCs were co-cultured with infected SupT1-shortBST2 or SupT1-longBST2 cells. In both cases, the presence of Vpu in virus-producing cells attenuated production of IFN-I by pDCs to a similar extent (S5 Fig). While this apparent discrepancy may result from the already potent repression observed with BST2-expressing SupT1 cells (Fig 7D and 7E), it may also well be possible that our co-culture assay is not sensitive enough to detect such a difference in repression or that expression levels of single isoforms in the cell lines may be a contributing factor. Nevertheless, our data indicate that in WT HIV-infected T cells, significant amounts of short BST2 homodimers as well as a smaller fraction of long BST2 homodimers (and possibly heterodimers) are re-distributed outside virus assembly sites in the presence of Vpu where they would be potentially accessible to engage the ILT7 pDC receptor. Whether BST2 molecules remaining outside viral assembly sites interact with ILT7 at contact zones of virological synapses between infected donor T cells and target pDCs is an interesting question that requires further study.
Our results suggest that through a sophisticated targeted regulation of specific BST2 isoforms involving surface downregulation and/or exclusion from viral assembly sites, Vpu could promote HIV-1 release while at the same time interfering with pDC antiviral responses through ILT7 activation (Fig 10). Interestingly, a recent study analyzed the sensitivity of the long and short BST2 isoforms to counteraction by primate lentiviruses and found that the differential targeting of BST2 isoforms by Vpu appears a specific property of proteins encoded by pandemic group M HIV-1 [34]. In contrast, both isoforms of the human and rhesus macaque BST2 demonstrated similar sensitivities to viral countermeasures by HIV-2 and SIVmac, respectively. It is interesting to note that the viral antagonists used by HIV-2 (Env) and SIVmac (Nef) counteract BST2 by a non degradative process, which involves removal of the protein from the cell surface through enhanced internalization and intracellular sequestration [39–42]. Vpu proteins encoded by SIV strains from African guenons as well as a Vpu protein from a recently isolated highly pathogenic HIV-1 group N from Togo (N1.FR.2001) [43] antagonized both isoforms, although in the latter case, N1.FR.2001 Vpu was unable to induce the degradation nor the downregulation of either isoforms [34]. Moreover, it was recently shown that HIV-1 group O strains have evolved to use Nef and not Vpu to counteract human BST2 [44]. However, consistent with their ability to target a domain in the N-terminal cytosolic region of human BST2 that is missing in short BST2, group O Nef did not antagonize the short isoform. Thus, primate lentiviruses diverge in their ability to differentially target BST2 isoforms with Vpu proteins from the pandemic group M showing a unique and conserved ability to counteract differently the short and long BST2 isoforms. Whether this targeted regulation of BST2 isoforms by Vpu may have contributed to the successful spread of HIV-1 group M at least in part by allowing a suppression of innate sensing by pDCs, remains an interesting possibility that will require further studies.
Activation of the ILT7–FcεRIγ complex by BST2 was found to initiate the ITAM-mediated activation of Src and Syk kinases in pDCs, a condition that ultimately inhibits production of IFN-I and other proinflammatory cytokines [25]. Our binding and functional studies support previous results from Cao and colleagues showing that BST2 binds and potently activates ILT7. Recently, the role of the ILT7-BST2 pair in the modulation of IFN-I production by pDCs has been challenged [45]. This study showed that treatment of PBMCs with the BST2 mAb 26F8 failed to enhance IFN-I release following TLR9 agonist treatment, despite the ability of 26F8 to block ILT7/BST2 binding. While these findings appear to challenge the physiological relevance of the BST2/ILT7 interaction, there are several possible alternative explanations for these results. The most important among these relates to the experimental system used in that study. In contrast to our co-culture system where an initial contact between infected T cells and pDCs is required for sensing to occur, treatment of PBMCs with TLR9 agonists activates pDCs directly without the need for cell-to-cell contact. Interestingly, all our attempts to differentially impact pDC ability to produce IFN-I through co-coculture with CD4+ T cells expressing or not BST2 in presence of TLR agonists were unsuccessful. In contrast, a significant reduction of IFN-I production by pDCs treated with TLR7 agonist could be observed upon co-cultures with HEK293T cells overexpressing BST2 (Fig 7B), suggesting that the concentration of BST2 at the cell-to-cell contact might also be an important factor driving the activation of ILT7. Furthermore, since the interaction between ILT7 and BST2 needed to be stabilized with cross-linking agents in cellulo (Fig 8C, 8D and 8E) and that close cell-to-cell contact was required to achieve activation of ILT7 by BST2-expressing transformed or cancer cells [25], it remains to be demonstrated whether contacts between activated pDCs and bystander normal T cells (as opposed to transformed or infected cells) would be frequent and sustained enough to engage the BST2/ILT7 negative feedback system. Indeed, it was reported that pDCs preferentially form conjugates with herpes simplex virus (HSV)-infected cells compared to uninfected cells [46]. Clearly, more studies will be required to define the determinants that drive BST2/ILT7 interactions at infected T cell/pDC contacts.
Studies in humanized (hu) mice indicated that Vpu-mediated BST2 antagonism promoted HIV-1 replication and propagation in vivo, especially at early times post-infection when the predominant mode of viral transmission is likely to be by cell-free viruses [47,48]. Under these conditions, Vpu appeared to be important in ensuring the efficient initial viral expansion that is most likely necessary to enable dissemination to local lymphoid tissues and establishment of infection [49]. Given the requirement of Vpu-mediated BST2 antagonism for efficient innate immune evasion, it is tempting to speculate that control of pDC antiviral responses during early infection could also be important to ensure sufficient local viral expansion so that widespread dissemination could occur. Indeed, consistent with their role as a major source of IFN-I, pDCs were found to effectively suppress HIV replication in hu-mice. Depletion of pDCs prior to HIV infection in that model prevented induction of IFN-I and IFN-stimulated genes (ISGs) and increased viral replication and dissemination [5]. Therefore, by limiting IFN-I and proinflammatory cytokines production by pDCs in the initial phase of infection, Vpu may contribute to the increased transmission fitness of T/F viruses by enabling efficient viral spread in an environment where expression of BST2 and other ISGs remain low [49].
Clearly, a better understanding of the mechanisms underlying Vpu-mediated inhibition of IFN-I production by pDCs during sensing of HIV infected cells may provide important insights into viral transmission and pathogenesis during acute infection
Rabbit polyclonal anti-BST2, anti-Vpu, and anti-p17 antibodies (Abs) were previously described [32,50,51]. Mouse anti-BST2 (mAb 26F8), anti-ILT7 (17G10.2) and their respective isotype controls were purchased from eBiosciences. The following mouse mAb: anti-HA (HA.11 Clone 16B12, formerly Covance), anti-CD3_Pacific Blue, anti-ILT7_PE, anti-ILT7_alexa647 and anti-CD4_PerCP/Cy5.5 were purchased from Biolegend, while anti-CD14_PE/Texas Red and anti-CD303 (BDCA-2)_APC were purchased from Caltag and Miltenyi, respectively. Anti-HIV-1 gp120 Monoclonal 2G12 (anti-Env) was obtained through the NIH AIDS Reagent Program [52–56]. All secondary Abs used for flow cytometry and western blot were purchased from Life Technologies and BioRad, respectively. Goat anti-human IgG was obtained from Abcam. Human rIFN-α2a was purchased from PBL. TLR7 and 9 agonists, TLR9 antagonist and their respective controls were obtained from InvivoGen: TLR7 agonists Imiquimod (final concentration: 2.5 μg/ml) and R848 (10 μg/ml); TLR9 agonist ODN 2216 CpG-A (5 μM); TLR9 antagonist or its control (ODNTTAGGG and ODNCtrl, 100nM). TLR7/8/9 antagonist or its control, were kindly provided by Idera Pharmaceuticals (final concentration: 500 nM) [27]. Fusion (T-20, 0.5 μM) and reverse transcription (3TC, 300 μM) inhibitors were obtained from the NIH AIDS Reagent Program, Division of AIDS, NIAID. Various types of siRNAs and transfection conditions were tested in order to select a combination that did not trigger IFN-I production by pDCs. Among the tested conditions, ON-TARGETplus siRNAs transfections using Oligofectamine (Invitrogen) was selected as it effectively depleted the target gene, did not trigger spontaneous IFN-I production after transfection, and displayed low toxicity in treated pDCs. ON-TARGETplus SMARTpool siRNAs targeting ILT7 and negative control C8b were obtained from Thermo Scientific and were transfected with Oligofectamine reagent according to the manufacturer’s recommendations. Purity and viability of transfected pDCs after co-culture was examined by flow cytometry (Forward Scatter / Side Scatter and BDCA-2+) and assessment of IFN-I production capacity.
SupT1 and MT4 T cells were obtained from the NIH AIDS reagents program [57] while HEK293T and HEK-blue human IFN reporter cell lines were obtained from ATCC and InvivoGen, respectively. HEK293T cells were transiently transfected using lipofectamine 2000 (Invitrogen Inc). The ILT7 NFAT-GFP reporter cell lines were a generous gift from Dr. Yong-Jun Liu [24]. In the ILT7+ NFAT-GFP reporter mouse cell line (CT550), which expresses ILT7 and FcϵRIγ, GFP is driven by an NF-AT promoter (NFAT-GFP) and results in GFP expression in response to ILT7 surface ligation. The negative control ILT7- NFAT-GFP reporter mouse cell line (CT59Fc) expresses FcϵRIγ and NFAT-GFP, but not ILT7.
A panel of full-length transmitted/founder (T/F) HIV-1 infectious molecular clones (Cat #11919) was obtained through the NIH AIDS Reagent Program [58–61]. Except for T/F CH077, all viruses used were derived from a pNL4.3 construct (X4 WT), in which the nef gene is followed by an internal ribosome entry site that allows expression of GFP (pNL4.3-GFP) [62]. The Vpu-deficient mutant (dU) was generated by subcloning a SalI-KpnI fragment from the Vpu-defective pNLVpuDEL1 [63] into pNL4.3-GFP. The virus encoding a TM domain mutant of Vpu (Vpu A10-14-18L) attenuated in its ability to bind and antagonize BST2 [28] was generated by site-directed mutagenesis from pNL4.3-GFP. The infectious CCR5-tropic GFP-marked proviral DNA (pNL4.3-Ada-GFP; R5 WT) and Vpu-deficient mutant (R5 dU) were previously described [48]. Briefly, the pNL4.3 backbone was rendered CCR5-tropic by transferring the SalI-BamHI fragment from a well-characterized CCR5-tropic construct Ada (NLHXADA), which encodes Vpu and Env proteins from ADA [64]. GFP-marked NL4.3 viruses expressing T/F Suma Vpu (pNL-Suma) or T/F CH077 Vpu (pNL-77) were generated using the overlapping PCR method. PCR fragments amplified from pNL4.3 provirus, from the Sal I site to the beginning of Vpu (ATG) and from the end of Vpu (stop codon) to the BamH1 site in env, were fused to PCR-generated DNA fragments encompassing the Vpu open reading frame from either T/F Suma or CH077 proviruses. The resulting chimeric fragments were digested with Sal I and BamH1 and cloned back into pNL4.3 backbone. The sequence of the Vpu-Env region of all provirus constructs used was validated by automated sequencing.
ILT7 expressing plasmid, pMX-Puro-HA-ILT7, and BST2-Fc expressing plasmid were a generous gift from Dr. Wei Cao (Department of Immunology, University of Texas, MD Anderson Cancer Center, Houston) [25]. The BST2-Fc construct expresses a soluble form of BST2 lacking the cytoplasmic tail, the transmembrane region and the GPI anchor. The HA-tagged ILT7 open reading frame was sub-cloned from its original backbone to pcDNA3.1. soILT7-HA was amplified by PCR from the pcDNA3.1-HA-ILT7 plasmid. The encoded N-terminal signal peptide and HA-tag were maintained but the amino acid (aa) sequence was truncated after residue 435 by adding a premature stop codon right before the TM domain (HA-ILT7). The soILT7-HA DNA fragment was inserted back into the pcDNA3.1 backbone. For Surface Plasmon Resonance (SPR) analysis, the synthetic codon-optimized ILT7 gene (Eurofins Genomics) encoding residues 24 to 435 (bacILT7) was cloned into the transfer vector pFL, followed by Tn7-based transposition into the EMBacY bacmid to generate a recombinant baculovirus [65]. The extracellular domain of BST2 (residues 47 to 159) was cloned into the pBADM30 expression vector in order to construct a His-tagged GST-N-terminus fusion protein. The original plasmid for subcloning of BST2-dGPI, Tetherin delGPI, was a generous gift from Dr. Paul Bieniasz [17]. Short (delta Met1) or long (delta Met13) BST2 were generated by PCR and fused to a weak kozak sequence. WT BST2, BST2-dGPI, short BST2 and long BST2 open reading frames were cloned into the pcDNA3.1 backbone for transient transfections and in pLenti-CMV/TO_Puro_DEST (Addgene) for the generation of stable cell lines.
MT4 cells were transduced using lentiviral vector particles containing shRNA targeting BST2 (Clone ID: TRCN0000107018, from OpenBiosystem) or control shRNA (target sequence: 5’CAACAAGATGAAGAGCACCAA3’). SupT1 cells stably expressing the different BST2 isoforms (WT BST2, dGPI, short and long BST2) as well as control SupT1 cells (empty) were established by lentiviral vector transduction, based on pLenti-CMV/TO_Puro_DEST plasmids. Briefly, recombinant lentiviral particles were generated by transfecting HEK293T cells with the lentiviral vector together with psPAX2, a plasmid encoding HIV-1 Gag/Pol, Tat and Rev, as well as with pVSVg, a vector expressing the G glycoprotein of vesicular stomatitis virus (VSV), as previously described [66]. Two days post-transfection, the culture media containing the lentiviral particles was used to transduce SupT1 or MT4 cells. Cell lines expressing short or long BST2 isoforms were evaluated by Western Blot following immunoprecipitation and PNGase F glycosidase treatment to remove carbohydrate modifications (New England Biolabs, Inc.). Briefly, cell lysates were immunoprecipitated with rabbit anti-BST2 Abs, digested with PNGase F overnight at 37°C and analyzed by 16% Tris-Tricine SDS-PAGE.
Infectious HIV-1 viruses T/F CH077, GFP-marked NL4.3 or VSV-G-pseudotyped NL4.3-Ada virus derivatives were generated by lipofectamine transfection of HEK293T cells. Supernatants containing virus were harvested 2 days post transfection, clarified, pelleted by ultracentrifugation and titrated using the TZM-bl indicator cells as described previously [51].
Release of virus particle was assessed by Western blot as described previously [67]. Viral particle release efficiency was evaluated by determining the ratio of virion-associated Gag (p24) signal over the total intracellular Gag (p24 + p55) signal measured by scanning densitometry analysis of Western blots. Viral release efficiency was normalized to the value obtained in cells infected with WT virus, which was set at 100%.
Peripheral blood samples were obtained from healthy adult donors who gave written informed consent in accordance with the Declaration of Helsinki under research protocols approved by the research ethics review board of the IRCM. PBMCs were isolated by Ficoll-Paque centrifugation (GE Healthcare) and cultured in RPMI-1640 media supplemented with 10% FBS. CD4+ T lymphocytes were isolated by negative selection using a CD4+ T Cell Isolation Kit (Miltenyi Biotec). Enriched CD4+ T cells were then activated using PHA-L (5 μg/mL) for 48 h and maintained in RPMI-1640 complete medium supplemented with IL-2 (100 U/mL). Activated primary T cells were infected 5 days post-isolation. Human pDCs were enriched by negative selection using the Diamond Plasmacytoid Dendritic Cell Isolation Kit II (Miltenyi Biotec). Surface phenotyping was carried-out by flow cytometry as described.
Two days prior to co-culture, T cells were infected with T/F CH077 or different GFP-marked pNL4.3 viruses, at different MOIs. Infection rates were calculated by measurement of GFP+ or intracellular p24+ cells by flow cytometry, as previously described [48]. Cultures with a range of 20–50% infected cells were subsequently used for co-cultures. Target and donor cell were mixed at a ratio of 3:1 (PBMC:T cell) or 1:5 (pDC:T cell) in a final volume of 250 μl and cultured in U-bottom 96-well plates for 18–22 h. Fusion (T-20, 0.5μM) or reverse transcription (3TC, 300μM) inhibitors were added to infected MT4 or to freshly isolated PBMCs, respectively, for 1 h prior to co-culture. For TLR antagonist experiments, freshly isolated PBMCs were pre-treated with either TLR9 antagonist or its control (ODNTTAGGG and ODNCtrl, 100nM) or TLR7/8/9 antagonist or its control (500nM) for 1 h prior to co-culture with infected cells. Agonist treatments (ODN2216, 50fM for TLR 9 and R848 10ug/ml for TLR7) were used as positive controls. For BST2 blocking experiments, mock or infected MT4 cells were pre-treated for 30 min with anti-BST2 rabbit polyclonal or pre-immune Abs at 37°C, prior to the 18h co-culture with PBMCs. Alternatively, HEK293T cells were transfected with plasmids (pLenti-CMV/TO_Puro_DEST) encoding for WT BST2 or BST2-dGPI 2 days prior to co-culture with PBMCs. After 6 h of co-culture, TLR7 agonist Imiquimod was added to a final concentration of 2.5 μg/ml and cells were kept in co-culture for an additional 18 h. In all conditions, co-cultures were then transferred to a V-bottom 96-well plate, and centrifuged for 5 min at 400g. Supernatants were then used to quantify the amounts of IFN-I produced as described in the supplemental information. Each experimental replicate (n) was performed using cells from a different donor.
BST2 cell-surface staining and flow cytometry analysis of live cells was performed as previously described [51]. In all histograms shown, mean fluorescence intensity (MFI) values are shown for each sample. Surface phenotyping was carried-out using multi-parametric surface flow cytometry staining. Briefly, freshly isolated PBMCs or pDCs (2 x 106 cells/ml and 5 x 104 cells/ml, respectively) were washed with cold PBS/EDTA/FBS, blocked with human IgG, and stained with the appropriate fluorochrome-conjugated surface cellular marker Abs for 60 min at 4°C. The CD4+ subpopulations of T cells were defined as CD3+/CD4+/CD14-. The monocytes were defined as CD3-/CD14+, while the pDC population was defined as CD3-/CD14-/BDCA-2+/ILT7+. Cells were washed, re-suspended in PBS and analyzed using a Cyan flow cytometer with FlowJo software (Treestar).
Detection of bioactive human IFN-I was performed using reporter cell line HEK-Blue IFN-α/β (InvivoGen) as previously described [66]. IFN-I concentration (U/ml) was extrapolated from the linear range of a standard curve generated using known amounts of IFN-I.
Primary CD4+ T cells, MT4 and SupT1 cell lines were infected with VSV-G-pseudotyped NL4.3-Ada (WT or dU) viruses. Forty-eight hours post-infection, cells were immunostained with anti-BST2 and anti-Env Abs for 45 min at 4°C prior to extensive washes. Cells were then plated on polyD-lysine-treated coverslips and fixed for 30 min in 4% PFA. Viral matrix p17, a product of Gag polyprotein cleavage by viral protease during viral budding at the surface of the infected cells, was used as marker of assembling HIV-1 particles and was detected with a specific antibody that does not recognized immature Gag products [32]. To detect p17 fixed cells were permeabilized in Triton 0.2% for 5 min, incubated for 2 h at 37°C in 5% milk-PBS containing anti-p17 Abs, washed and incubated with the appropriate secondary Ab for 30 min at room temperature. All analyses were acquired using a 63× Plan Apochromat oil immersion objective with an aperture of 1.4 on an LSM710 Observer Z1 laser scanning confocal microscope coupled with a Kr/Ar laser (Zeiss). Surface BST2 was quantified by measuring raw integrated signal density using ImageJ software on manually selected cells.
The supernatant of SF21 insect cells secreting bacILT7 was collected 4 days post infection, dialyzed against buffer A (20 mM Tris, 150 mM NaCl, pH 7.2) and concentrated 2-fold. The supernatant was then applied to a nickel-affinity chromatography (Qiagen) column. The column was washed sequentially with buffer A containing 10, 50 and 70 mM imidazole, followed by elution of bacILT7 with buffer A containing 300 mM imidazole. The eluted fractions were pooled and dialyzed extensively against buffer B (20 mM Tris, 150 mM NaCl, 10% glycerol). Analytical size exclusion chromatography showed that the majority of the protein eluted in a peak at 13 mL from a Superdex 200 column. BacILT7 was dialyzed against HBS-PE and cleared by centrifugation at 100,000 g for 20 min. GST-BST2 was expressed in E. coli Rosetta (DE3) cells (Novagen) and purified in buffer C (20 mM Tris, 100 mM NaCl, pH 7.5) by nickel-affinity chromatography (Qiagen) followed by size-exclusion chromatography on a Superdex 200 column (GE Healthcare) in buffer D (20 mM HEPES, 100 mM NaCl, 10 mM EDTA, pH 7.5). Surface plasma resonance was performed on a Biacore 3000 (GE Healthcare) system using HBS-PE as a running buffer (10 mM HEPES pH 7.5, 150 mM NaCl, 3 mM EDTA and 0.005% Tween-20). GST-BST2 was diluted to 5 μg/ml in 10 mM sodium acetate (pH 4) buffer and covalently immobilized to the surface of a CM5 sensor chip by amine coupling according to the manufacturer’s instructions, yielding an Rligand of 6550 RU. A reference flow cell was generated by amine coupling of GST alone (Rligand = 1028). BacILT7 was serially diluted into running buffer and passed over the chip at a flow rate of 10 μl/min. The response from the GST-coated reference cell was subtracted from the response resulting from specific binding to the target protein. Regeneration of the sensor chip was achieved with 10 mM HCl for 60 seconds. The spikes present in the sensorgrams are due to a delay in the bulk refractive index change between the flow cells, which is exacerbated by the 10 μl/min flow rate. Data were analyzed with the BIAevalution software version 4.1.
HEK293T cells were transiently transfected with a control empty plasmid, with BST2-Fc or with soILT7-HA expressing-plasmid using lipofectamine 2000 (Invitrogen Inc). Cell culture medium was replaced 6 h post transfection with serum-free medium (DMEM supplemented with Nutridoma-SP (Roche, Life Science)). Supernatant containing the soluble protein as well as control were collected 48 h post transfection, clarified by centrifugation (400g, 5 min) and filtered through a 0.2 μm pore size membrane.
HEK293T cells were transfected with an empty plasmid or with a construct encoding the full length ILT7. Cells were detached and incubated with control supernatant or with BST2-Fc-containing supernatant (approximately 300 μg/ml) for 30 min at 4°C, prior to crosslinking with DTSSP (3,3´-dithiobis [sulfosuccinimidylpropionate], Thermo Scientific) for 30 min at room temperature. The crosslinking reaction was stopped by addition of 1 M Tris, pH 7.5. Cells were washed and then stained for surface BST2-Fc using on one hand rabbit polyclonal anti-BST2 serum and goat anti-rabbit IgG coupled with Alexa 633 to label the BST2 portion and on the other hand anti-human IgG coupled with Alexa 633 to label the Fc fragment of the recombinant protein. Cells were analyzed using a Cyan ADP flow cytometer and FlowJo software (Treestar).
In situ proximity ligation assay (PLA) was performed using the Duolink kit 613 (Sigma Aldrich). Briefly, ILT7+ NFAT-GFP or ILT7- NFAT-GFP reporter cells were incubated with control supernatant (CTRL sup) or with BST2-Fc-containing supernatant (BST2-Fc sup, approximately 300 μg/ml) prior to crosslinking with DTSSP for 30 min at room temperature. The crosslinking reaction was stopped by addition of 1M Tris, pH 7.5. Cells were washed, adhered on poly-L Lysine slides, fixed using a 4% PFA solution and blocked using the blocking solution provided in the PLA kit. Fixed cells were then incubated with the following Abs: mouse mAb antibody against ILT7 or rabbit polyclonal serum against BST2. The Duolink system provides oligonucleotide-labeled secondary Abs (PLA probes) to each of the primary Abs that, in combination with a DNA amplification-based reporter system, generate a signal only when the two primary Abs are in close proximity (less than 40 nm). Following the manufacturer’s recommendation, after addition of the PLA probes, the oligonucleotides were ligated and amplified using the ligase and polymerase provided. Finally, nuclei were counter-stained using DAPI. The signal from each detected pair of primary Abs was visualized as a red spot using fluorescence confocal microscopy.
Standard ELISA plates were sterilized by UV treatment. Plates were treated with either anti-ILT7 Abs or BST2-Fc. For treatment with ILT7 Abs, Alexa647-conjugated Anti-ILT7 Abs (diluted in PBS to 1 μg/ml) were allowed to adhere to plates for 18 h at 4°C. For BST2-Fc treatment, plates were first coated with goat anti-human IgG (diluted in PBS to 10 μg/ml) for 2 h at 4°C, washed with PBS and incubated with BST2-Fc-containing supernatant (approximately 300 μg/ml) for an additional 16 h at 4°C. The following day, ILT7+ NFAT-GFP reporter cells were added and plates were incubated for 18 h at 37°C. To test the effect of unbound ILT7 Abs (soluble ILT7 Abs), Alexa647-conjugated Anti-ILT7 Abs (diluted in media to 1 μg/ml) were added directly to ILT7+ NFAT-GFP reporter cells prior to an overnight incubation. Cells were collected and analyzed by flow cytometry for GFP expression. For ILT7 activation experiments involving BST2-expressing cells, BST2-expressing or control cell lines (50,000 cells/well) were plated in 24 well plates 18–24 h prior to co-culture with either ILT7+ NFAT-GFP or control ILT7-NFAT-GFP reporter cells (100,000 cells/well). The co-cultures were maintained for an additional 18–24 h, at which time samples were analyzed by flow cytometry for GFP expression. For blocking experiments, anti-ILT7 Abs (10 μg/ml) were added to reporter cells or anti-BST2 (30 μl of polyclonal rabbit serum) to BST2-expressing or control cells for 1 h prior to co-culture.
Statistical analysis was performed using repeated measures ANOVA, with Bonferroni’s multiple comparison test or two-tailed paired Student’s t-tests. A p value of <0.05 was considered significant. The following symbols were used throughout the manuscript: *** p<0.001, ** p<0.01, * p<0.05, ns not significant (p>0.05).
|
10.1371/journal.ppat.1004932 | Vaccine-Elicited Tier 2 HIV-1 Neutralizing Antibodies Bind to Quaternary Epitopes Involving Glycan-Deficient Patches Proximal to the CD4 Binding Site | Eliciting broad tier 2 neutralizing antibodies (nAbs) is a major goal of HIV-1 vaccine research. Here we investigated the ability of native, membrane-expressed JR-FL Env trimers to elicit nAbs. Unusually potent nAb titers developed in 2 of 8 rabbits immunized with virus-like particles (VLPs) expressing trimers (trimer VLP sera) and in 1 of 20 rabbits immunized with DNA expressing native Env trimer, followed by a protein boost (DNA trimer sera). All 3 sera neutralized via quaternary epitopes and exploited natural gaps in the glycan defenses of the second conserved region of JR-FL gp120. Specifically, trimer VLP sera took advantage of the unusual absence of a glycan at residue 197 (present in 98.7% of Envs). Intriguingly, removing the N197 glycan (with no loss of tier 2 phenotype) rendered 50% or 16.7% (n = 18) of clade B tier 2 isolates sensitive to the two trimer VLP sera, showing broad neutralization via the surface masked by the N197 glycan. Neutralizing sera targeted epitopes that overlap with the CD4 binding site, consistent with the role of the N197 glycan in a putative “glycan fence” that limits access to this region. A bioinformatics analysis suggested shared features of one of the trimer VLP sera and monoclonal antibody PG9, consistent with its trimer-dependency. The neutralizing DNA trimer serum took advantage of the absence of a glycan at residue 230, also proximal to the CD4 binding site and suggesting an epitope similar to that of monoclonal antibody 8ANC195, albeit lacking tier 2 breadth. Taken together, our data show for the first time that strain-specific holes in the glycan fence can allow the development of tier 2 neutralizing antibodies to native spikes. Moreover, cross-neutralization can occur in the absence of protecting glycan. Overall, our observations provide new insights that may inform the future development of a neutralizing antibody vaccine.
| Here we show that native HIV-1 Env spikes expressed in a natural membrane context can induce potent tier 2 nAbs in rabbits. These antibodies reacted exclusively with epitopes present on these trimers and not with isolated Env subunits. Intriguingly, the neutralizing sera were found to take advantage of natural gaps in the carbohydrate defenses of Env spikes of the vaccine strain. Some sera were able to neutralize heterologous isolates, provided that a key, regulating glycan was removed. Overall, these findings suggest that native, membrane-expressed trimers hold promise for further development as vaccine candidates. In the future, by adapting our current findings, we might be able to encourage nAb development to key conserved sites by introducing additional, targeted gaps in the trimer's glycan shell. We suggest that the rare ability to predictably induce potent autologous neutralizing antibodies to field isolates, as we report here, provides a foundation for exploring new strategies aimed at inducing neutralization breadth which is widely expected to be essential for vaccine-induced protection.
| Eliciting broadly neutralizing antibodies (bnAbs) is a major goal of HIV-1 vaccine development [1,2]. NAbs block infection by binding to native Env spikes, consisting of trimers of gp120/gp41 heterodimers [2,3]. However, the compact, sequence-diverse, and heavily glycosylated nature of these trimers allows the virus to largely evade neutralization [4,5].
For a neutralizing antibody vaccine to be sufficiently effective, it will have to overcome at least three challenges: i) to consistently induce nAbs in all vaccinees, ii) to induce nAbs that can potently neutralize tier 2 field isolate(s) resembling transmitted strains, and iii) to induce nAbs that are effective against a broad spectrum of tier 2 strains. An ideal vaccine would resolve all these challenges simultaneously. However, most current vaccine candidates usually elicit weak or undetectable autologous tier 2 nAbs, let alone any breadth [1,2,6]. In natural infection, autologous nAbs typically develop within a few months and invariably precede any bnAb development [7]. This may be a reflection of the unprecedented sequence diversity that makes cross-reactive epitopes extremely rare among the exposed targets available on native trimers. A plausible solution may therefore be to first develop a platform that consistently elicits potent autologous tier 2 nAbs, then to use heterologous boosts to try to recapitulate the steps in nAb breadth development in natural infection [8–11]. In other words, we might implicitly solve the challenges described above in a stepwise manner.
Resolving the first challenge (consistent nAb induction) may be facilitated by ensuring that relevant epitope(s) are well-exposed. For example, previous studies have reported that several animals that received JR-FL strain-based immunogens developed modest nAb responses that target the CD4 binding site (CD4bs) [12,13]. To resolve the second challenge (inducing potent tier 2 nAbs), clearly, nAb titers should be sufficient to protect against incident infection. Studies suggest that a ~1:200 nAb ID50 titer (in the TZM-bl assay) can protect against low dose SHIV challenge [14–19]. However, factors such as the nature of the challenge virus, its dose, and nAb specificity complicate any firm estimates. Conservatively, an ID50 titer >1:1,000 might be expected to be protective. In one study, rabbits immunized with a JR-CSF gp120 DNA prime-gp120 protein-boost regimen induced exceptional nAb ID50 titers of >1:10,000 to the tier 2 index virus in the TZM-bl assay, and targeted epitopes involving the gp120 C3/V4 region [20]. Other studies have also shown that DNA prime-soluble protein boost regimens frequently elicit improved nAbs compared to protein-only regimens, although titers usually fall short of what may be protective [13,21,22]. These findings provide reasons to be optimistic that the first two challenges in neutralizing antibody vaccine development can be addressed.
Since functional, trimeric Env spikes stringently resist binding by all but the most precisely targeted nAbs, perhaps only these spikes themselves possess the necessary selectivity to elicit nAbs in a vaccine setting. Indeed, the slow progress in nAb vaccine development may derive from the fact that most Env vaccine candidates insufficiently resemble native spikes [23] and consequently elicit largely “off target” (i.e. non-neutralizing) antibodies [1,2,6]. In an attempt to address this problem, one group generated a "near native" soluble trimer, termed BG505 SOSIP.664, that elicits largely consistent and potent autologous nAbs [24]. This, and any other vaccine approaches based on authentic Env spikes clearly deserve further attention.
To achieve a "near native" conformation, soluble Env vaccine candidates require mutations to increase their stability. However, this comes at a price: to varying extents, these mutations inevitably sacrifice a fully native conformation [23]. In contrast, Env trimers expressed in a natural lipid membrane context do not require trimer-stabilizing mutations, and, unlike their soluble counterparts, fully resemble functional spikes found on infectious virus [25]. Virus-like particles (VLPs) provide one platform for testing membrane-expressed native trimer vaccines. In support of this approach, it is worth noting that all licensed infectious disease vaccines (e.g. HBV, RV, HPV) and many others in development (e.g. influenza, malaria (RTS,S), parvovirus, NDV, RSV, norovirus) are particle-based [26,27]. In the HIV arena, particulate vaccines have so far been explored in the forms of live inactivated viruses, VLPs, liposomes and virosomes (many references are cited in [28]), although none have yet demonstrated a great capacity to elicit tier 2 nAbs.
At least two major factors could underlie the lack of progress in developing nanoparticle vaccines to prevent HIV-1 acquisition. First, germline antibody precursors heavily favor protein-based epitopes over glycan epitopes, as glycans are generally considered to be “self” antigens. Given that accessible protein sites on the trimer are protected by a heavy carbohydrate shell [5,29,30], the germline antibody repertoire may therefore have a limited capacity to engage the trimer, negatively impacting nAb development [8,31–34]. This constraint may be particularly relevant in small animal models whose antibody repertoires may not be well equipped to recognize such challenging antigens [35,36]. By comparison, other, simpler forms of Env such as the gp120 monomer are more accessible and can therefore engage antibody germlines more easily. However, an important drawback is that they lack the ability to selectively elicit nAbs. Despite these challenges, a growing portfolio of broadly neutralizing human monoclonal nAbs (mbnAbs) has revealed various new ways that trimer defenses can be breached and provide paradigms for vaccine design. In many cases, these mAbs target recessed protein-based epitopes that are either bordered by glycans or make direct contacts with glycans [37–49].
A second factor that may underlie the lack of progress in particulate HIV-1 vaccine development may be that their surfaces are contaminated with non-functional forms of Env, including uncleaved (UNC) gp160 and gp41 stumps. These aberrant forms of Env may promote the development of non-neutralizing responses, perhaps at the expense of the development of neutralizing responses directed to the more compact native trimer [28,50,51]. In other words, they may act as antigenic decoys. To address this problem, we previously showed that protease treatment can selectively remove non-functional Env from VLP surfaces, leaving native trimers intact. The resulting particles are termed “trimer VLPs” [25,52]. Strikingly, the IC50 titers of monoclonal antibody (mAb) binding to trimer VLPs and neutralization correlate well [25].
To evaluate the ability of native trimers to induce nAbs, here we immunized rabbits and guinea pigs with high doses of trimer VLPs. Two rabbits developed remarkably potent serum nAbs. We compared these sera to another rare, potent JR-FL neutralizing serum generated in a rabbit immunized with DNA that expresses native trimers followed by a soluble protein boost. All 3 sera targeted quaternary epitopes that took advantage of holes in the trimer’s carbohydrate shell left by the natural absence of glycans in the C2 domain of JR-FL gp120. The VLP sera were also able to neutralize other clade B tier 2 isolates when the same glycan-deficient gap was introduced, suggesting that they target a conserved site to which access is usually regulated by a glycan. We discuss the impetus of these results for the further development of trimer VLP immunogens.
Our prior work suggested that the ability of VLPs to induce tier 2 nAbs may be improved by eliminating antigenic interference by non-functional forms of Env [50–52], by increasing the immunogen dose, and by use of a model species with a sufficiently complex antibody repertoire to enable responses to the native Env trimer [28]. Fig 1 provides an overview of a panel of reference sera and five groups of small animal vaccine sera. The reference panel includes four HIV-1 donor plasmas (1702, N160, 1686 and BB34) [53,54], an uninfected human control plasma (210), and an anti-JR-FL gp120 monomer serum pool from rabbits (described previously as "R1" in ref. [28]).
To facilitate comparisons between different vaccine regimens, all animals in groups 1–5 of Fig 1 were immunized with vaccines that present various forms of JR-FL Env. Animals in immunization groups 1–4 were immunized with VLPs bearing gp160∆CT Env in AS01B adjuvant [28,51,53,55]. An E168K mutation, with or without an additional N189A mutation, was used to partially or completely introduce the broad PG9 mAb epitope [25,52,56]. We used higher VLP doses here than in our previous studies [28,50]. Normalized VLP doses used in guinea pigs (375μg/kg; group 3) were higher than those used in rabbits (150μg/kg; groups 1, 2 and 4), assuming mean masses of 0.8kg for guinea pigs and 4kg for rabbits.
VLP immunogens administered to animal groups 1–3 (Fig 1) were treated with proteases to remove non-functional Env, leaving native Env trimers intact (termed “trimer VLPs”) [25,52]. Groups 2 and 3 received VLPs bearing "SOS" mutant Env that introduces a gp120-gp41 disulfide bond [57]. Previous studies have shown that the various Env modifications in these VLP immunogens (i.e. gp41 tail truncation, E168K, N189A and SOS mutations) all have negligible effects on the tier 2 phenotype, compared to the full-length, unmutated JR-FL parent, thereby justifying their use here [53,55,56]. For reference to the trimer VLPs administered to groups 1–3, undigested VLPs were used to immunize group 4 rabbits. During the course of these immunogenicity studies, the plasmids used to express VLPs changed (Fig 1). Thus, 1st generation VLPs, were expressed using the subgenomic pNL-LucR-E- plasmid (abbreviated as pNL-Luc) to induce budding [28]. Conversely, 2nd generation VLPs were expressed using a plasmid expressing SIV p55. In this case, a Rev-expressing plasmid was also co-expressed to enhance Env mRNA export and thereby boost Env expression (Rev is naturally encoded by pNL-LucR-E- used in 1st generation VLPs). In a BN-PAGE analysis, 2nd generation VLPs exhibited markedly improved trimer expression compared to 1st generation VLPs (S1 Fig compare lanes 7 and 8). Like their predecessors, 2nd generation trimer VLPs were also preferentially recognized by neutralizing mAbs [25]. The SOS mutation led to improved trimer expression compared to WT (S1 Fig, compare lanes 5 and 6 to lane 7; [51,53,55]). Protease digestion substantially (albeit incompletely) cleared non-functional Env, including UNC gp160∆CT monomers and gp41 stumps (S1 Fig, compare lanes 1–4 to lanes 5–8; [52]). The residual undigested monomer (see lanes 7 and 8 in S1 Fig) is probably a minor species of UNC gp160 that bears complex glycans, enabling it to survive protease treatments [52]. Overall, this analysis confirms that animals in groups 1–3 were immunized with VLPs bearing predominantly native trimer (S1 Fig, lanes 6 and 8), whereas group 4 animals received VLPs that bear a higher proportion of non-functional Env (S1 Fig, lane 2).
The success of trimer VLPs as immunogens could be adversely affected by protease damage. On the other hand, we know from our previous work that, perhaps surprisingly, VLPs remain fully infectious following protease treatment [25,52]. We also know that trimer VLPs remain intact during ELISA analysis [25,52]. To investigate the stability of our VLP immunogens in more detail, we followed their decay over time at 4°C and 37°C using infectivity and BN-PAGE as readouts of trimer function and stability, respectively (S2 Fig). In brief, we found that trimer VLP infectivity decayed more rapidly (t1/2 of 1.4h) than untreated VLPs (t1/2 of ~64.5h) at 37°C (S2A Fig). However, residual infectivity (i.e. functional trimer) was nevertheless still detected at 72h. At 4°C full infectivity was retained indefinitely, regardless of protease digestion (S2A Fig). These observations were perfectly complimented by the survival of native trimer under the same conditions, as measured by BN-PAGE-Western blot (S2B Fig). Overall, we conclude that, while trimer VLPs survive protease treatments, they are prone to subsequent decay at physiologic temperatures. Nevertheless, since native trimers and infectivity can still be detected at 72h, they may survive sufficiently long in vivo to be able to induce nAbs.
To provide a comparison to our VLP sera, another group of rabbit sera (group 5 in Fig 1) consisted of the best responders from 4 different DNA prime-soluble Env boost immunogenicity studies. The most potent group 5 serum was from animal 647 that had been immunized with pSVIII SOS gp160∆CT plasmid DNA, followed by a single gp140 foldon (gp140F) trimer boost. Nineteen other rabbits immunized with the same or related DNA prime-boost or gp140F only regimens did not develop potent autologous nAbs (S1 Table). Interestingly, the 647 animal was also the only one to develop high titer nAbs against the tier 1A MN strain after 3 DNA primes, i.e., before it was boosted with gp140F trimers (S1 Table). Thus, a particularly effective response to DNA priming may have imprinted tier 2 JR-FL nAbs that were expanded after a single protein boost. Notably, the neutralizing ID50s (particularly against tier 1 viruses) in many animals decreased following the second protein boost (S1 Table). One explanation might be that the rest period between protein boosts 1 and 2 may have been insufficient for B cells to return to a resting state (4 weeks). Another possibility is that there may be competition between lineages initiated by DNA priming and those initiated by the first protein boost. If the protein-initiated responses predominate, this could explain the transient dip in titers at the second boost. Yet another possibility could be an increasing focus on strain-specific epitopes (i.e. JR-FL-specific epitopes not present on the MN or SF162 strains).
The 2922 serum of group 5 was the most potent of 10 rabbit sera arising from a gp120 DNA-soluble gp120 monomer boost vaccination study (S2 Table; [20]). It is worth noting that the autologous nAb ID50s here were weaker than those observed in a previous study using the same regimen based on the JR-CSF strain [20]. This suggests that the Env clone markedly affects nAb induction by this regimen. Serum 849 was derived from an animal immunized with JR-FL SOSIP gp140 in a DNA prime-protein boost regimen [58]. Finally, the 7672 serum resulted from immunization with a mutant gp120 DNA prime-protein boost regimen that contained a graft of the MPER region within the V2 loop (G2C mutant) [59].
We previously showed that neutralization exhibits an excellent correlation with the ability of antibodies to recognize native trimers as observed in BN-PAGE-Western blot [3,25,50–54,64,65]. Consistent with their neutralizing activities, mAb b12 and neutralizing sera from animals 613, 647, 849, 2922 and 7672 all bound to the native SOS E168K trimer, as evidenced by depletion of the unliganded trimer (Fig 3A, lanes 2, 6 and 8–11). Trimer binding by serum 849 was rather weak, consistent with its modest neutralizing ID50 (Fig 2B). All other samples failed to bind parent JR-FL trimers, consistent with their lack of neutralizing activity (Fig 3A). Recognition of SOS D368R mutant trimers by the 613, 647, 2922 and 7672 sera suggest that these sera recognize native trimers via D368-independent epitopes. In similar studies, serum 347 also recognized native trimers in a D368-independent manner. In contrast, binding by mAb b12 and serum 849 (SOSIP) was eliminated, consistent with their dependency on residue D368 for trimer binding [12]. It is worth noting here that IgG-trimer complexes are rarely visible in these experiments, as we reported previously [25]. This is likely to be due to IgG bivalency and flexibility. Thus, a single IgG can potentially engage two trimers, two or three individual IgGs can engage a single trimer and higher order IgG-trimer complexes are possible. This plethora of possibilities probably explains the general lack of well-defined trimer-IgG complexes in these experiments. In fact, several products are often observed, reflecting the various complex combinations. This issue may be exacerbated by IgG flexibility, which may cause band smearing.
To detect any MPER nAbs, we evaluated sera in a post-CD4CCR5 neutralization assay format we described previously [55], and observed no activity (S4 Table). In contrast, the previously reported BB34 serum (serving here as a reference control) showed some activity [54]. To test whether these sera share contacts on Env with known broadly neutralizing mAbs, we also examined the effects of various knockout mutants on neutralization sensitivity, including N160A, N295Q, and N332Q. None of these mutations eliminated the neutralizing activities of the 613, 347 and 647 sera, suggesting that they must contact the trimer via other sites (S4 Table).
The poor tier 1 neutralizing activities of our potent vaccine sera (S3 Table) raise the possibility that nAbs may target quaternary tier 2 epitope(s) that are inaccessible on more sensitive HIV-1 strains. To address this possibility, we tested the ability of JR-FL monomeric gp120 and gp140F trimer to interfere with their neutralizing activities (S7 Fig). A D368R mutation was introduced into these soluble Env competitors to prevent them from binding to cellular CD4 and thereby directly inhibiting infection. We already showed that several of our sera (613, 347, 647, 2922, and 7672) recognize native Env trimers independently of the D368R mutation (Fig 3). Therefore, any interference would indicate that nAb epitopes are present on these soluble forms of Env. In S7 Fig, gp120 monomer and gp140F trimer both interfered with mAb 2G12 neutralization, but not with b12 neutralization. The latter was expected, because the D368R mutation is known to ablate b12 recognition. The soluble Envs also interfered with neutralization by group 5 serum 2922, as expected, considering this animal received a gp120 DNA prime, gp120 monomer boost regimen. In contrast, the sera from animals 613, 347 or 647 were unaffected (S7A and S7B Fig), suggesting that their neutralizing epitopes are not expressed on soluble forms of Env. This contrasts with the soluble Env interference we observed with VLP sera in our previous study [28].
Given that gp140F trimers were used as a protein boost in animal 647, the lack of D368R gp140F trimer interference was rather surprising. However, since quaternary nAbs developed after only one protein boost (S1 Table), then the boosting may have merely expanded antibodies that had been imprinted by DNA priming. The exceptional MN neutralizing activity present in this animal before protein boosting (S1 Table) suggests an unusually strong response to DNA priming in this animal that could be consistent with such imprinting. Nevertheless, the tier 1 nAbs against the JR-FL A328G were susceptible to monomeric gp120 interference, suggesting they are mediated by antibodies that do not depend on quaternary epitopes, unlike the tier 2 nAbs in this serum (S7C Fig).
We next assessed the ability of repeated serum adsorption to densely packed cells expressing native Env trimers to deplete neutralizing activity [44]. Prior to adsorption, pure IgG was extracted from the 613 and 7672 sera (the latter was included as a control). IgG concentrations were then adjusted to match that of their respective parent sera (as verified by ELISA). After adsorption to cells, IgG was again purified and then adjusted back to its pre-adsorption volume. This process resulted in heavily depleted VRC03 and 2G12 neutralization (30- to 100-fold reduction; Fig 4A). NAbs in the 613 and 7672 sera were also dramatically depleted (>100 fold). However, the 613 serum IgG was only modestly depleted by adsorption (~4-fold; Fig 4B). Therefore, the loss in neutralizing activity was due to specific adsorption to the native trimer, rather than non-specific antibody loss during the adsorption process. This data supports the idea that the 613 serum contains nAbs directed to quaternary epitope(s).
To further map the potent vaccine sera, we assessed their ability to compete with various mAbs for binding to trimer VLPs [28,60]. Here, competitions reflect direct epitope overlaps or conformational relationships on the native Env trimer. For simplicity, we infer them to indicate epitope overlaps. In addition to our vaccine sera, for reference purposes, three previously mapped HIV-1 plasmas were also analyzed ([54]; see Fig 1). All rabbit serum-mAb competitions were referenced to a rabbit prebleed control. Similarly, HIV-1 donor sera were referenced to the HIV-1 seronegative human plasma from donor 210. Our panel of biotinylated mAbs is organized horizontally in Fig 5, starting with those directed to epitopes at the membrane-distal apex of the trimer on the left and ending on the right with those directed to the membrane-proximal ectodomain region (MPER). Competitions were considered as significant when biontinylated mAb binding was reduced to <50% of control levels (indicated by colored squares). A non-neutralizing reference serum from rabbit 618 did not compete with any mAb to <50%, suggesting that this cutoff is sufficiently stringent to rule out any non-specific inhibition by our vaccine sera. Multiple representatives of major epitope clusters were assayed to ensure the consistency of any inhibitions within these clusters. Each assay was repeated at least 2 times. The generally low standard errors we observed in S8 Fig reveal the consistent patterns in these repeats. In S9A Fig, we show that each mAb inhibited its biotin-labeled counterpart and other mAbs within each epitope cluster to <10% of control levels (diagonal boxes from top left to bottom right). In representative inter-cluster competitions shown in S9A Fig, as expected, 2G12 was unable to inhibit CD4bs nAb binding. Similarly, various gp120-specific nAbs were unable to inhibit the binding of biotinylated 4E10. However, a competitive inter-cluster relationship between PGT125 and CD4bs nAbs was observed, consistent with the known overlap between these epitopes [42,66]. As with the serum-mAb competitions (S8 Fig), the standard error of repeated mAb-mAb competitions (all assays were performed at least twice) were generally low. Representative titrations of biotinylated mAbs in the presence or absence of unlabeled mAb competitors are shown in S10 Fig and exemplify the strong mAb-mAb self-competition in S9 Fig.
Of all the polyclonal samples, the potently neutralizing HIV-1-infected donor plasma 1686 showed the highest overall competition, strongly inhibiting binding by all 4 CD4bs mAbs (Fig 5). This agrees with our previous mapping of this sample using other methods [54]. The 1686 plasma also competed with mAb PGT125 (Fig 5), consistent with the PGT125-CD4bs overlap mentioned above (S9 Fig). Competition of most other mAbs we tested was marginal, except for 4E10. In S10 Fig, representative titrations show strong plasma 1686 competition of mAb VRC03 directed to the CD4bs, but little or no competition of mAb PG9. The other 2 more weakly neutralizing human plasmas (BB34 and 1702) showed modest CD4bs competition, as well as 4E10 competition, consistent with their known MPER nAb activities ([54]; S4 Table). The marginal (but consistent) competition of VRC03 and to some extent with PG9 by these two plasmas is shown in S10 Fig. Overall, these findings confirm the specificity and reproducibility of this competition assay.
All the neutralizing rabbit sera showed significant competition with CD4bs mAbs VRC03, VRC07, b12 and 1F7 (Figs 5 and S10). Of these, the 347 serum exhibited the strongest activity and, like the human 1686 plasma also competed moderately with mAb PGT125. This serum also competed modestly, but consistently with mAbs PG9 and PG16 (Figs 5 and S10). We suggest that serum 347 binding to trimer has an allosteric effect on the integrity of these V2/quaternary epitopes. The vaccine sera showed little competition with other mAbs. The 613 and 7672 sera showed a similar pattern, competing with all 4 CD4bs nAbs, with possible weak inhibition of PGT125, PG9 and PG16. The 647 and 2922 sera showed weaker competition that was also focused on the CD4bs. In the case of the 647 serum, the unexpectedly weak competition was because limited sample availability meant that we had to use serum from a bleed taken 4 weeks after the immune bleed used elsewhere in this study, by which time the neutralizing ID50 had fallen to ~1:80.
The patterns of inhibition by vaccine sera are represented as raw titrations in S10 Fig, where VRC03 binding is clearly perturbed, but the effects on PG9 and 2G12 are weak or absent, respectively. Although the extent of competition by vaccine sera was in no case as complete as with mAb-mAb competitions (S10 Fig), competition patterns were consistent between assays (S8 Fig). Furthermore, in most cases, inhibition of all four different mAbs within the CD4bs cluster was detected (Fig 5). The competition profiles of our vaccine sera patterns are also strongly supported by our observing expected patterns of competition by HIV-1 donor sera. It may be no coincidence that all vaccine sera overlap with the CD4bs, considering that this is one of the few partially exposed protein patches on the otherwise densely glycan-populated trimer. The observation that the 347 and 7672 sera exhibited stronger competition of this site than the 613 serum, despite having somewhat lower neutralizing ID50s may indicate that their neutralizing activities have a greater overlap with the CD4bs.
Like all other vaccine sera described to date, none of our sera significantly neutralized tier 2 viruses, aside from the autologous JR-FL parent virus. The lack of JR-CSF cross-neutralization is sobering, given that this strain was isolated from the same donor as JR-FL and therefore exhibits a degree of sequence homology. This lack of JR-CSF cross-neutralization, however, has the benefit of allowing us to map our sera by measuring their activities against a series of JR-FL Env-based chimeric pseudoviruses with various JR-CSF Env domain swaps [20,67]. Most of these chimeras were infectious (S11A Fig). All of the functional clones resisted neutralization by the V3 loop-specific mAb CO11 [28], suggesting that they retain a tier 2 phenotype. The potent sera from animals 613 and 647 neutralized all of the functional chimeras (S11A Fig). This suggests that they target epitopes involving the JR-FL C1 and/or C2 regions that were not represented in the chimera neutralization analysis, as they were not sufficiently functional (S11A Fig). In contrast, sera 2922 and 7672 targeted epitopes affected by the C-terminal portion of gp120 (S11A Fig). Specifically, the 2922 gp120 DNA prime-protein boost sera targeted the C3V4 region, which is similar to the epitope targets of the sera previously reported using this regimen using the JR-CSF strain [20], suggesting a common nAb development pathway. The 7672 G2C gp120 prime-boost serum was knocked out by a C3-C5 chimera (clone 8070), although its activity could not be further resolved by less substantial domain swaps (S11A Fig). These results are summarized in Fig 6A.
Since the competition data in Fig 5 suggested that our potent sera target CD4bs-like epitopes, we first investigated the possible role of the C2 region in the 613 and 647 epitopes, due to its close proximity to the CD4bs. Alignment of JR-FL and JR-CSF C2 regions (Fig 6B) reveals 8 amino acid differences, including 3 additional glycan sequons in JR-CSF. These substantial differences could account for the lack of infectivity of the C2 domain swap mutants (8076 and 8086) in S11A Fig—JR-FL trimers may be unable to accommodate all 3 extra glycans. An analysis of JR-FL-based point mutants at the 8 variant sites revealed that the 613 serum was sensitive to a D197N mutation at the base of the V1V2 loop that introduces one of the three additional JR-CSF glycans (Fig 6B and 6C). Several other D197 mutants (D197N+S199A, D197Q and D197K) that do not introduce a new glycan had no effect, suggesting that the N197 glycan regulates 613 serum neutralization (Fig 6C). None of the other C2 mutants affected neutralization by the 613 serum. It is worth noting that the introduction of the N197 glycan into JR-FL trimer also leads to a ~4 fold reduction in b12 IC50 (Fig 6C), consistent with its role in a putative “glycan fence” that regulates access to the CD4bs [64,67]. In contrast to the 613 serum, the 647 serum was affected by mutations of residues D230, P238 and E269 in loops A and C of the C2 region that straddle a disulfide-bonded hairpin (Fig 6B and 6C). In this case, neutralization was not exclusively eliminated by the introduction of a new glycan at residue 230, but instead depended on the presence of an aspartic acid at this position. The resistance of all these mutants to the CO11 V3 mAb suggests that they retain a tier 2 phenotype. Therefore, we can conclude that these mutants directly affected serum recognition rather than having an adverse effect on trimer folding.
Taken together, the above mapping data suggest that both of our potent sera take advantage of the relatively sparse glycan coverage of the JR-FL C2 region. Notably, the N197 glycan is conserved in 98.6% viruses of a multi-clade panel of >4,000 sequences (Fig 6B). Further experiments using a set of JR-CSF Env-based chimeras (S11B Fig) revealed that two mutants (8073 and 8087) that eliminate the N197 glycan knocked in 613 serum neutralization sensitivity, suggesting that the N197 glycan regulates tier 2 cross-reactivity. These mutants exhibited increased b12 sensitivity [67], consistent with a role of N197 glycan in the glycan fence. In contrast, all JR-CSF-based chimeras remained insensitive to the 647 serum, suggesting that, in this case, neutralization is highly strain-specific and context dependent. An expanded set of JR-CSF-based chimeras similar to the JR-FL chimeras in S11B Fig were also tested, but none were sensitive to the 613 or 647 sera.
We next asked if the N197 glycan regulated the sensitivity of other isolates to our sera, as it did for the JR-CSF isolate. To answer this question, we generated N197 glycan-deficient mutants in 28 other tier 2 isolates sampled from several clades, all of which resisted 613 serum neutralization in the presence of this glycan (S5 Table). Remarkably, the removal of the N197 glycan rendered 9 of these viruses (i.e., 31%) sensitive to the 613 serum (S5A Table and Fig 7). All 9 viruses were from clade B and constituted 50% of the clade B N197 knockout mutants tested (n = 18; Fig 7). Our other neutralizing trimer VLP vaccine serum (from animal 347) neutralized 3 of these 9 viruses (JR-FL, JR-CSF and ADA; S5A Table). Neither serum neutralized any of the 11 non-clade B viruses, which included Envs from clades A, C and several Chinese B' isolates [68]. Not surprisingly, the 647 serum did not neutralize any of the N197 mutants.
To properly interpret these findings, it was important to monitor any overt changes associated with the removal of the N197 glycan that might signify a loss of tier 2 phenotype and therefore may impact sensitivity to our sera. Accordingly, our mutants were further characterized using a panel of mAbs (S5 Table). These included 6 non-nAbs, 14e, 39F, b6, F105, 17b and 48d, directed to V3, CD4bs and CD4i epitopes. In addition, we monitored sensitivity to a weakly neutralizing HIV+ serum, BB68 [54]. We arbitrarily ascribed global sensitivity to any N197 mutant that was sensitive to at least 2 of the 6 non-nAbs. Six N197Q mutants (JR-CSF, ADA, SC422, YU2, PVO.4 and BaL) were thus found to be globally sensitive and were therefore excluded from further analysis (S5C Table). Nevertheless, since 5 of these mutants were sensitive to the 613 serum and 4 were sensitive to the 347 serum, we decided to remake 4 of them (JR-CSF, ADA, YU2 and PAVO) as N197D mutants in the hope that the different amino acid exchange would preserve a tier 2 phenotype. A N197D BaL mutant was not made, however, because the parent virus was sensitive to two of our non-nAbs (S5C Table). All 4 new N197D mutants retained a tier 2 phenotype. Moreover, JR-CSF and ADA mutants were sensitive to both the 613 and 347 sera, the YU2 N197D mutant was sensitive to the 613 serum and the PVO.4 mutant resisted both sera. Thus, all 9 613-sensitive mutants shown in Fig 7A were verified to have a tier 2 phenotype, consistent with modest tier 2 breadth within clade B by recognition of the surface protected by the N197 glycan. The 347 serum also exhibited modest breadth that might have been limited by the fact that it was taken at bleed 3 rather than bleed 4. Although this “breadth” depends on the absence of the N197 glycan, since all the mutant viruses retain a tier 2 phenotype, our data suggest that addressing the key problem of breadth in vaccine design may in future be possible.
We next measured the sensitivity of N197 mutant panel to various neutralizing ligands to determine any patterns that might partition with serum 613 sensitivity (S5 Table), using a previously published fingerprinting analysis [69]. Computational analysis suggests that strain selection can improve the predictive accuracy of this approach, i.e., the ability to discriminate between the different specificities. A recently developed method for virus panel selection was applied to determine any subsets (including all 9 613 serum-sensitive viruses) of the 29 member N197 mutant panel in S5 Table that are more suitable for the fingerprinting analysis compared to the full panel [69]. All possible panels of sizes 20–28 were evaluated. The best panel included 23 N197 mutant strains. Although this was preferable to the full 29-strain panel, neither were optimal and could therefore adversely affect the reliability of any predictions. Our panel size was limited by the practical challenge of making N197 mutants for each Env clone and performing a full neutralization analysis. In the fingerprinting analysis, neutralization via a particular epitope is typically associated with delineation values of >0.3, and preferably >0.4. However, we have previously confirmed positive signals of >0.2 [5,49,69]. In this case, a 23-strain panel and 12 bnAb specificities were used for fingerprint delineation. The 1F7 mAb was included because it is known to be N197-sensitive [67]. The strongest fingerprint signal for the 613 serum was for PG9-like antibodies (Fig 8). It may be no coincidence that all 9 of the 613-sensitive N197 mutants were also sensitive to at least one of the mAbs PG16, PGT145 and VRC26, whose epitopes depend on a compact quaternary conformation (S5 Table). This was not true for 4 of the 9 clade B 613 serum-resistant mutants (1168, QH0515.01, 6101 and BL01). Several non-B clade mutants also resisted these bnAbs. It should be noted that the 613 serum can neutralize the parent JR-FL strain in the absence of a lysine at residue 168 of the V2 loop that is critical for known V2 quaternary-dependent neutralizing mAbs. Therefore, in keeping with the marginal competition in Fig 5, it is unlikely that the 613 serum targets this epitope cluster. Rather, we suggest that trimer compactness is important for neutralization by both the 613 serum and by broadly neutralizing V2 quaternary mAbs.
Cumulatively, CD4bs nAbs also showed a strong fingerprint signal; VRC01-like signals were the strongest, along with detectable b12-like and 1F7-like signals (Fig 8). This is consistent with the solid 613 serum-mediated competition of CD4bs mAbs (Fig 5). Overall, N197 mutants exhibited a trend for slightly greater sensitivity to CD4 ligands. Specifically, 84.9% of virus-ligand combinations in which neutralization was detectable (146 of 172 virus-ligand combinations) exhibited a >2 fold increase in sensitivity. This is consistent with its role in the glycan fence that regulates CD4bs access. There were no consistent effects on several other bnAb epitopes analyzed in S5 Table.
Despite these observed PG9-like and CD4bs-like patternsof serum 613, none of the delineation signals were exceptionally strong. This is consistent with a unique epitope that does not fall entirely into any known category. Rather, it suggests a CD4bs-like quaternary epitope that exhibits some characteristics of both clusters. An alignment of the 18 clade B sequences derived from Fig 7 and S5 Table did not reveal any consistent elements (e.g. glycans, V1V2 length, CD4 contact regions) that partitioned with 613 serum sensitivity (S12 Fig). Similarly, an investigation of a variety of phylogenetic tree models (neighbor join, UPGMA, maximum parsimony) and substitution models (Poisson, p-distance) rooted to various reference sequences failed to identify any distinct clustering of 613-sensitive viruses, in part due to the low confidence of bootstrap values. Ultimately, high sequence variability defied any attempt to decipher common element(s) that might predict sensitivity. Perhaps the only clear conclusion we can make is that it is probably no coincidence that all the sensitive viruses were derived from clade B.
Here, for the first time, we showed that authentic trimers can elicit potent nAbs directed to quaternary epitopes that take advantage of strain-specific holes in the glycan shield. This establishes a mechanism by which autologous antibodies can potently neutralize tier 2 isolates in a vaccine setting. The binding sites of two potent sera (from animals 613 and 647) are perhaps best visualized as footprints on the near native BG505.SOSIP.664 Env trimer model (Fig 9) [5,30]. Given the collective features of the 613 serum nAb—i) CD4bs overlap, ii) N197 glycan-sensitivity, iii) modest PGT cluster II overlap, iv) quaternary dependency, and v) CD4bs/PG9-like fingerprint—we suggest that it targets a CD4bs-like hole in the glycan shield that is accessible only on the trimer, possibly close to the gp120 protomer interface. To reach the CD4bs cavity, antibodies must navigate through a putative "glycan fence" [64], consisting of glycans N197 (in red in Fig 9) [70], N276 [10,32], N362, N386 (all in magenta), and possibly others, depending on the Env strain (Fig 9A). The difficulty of this challenge is illustrated by the exquisite docking of CD4bs mAb VRC01 on the trimer (Fig 9B). The 613 serum nAb must also navigate this fence, albeit with the N197 glycan missing in the JR-FL strain (modeled in Fig 9C and 9D).
Serum 647 nAbs make contact with loops A and C of the C2 region that straddle the inner and outer domains (labeled red in Fig 9E). In this BG505 trimer model, we labeled residue 241 in black to signify the presence of a glycan in the JR-FL trimer that happens to be absent in BG505 Env. Conversely, we removed a glycan at residue 234, consistent with its absence in JR-FL Env trimers. Like the 613 serum nAb, the 647 serum appears to take advantage of a rare glycan-deficient patch (modeled in Fig 9F). The position of the 647 epitope on the lower edge of the gp120 outer domain is compatible with a lateral angle of approach similar to that of recently reported mAb 8ANC195, which also targets the C2 hairpin and competes with CD4bs antibodies [47,71]. However, unlike the 647 serum, mAb 8ANC195 recognizes the common N234 glycan that is absent in JR-FL [72]. Even when this glycan was introduced into JR-FL Env, 8ANC195 neutralization remained undetectable and thus prevented us from being able to compete serum 647 and mAb 8ANC195 in the trimer VLP ELISA (as in Fig 5). This difference in specificity might in part explain 8ANC195's breadth, as compared to the strain-specificity of the 647 serum. The N234 glycan present in BG505 and 79.7% of 4,265 published strains is added back in Fig 9G, revealing a possible clash with 647 serum nAb binding. Overall, the epitopes of our 3 potent sera add to previous evidence that the JR-FL Env can quite frequently induce nAbs that overlap the CD4bs [12,13]. This may be a result of strain-specific gaps in its glycan armor of the gp120 C2 region that protects the CD4bs. Similar glycan-deprived sites can also be targets for autologous neutralization in natural infection [9,11].
The potent nAb responses in rabbits 613 and 347 far exceeded any we observed in previous VLP immunogenicity studies [28,50]. This may be in part due to improvements that were suggested by these earlier studies: i) that high doses may be important, ii) that rabbit models may be more competent as nAb vaccine models, and iii) that it may be important to refocus responses on the native trimer by eliminating antigenic interference by non-functional Env. The use of more efficient expression methods should in future help us to ensure that doses are sufficient to induce maximum responses. Our newly adopted Gag-only plasmids (as compared to subgenomic NL-Luc) are also safer because they lack a Ψ packaging signal and therefore do not carry a genetic payload [73]. In contrast to rabbits, the relatively high gp120 binding (Fig 2) and tier 1 nAb titers (S3 Table) but lack of tier 2 nAbs (Fig 2) in guinea pigs suggest that VLPs are immunogenic, but that this species struggles to target the native Env trimer and instead target degraded forms or Env that may appear later.
Different stochastic processes might underlie the inconsistent nAb development to the various immunogens we tested. In the case of the gp120 prime-protein boost regimen used in the 2922 animal, since monomeric gp120 is not an authentic target, it may only rarely induce antibodies that happen to cross-react with native trimers and instead largely induces ‘off target’ responses. In the case of regimens that involve authentic trimer as an antigen, different challenges come into play, principally that it is a complex antigen that presents few opportunities for antibody development. For this reason, tier 2 nAbs developed in only 1 of 20 animals that received a DNA prime that expresses largely authentic Env trimer (S1 Table). An additional factor may have been this plasmid DNA expresses only modest amounts of trimer to ensure a mature, authentic product, but that could adversely affect the robustness of priming. Paradoxically, however, the rare tier 2 nAbs observed in animal 647 were associated with unusually effective DNA priming, as evidenced by an exceptional ID50 to the MN strain (S1 Table). The distinct specificities of the tier 1 and 2 nAbs in this animal (S7 Fig) suggest the development of separate antibody lineages: potent tier 1 nAbs that recognize non-native forms of Env are produced during DNA priming, as well as trimer-specific tier 2 nAbs that are expanded upon the first protein boost. Clearly, this effect may be difficult to replicate with any reliability. In contrast, our data show that trimer VLPs induce only limited tier 1 nAbs and generated tier 2 nAbs more frequently, albeit still in a minority of vaccinees. This inconsistency may relate to limited ability of small animal antibody repertoires to find solutions for binding this complex antigen—a problem that may be exacerbated by the low Env spike density on VLP surfaces that may limit crucial signals necessary for affinity selection [31,74]. Overall, these observations suggest that, despite the exposed protein patch in the C2 region, the native JR-FL trimer remains poorly immunogenic.
Clearly, it will be important to try to improve the consistency of vaccine nAb responses. One approach may be to combine the approaches used here in DNA prime-VLP boost regimens [13,20,22,75]. A more intriguing approach might be to engineer additional holes in the glycan shield to encourage nAb development [32,76]. For example, CD4bs nAb development might be facilitated by selectively removing glycan fence posts to avoid clashes [32,67,70]. One concern with this strategy is that glycan-depleted trimers may be overtly sensitive to non-neutralizing antibodies, as occurs with the removal of the N301 glycan [64,77]. In addition, some glycans cannot be removed without adversely affecting trimer folding [29]. Nevertheless, in our experience, most single glycans can be removed without global effects on trimer folding and maintain a tier 2 phenotype (i.e., non-neutralizing epitopes remain occluded). In fact, the conservation and use of glycosylation sites vary considerably between isolates. It is fairly common for tier 2 isolates to lack one or two glycans, like JR-FL, that are present on the majority of other circulating tier 2 viruses. Indeed, modestly glycan-depleted tier 2 viruses can in some scenarios transmit infection, perhaps because these variants can facilitate acquisition in settings where nAb resistance is unimportant [8,78–80].
To our knowledge, our data provide the first clear evidence of potent, vaccine-elicited nAbs directed to a reasonably well-conserved site on tier 2 viruses—albeit one that is camouflaged by a glycan on most strains. In contrast, recent evidence suggests that broad HIV-1 neutralization almost always requires glycan recognition—as opposed to other viruses such as influenza, where, despite high glycan context, no glycan recognition occurs [5]. Our results therefore show how effective the glycan shield is in preventing the development of broad neutralization. With this in mind, how might we try to induce cross-reactive tier 2 nAbs in the future? One strategy might be to evolve breadth from initial autologous nAbs using modified boosts. After all, in both vaccine and natural infection settings, initial nAbs are invariably autologous and there may be no practical way to elicit bnAbs directly. The mechanisms of nAb breadth development in natural infection provide model scenarios that might be mimicked by vaccine strategies [8,9,11,81–85]. For example, breadth sometimes develops as a result of virus escape from autologous nAbs in which a newly introduced glycan creates a new broad site of vulnerability [8–11]. Indeed, while glycans are largely immunosilent, bnAbs can eventually develop that recognize composite epitopes involving glycan contacts. The evolutionary changes in nAb specificity needed to accommodate a newly added glycan might be either to avoid clashes, perhaps by altering the angle of approach to its epitope (e.g. CD4bs [86]) or to incorporate the glycan as part of its epitope. Either of these scenarios might be a useful first step towards cross-reactivity. For example, it might be possible to evolve changes in neutralizing antibodies like those observed in animal 613 by using appropriate boosts to encourage the development of antibody variants that are no longer sensitive to the N197 glycan.
In summary, this study introduces some encouraging new concepts for HIV-1 nAb vaccine development. Specifically that i) native trimers presented in situ can induce tier 2 nAbs, ii) nAbs recognize quaternary epitopes, iii) nAbs target vulnerable gaps in the glycan shield, and iv) these nAbs can neutralize other tier 2 viruses provided that appropriate gaps are artificially introduced in their glycan shell. The advances we made here with VLP immunogens required us to address various limiting factors we identified in preceding studies. A continued commitment to further iterative, rational improvements may in future bring us closer to our goal of inducing tier 2 breadth.
Monoclonal antibodies (mAbs) were obtained from their producers, the AIDS reagent repositories of the UK Medical Research Council and the NIH, or were purchased from commercial suppliers. Further information on these mAbs can be found at the web link: (www.hiv.lanl.gov). The mAb panel included the following (originators given in parentheses): 2G12 (Katinger), directed to a unique glycan-dependent epitope of gp120 [87]; 14e, 39F and CO11 (J. Robinson), directed to the gp120 V3 loop [25,28,53]; b12 and b6 (Burton), VRC01 and VRC03 (Mascola), 8ANC131 and 3BNC117 (Nussenzweig), CH103 (Haynes), HJ16 (Lanzavecchia), F105 (Posner) 15e (J.Robinson) and 1F7 (Katinger), directed to epitopes that overlap the CD4bs [25,38,40,67,72,83]; 8ANC195 (Nussenzweig), directed to the C2 region [47,71]; 17b and 48d (J.Robinson), directed to CD4-induced (CD4i) epitopes of gp120; PGT121, PGT125 and PGT128 (Burton) directed to epitopes involving the base of the V3 loop of gp120 and the N332 glycan [42]; PG9, PG16 and PGT145 (Burton), and CAP256-VRC26.08 (abbreviated as VRC26; directed to quaternary, glycan-dependent epitopes that involve the V2 loop [37,56,84]; 35O22 (Connors), directed to a quaternary epitope of the gp120-gp41 interface [49]; PGT151 (D. Burton), directed to a quaternary gp41 epitope [43]; 7B2 and 2.2B (J. Robinson), directed to the gp41 cluster I and II epitopes, respectively [11]; 4E10, 2F5 (Katinger) and 10e8 (Connors), directed to the gp41 membrane-proximal ectodomain region (MPER) [45].
Recombinant monomeric JR-FL gp120 produced in CHO cells and soluble CD4 (sCD4) consisting of its 4 outer domains were gifts from Progenics Pharmaceuticals (Tarrytown, NY). 2-domain CD4-Ig was a gift from Marie Pancera (NIH). Recombinant HXB2 gp41 was obtained from Meridian Life Science (Catalog#VTI310; residues 546–682, Saco, Maine).
Plasmid pCAGGS was used to express JR-FL gp160∆CT on VLP surfaces [51,55]. Gp160∆CT is truncated at amino acid 709, leaving a 3 amino acid gp41 cytoplasmic tail. This increases native trimer expression and can be used to produce pseudoviruses with similar neutralization sensitivity profiles compared to their full-length gp160 counterparts [53]. The use of the JR-FL Env strain has several rare advantages, including efficient expression and gp120/gp41 processing [51,53]. Mutants were generated by Quikchange (Agilent Technologies) and were numbered according to the HXB2 reference strain [54]. "SOS" mutations (A501C and T506C) introduce an intermolecular disulfide bond between gp120 and gp41 [57]. E168K and N189A mutations knock in the broadly neutralizing “PG” epitopes that are normally absent in the JR-FL isolate and increases trimer expression [25]. The A328G mutation dramatically enhances neutralization sensitivity, otherwise known as "global" sensitivity [25]. The D368R mutation abrogates CD4 binding capability [13].
Various plasmids expressing other Env gp160s were obtained from the NIH AIDS repository, including clade B and C virus panels. We also obtained a set of Chinese B' Env-expressing plasmids from Dr. Zhiwei Chen [68]. Mutant versions of these sequences in which the N197 glycan is eliminated were used for the fingerprinting analysis. Codon optimized soluble JR-FL D368R gp140 foldon (gp140F) was expressed from a pCDNA3.1(-) vector using a CD5 leader and CMV promoter [13]. Another set of plasmids encoded a series of domain-exchanged hybrid gp160s using sequences from the JR-FL and JR-CSF isolates [20,67].
Other plasmids used to make VLPs included Env-deficient sub-genomic plasmids pNL4-3.Luc.R-E- and pSG3∆Env that have been described previously [50]. Plasmids pMV-2024 and pMV-0932 express full-length SIVmac251 (BK28) Gag and codon-optimized HIV-1 Rev, respectively, both under the control of a CMV promoter.
First generation VLPs were produced by co-transfecting 293T cells with an Env-expressing plasmid and pNL4-3.Luc.R-E-, using polyethyleneimine, as described previously [25]. Second generation VLPs were produced using plasmids pMV-2024 and pMV-0932 in place of pNL4-3.Luc.R-E-. Two days later, supernatants were collected, precleared by low speed centrifugation and pelleted at 50,000 x g in a Sorvall SS34 rotor. To remove residual medium, VLP pellets were diluted with 1ml of PBS, then re-centrifuged at 15,000 rpm and resuspended in PBS at 1,000 x the original concentration. VLPs were referred to as WT-VLPs or SOS-VLPs, depending on the form of Env displayed on their surfaces or as bald-VLPs, bearing no Env, produced by transfecting pNL4-3.Luc.R-E- alone [50]. "Trimer-VLPs" were made by digestion using a cocktail of proteases including proteinase K, subtilisin, trypsin and chymotrypsin, as previously described [25,52]. VLPs were inactivated using aldrithiol (AT-2) [50].
Reference controls included a pooled serum generated from six rabbits that were immunized with monomeric gp120 in AS01B, as described previously [28]. HIV-1-infected donor plasmas BB34, BB68, 1686, 1702 and N160 and uninfected control plasma 210 were also described previously [28,54].
ELISAs were used to measure serum binding to various antigens and were also used to determine their specificities [25,28,60]. Briefly, Immulon II plates were coated with 20x concentrated VLPs, recombinant gp120 or gp41 at 5μg/ml overnight at 4°C. Following a PBS wash and blocking, sera were titrated against each antigen in blocking buffer. Species-specific alkaline phosphatase anti-Fc conjugates (Accurate, Westbury, NY) and SigmaFAST p-nitrophenyl phosphate tablets (Sigma) were then used to detect binding. Plates were read at 405nm. Titers are taken when ELISA signals exhibited an optical density of 0.5 (approximately 3x background).
In competitive ELISAs, we determined the ability of sera at a 1:10 dilution to inhibit binding of graded doses of various biotinylated mAbs to SOS E168K trimer VLPs. MAbs were biotinylated using NHS-X-biotin reagent (Calbiochem). Biotinylated mAb binding was detected using streptavidin-alkaline phosphatase (Vector, Burlingame, CA), and developed as above. A prebleed rabbit serum or an HIV-1-seronegative plasma (donor 210) were used as reference controls. Titers of biotinylated mAbs (in μg/ml) measured at OD = 0.5 (approximately 3 x background) were determined in the presence of competitor and control samples. Then, competition was expressed as the % residual binding = [(titer in the presence of the control sample)/(titer in the presence of competitor) x 100].
Blue native PAGE (BN-PAGE) was performed as described previously [3,50,51,53]. Briefly, VLPs were solubilized in 0.12% Triton X-100 in 1 mM EDTA. An equal volume of 2x sample buffer (100 mM morpholinepropanesulfonic acid (MOPS), 100 mM Tris HCl, pH 7.7, 40% glycerol, and 0.1% Coomassie blue) was added. Samples were then loaded onto a 4–12% Bis-Tris NuPAGE gel (Invitrogen) and separated at 4°C for 3 hours at 100V. The gel was then blotted onto polyvinylidene difluoride membrane, destained, immersed in blocking buffer (4% nonfat milk in PBS) and probed with an anti-gp120 cocktail (mAb b12 and 39F at 1μg/ml) and/or a anti-gp41 cocktail (mAb 2F5, 4E10, 7B2, 2.2B at 1μg/ml). Blots were then probed by an anti-human Fc alkaline phosphatase conjugate (Accurate Chemicals) and developed using SigmaFast BCIP/NBT substrate (Sigma).
BN-PAGE “shift” assays were used to measure the ability of antibodies to bind and deplete the unliganded trimer [3,25,50,51,53]. VLPs were incubated with mAb or serum for 1h at 37°C, then washed with PBS and resolved by BN-PAGE-Western blot, as above.
Experiments were conducted to determine the stability of VLPs over time after digestion with proteases.
Heat-inactivated sera and protein A-purified serum IgGs were analyzed for neutralization of various pseudoviruses produced by co-transfecting either 293T cells [64] with an Env plasmid and pNL4-3.Luc.R-E- (CF2 assays) or pSG3∆Env (TZM-bl assays). Data is representative of at least 3 repeat assays performed in duplicate.
Mann-Whitney and ANOVA two-tailed tests were used to determine any statistically significant differences between study groups.
The neutralization sensitivities of a panel of N197 glycan knockout mutants to a vaccine serum and a panel of neutralizing antibodies was used as a basis for a bioinformatics-based approach to inferring the neutralizing specificities in the serum, according to a previously published protocol [5,49,69,72]. A cutoff of ID50 = 50 (since input data was ‘<50’ in most cases) was used to denote the presence or absence of neutralization. A reference set of 12 antibody specificities was used in the analysis, with one or more representative antibodies included for each specificity: VRC01-like (VRC01, 3BNC117, 8ANC131, CH103), b12-like (b12), CD4 (sCD4, CD4-Ig), HJ16-like (HJ16), 1F7-like (1F7), 8ANC195-like (8ANC195), PG9-like (PG9, PGT145), PGT128-like (PGT121, PGT128), 2G12-like (2G12), 2F5-like (2F5), 10E8-like (4E10, 10E8), 35O22-like (35O22).
The recently reported structure of BG505 SOSIP gp140 trimer ([5]; PDB id: 4TVP) was used to generate a visual model for the distribution of glycans on the native spike. First, atomic clashes present in the 4TVP crystal structure were relieved and missing side-chains rebuilt, by executing 1,000 symmetric ROSETTA-fixbb simulations, selecting the lowest scoring model, and then running a constrained ROSETTA-relax simulation. Each N-linked glycosylation motif for each of the 15 models was decorated with Man8GlcNAc2 glycans at sites of predicted oligomannose-type glycans and with Man5GlcNAc2 glycans at remaining sites. GlycanRelax [88] was used to approximate the conformational behavior of glycans in a glycoprotein context. For each model, 10 separate GlycanRelax trajectories of 10,000 cycles of MonteCarlo trials were carried out. Each glycan on the gp120 was allowed to move independently throughout the GlycanRelax minimization. A single low energy model was chosen for Figs. All Figs were generated in PyMOL Molecular Modeling Software (Version 1.5.0.4 Schrödinger, LLC).
The archived adult human plasmas used in this study have previously been described [28,54]. All donors provided written consent for the use of these samples. Institutional Review Board (IRB) approval for this project was obtained through the San Diego Biomedical Research Institute IRB Committee (approval number: IRB-14-04-JB; Federal Wide Assurance number: 00021327).
All immunization protocols for rabbits and guinea pigs were approved (protocol PRF2A) by the Explora Biolabs Animal Care and Use Committee (IACUC). Explora Biolabs’ animal welfare assurance (AWA) number is A4487-01. Pocono Rabbit Farm, Covance and Aldevron are all approved for rabbit and guinea pig immunizations by the Association for Assessment and Accreditation of Laboratory Animal Care (AALAC). The AWA number Pocono Rabbit Farm is A-3886-01. The protocol for rabbits immunized at Covance (Denver, PA), was approved (protocol 0310–11) by the Covance IACUC Committee (AWA number A-3850-01; NIH profile number 9640807). Protocols for rabbits immunized at Aldevron LLC (Fargo, ND) were approved (protocols 2-04-007-09-2005, 1-06-001-08-2007, and 1-07-005-10-2008) by the Aldevron IACUC Committee; AWA number A4422-01). All animals were fed, housed and handled in strict accordance with the recommendations of the NIH Guide for the Care and Use of Laboratory Animals, the Animal Welfare Act and Regulations and guidelines established in 1993 by the American Veterinary Medical Association Panel on Euthanasia. Two rabbit sera (7672 and 849) were archived from previous studies in which animal welfare information was documented [12,58,59].
|
10.1371/journal.pgen.1007658 | Whole exome sequencing of ENU-induced thrombosis modifier mutations in the mouse | Although the Factor V Leiden (FVL) gene variant is the most prevalent genetic risk factor for venous thrombosis, only 10% of FVL carriers will experience such an event in their lifetime. To identify potential FVL modifier genes contributing to this incomplete penetrance, we took advantage of a perinatal synthetic lethal thrombosis phenotype in mice homozygous for FVL (F5L/L) and haploinsufficient for tissue factor pathway inhibitor (Tfpi+/-) to perform a sensitized dominant ENU mutagenesis screen. Linkage analysis conducted in the 3 largest pedigrees generated from the surviving F5L/L Tfpi+/- mice (‘rescues’) using ENU-induced coding variants as genetic markers was unsuccessful in identifying major suppressor loci. Whole exome sequencing was applied to DNA from 107 rescue mice to identify candidate genes enriched for ENU mutations. A total of 3,481 potentially deleterious candidate ENU variants were identified in 2,984 genes. After correcting for gene size and multiple testing, Arl6ip5 was identified as the most enriched gene, though not reaching genome-wide significance. Evaluation of CRISPR/Cas9 induced loss of function in the top 6 genes failed to demonstrate a clear rescue phenotype. However, a maternally inherited (not ENU-induced) de novo mutation (Plcb4R335Q) exhibited significant co-segregation with the rescue phenotype (p = 0.003) in the corresponding pedigree. Thrombosis suppression by heterozygous Plcb4 loss of function was confirmed through analysis of an independent, CRISPR/Cas9-induced Plcb4 mutation (p = 0.01).
| Abnormal blood clotting in veins (venous thrombosis) or arteries (arterial thrombosis) are major health problems, with venous thrombosis affecting approximately 1 in every thousand individuals annually in the United States. Susceptibility to venous thrombosis is governed by both genes and environment, with approximately 60% of the risk attributed to genetic influences. Though several genetic risk factors are known, >50% of genetic risk remains unexplained. Approximately 5% of people carry the most common known risk factor, Factor V Leiden. However, only 10% of these individuals will develop a blood clot in their lifetime. Mice carrying two copies of the Factor V Leiden mutation together with a mutation in a second gene called tissue factor pathway inhibitor develop fatal thrombosis shortly after birth. To identify genes that prevent this fatal thrombosis, we studied a large panel of mice carrying inactivating gene changes randomly distributed throughout the genome. We identified several genes as potential candidates to alter blood clotting balance in mice and humans with predisposition to thrombosis, and confirmed this protective function for DNA changes in one of these genes (Plcb4).
| Venous thromboembolism (VTE) affects 1:1000 individuals in the US each year and is highly heritable [1, 2]. A single nucleotide variant (SNV) in the F5 gene, referred to as Factor V Leiden (FVL, p.R506G) is present in 5–10% of Europeans, conferring a 2–4 fold increased risk for VTE [3]. Although ~25% of VTE patients carry the FVL variant [4], only ~10% of individuals heterozygous for FVL develop thrombosis in their lifetime.
To identify genetic variants that could potentially function as modifiers for FVL-associated VTE risk, we recently reported a dominant ENU screen [5] in mice sensitized for thrombosis. Such sensitized screens have been previously successful in identifying modifier genes for various phenotypes [6–9]. This screen was based on mice homozygous for the FVL mutation (F5L/L) and haploinsufficient for tissue factor pathway inhibitor (Tfpi+/-). As previously reported, F5L/L Tfpi+/- mice exhibit normal embryonic development, although nearly all die of widespread, systemic thrombosis in the immediate perinatal period [10]. After ENU mutagenesis, 98 G1 F5L/L Tfpi+/- progeny survived to weaning (“rescues”) and 16 progeny exhibited successful transmission of the ENU-induced suppressor mutation. However, subsequent efforts to genetically map the corresponding suppressor loci were confounded by complex strain-specific differences introduced by the required genetic outcross [5]. Similar genetic background effects have complicated previous mapping efforts [11] and have been noted to significantly alter other phenotypes [12, 13]. Additional challenges of this mapping approach include the requirement for large pedigrees and limited mapping resolution, with candidate intervals typically harboring tens to hundreds of genes and multiple closely linked mutations.
High throughput sequencing methods have enabled the direct identification of ENU-induced mutations. Thus, mutation identification in ENU screens is no longer dependent upon an outcross strategy for gene mapping [14–16]. We now report whole exome sequencing (WES) of 107 rescue mice (including 50 mice from the previously reported ENU screen [5]). Assuming loss of gene function as the mechanism of rescue, these WES data were analyzed gene-by-gene to identify genes enriched with mutations (mutation burden analysis). The Arl6ip5 gene emerged as the top candidate suppressor locus from this analysis. However, an independent CRISPR/Cas9-generated Arl5ip5 mutant allele failed to demonstrate highly penetrant rescue of the F5L/L Tfpi+/- lethal phenotype. Surprisingly, a maternally inherited (not ENU-induced) de novo mutation (Plcb4R335Q) exhibited significant co-segregation with the rescue phenotype (p = 0.003) in an expanded pedigree.
In the previously reported ENU screen [5], viable F5L/L Tfpi+/- rescue mice were outcrossed to the 129S1/SvImJ strain to introduce the genetic diversity required for subsequent mapping experiments. However, complex strain modifier gene interactions confounded this analysis and resulted in a large number of “phenocopies” (defined as viable F5L/L Tfpi+/- mice lacking the original rescue mutation). To eliminate confounding effects of these thrombosis strain modifiers, we generated an additional 2,834 G1 offspring exclusively maintained on the C57BL/6J background. The 42 G1 F5L/L Tfpi+/- mice alive at 6 weeks of age were mated to F5L/L mice to test the heritability of the survival phenotype. Twenty-one of these 42 mice generated at least one litter and 15 (1 female, 14 males) produced ≥1 offspring with the F5L/L Tfpi+/- genotype. Fifteen new rescue pedigrees were established from this screen (S1 Table). All pedigrees were expanded until all rescues either were infertile or died without producing any progeny with rescue genotype. The frequency, survival, weight, and sex distributions of identified rescues were consistent with our previous report (S1 Fig). Though many of the pedigrees previously generated on the mixed 129S1/SvImJ-C57BL/6J background generated >45 rescue progeny per pedigree (8/16) [5], all pedigrees on the pure C57BL/6J background yielded <36 rescue mice (most generating ≤5 rescues) (S1 Table). Significantly smaller pedigrees in comparison to the previous screen (p = 0.010, S2 Fig) are likely explained by a generally positive effect of the hybrid 129S1/SvImJ-C57BL/6J strain background either directly on rescue fertility (hybrid vigor) or indirectly by reducing the severity of the F5L/L phenotype. Although a contribution from nongenetic factors cannot be excluded, the C57BL/6J and 129S1/SvImJ strains have been shown to exhibit significant differences in a number of hemostasis-related parameters, including platelet count and TFPI and tissue factor expression levels [17], with the genetic variations underlying such strain specific differences likely contributing to the genetic mapping complexity noted in the previous report [5].
As the rescue pedigrees were maintained on a pure C57BL/6J background, the only genetic markers that could be used for mapping were ENU-induced variants. WES of one G1 or G2 member of the three largest pedigrees (1, 6, and 13, S2 Table), identified a total of 86 candidate ENU variants that were also validated by Sanger sequencing analysis (S3 Table). Of these 86 candidate genes, 69 were present in the G1 rescue but not its parents (G0), indicating that they were likely ENU-induced variants. These 69 variants were then further genotyped in all other rescue progeny in the respective pedigrees. Given the low number of identified genetic markers (20–26 per pedigree), these three pedigrees were poorly powered (29.6%, 21.7% and 39.4%, respectively) to identify the rescue variants by linkage analysis (S3A–S5A Figs). None of the 19 ENU variants tested in pedigree 1 (S3B Fig), showed linkage with a LOD-score >1.5 (S3C Fig). Similarly, 26 and 24 variants analyzed in pedigrees 6 and 13, respectively (S4B and S5B Figs) also failed to demonstrate a LOD-score >1.5 (S4C and S5C Figs). Failure to map the causal loci in any of these pedigrees was likely due to insufficient marker coverage. However, in these analyses, we could not exclude the contribution from a non-ENU-induced variant [18] or an unexpectedly high phenocopy rate. While WES has been successfully applied to identify causal ENU variants within inbred lines [19] and in mixed background lines [20, 21], whole genome sequencing (WGS) provides much denser and more even coverage of the entire genome (~3,000 ENU variants/genome expected) and outperforms WES for mapping [15]. However, a WGS approach requires sequencing multiple pedigree members [16], or pooled samples at high coverage [15], resulting in considerably higher expense with current methods.
In order to identify exonic ENU mutations, a total of 107 G1 rescues (57 from the current ENU screen and an additional 50 rescues with available material from the previous screen [5]), were subjected to WES (S2 Table). From ~1.5 million initially called variants (34,000 in exonic regions) 6,735 SNVs and 36 insertions-deletions (INDELs) within exonic regions were identified as potential ENU-induced mutations, using an in-house filtering pipeline (see Materials and methods). The most common exonic variants were nonsynonymous SNVs (47%), followed by mutations in 3’ and 5’ untranslated regions (31%) and synonymous SNVs (15%). The remaining variants (7%) were classified as splice site altering, stoploss, stopgain, or INDELs (Fig 1A). T/A -> C/G (47%), and T/A -> A/T (24%) SNVs were overrepresented, while C/G -> G/C (0.8%) changes were greatly underrepresented (Fig 1B), consistent with previously reported ENU studies [22, 23]. Since ENU is administered to the G0 father of G1 rescues, only female progeny are expected to carry induced mutations on the X chromosome, while males inherit their single X chromosome from the unmutagenized mother. Among the called variants, all chromosomes harbored a similar number of mutations in both sexes, with the exception of the X chromosome where a >35 fold increase in SNVs per mouse was observed in females (Fig 1C). The average number of exonic ENU mutations for G1 rescues was ~65 SNVs per mouse (Fig 1D), consistent with expected ENU mutation rates [16, 23]. These data suggest that most called variants are likely to be of ENU origin.
WES data for 107 independent rescue mice were jointly analyzed to identify candidate genes that are enriched for potentially deleterious ENU-induced variants including missense, nonsense, frameshift, and splice site altering mutations (3,481 out of 6,771 variants in 2,984 genes, ~32.5 potentially deleterious variants per mouse, S4 Table). Similar mutation burden analyses have been used to identify genes underlying rare diseases caused by de novo loss-of-function variants in humans [24–27]. In our study, the majority of genes harbored only a single ENU-induced variant, with 15 SNVs identified in Ttn, the largest gene in the mouse genome (Fig 1E). After adjusting for coding region size and multiple testing (for 2,984 genes), the ENU-induced mutation burden of potentially deleterious variants was significantly greater than expected by chance for 3 genes (FDR<0.1, Arl6ip5, Itgb6, C6) and suggestive for 9 additional genes (FDR<0.25). Sanger sequencing validated 36 of the 37 variants in these 12 candidate genes (S4 Table). While in this study, stringent correction for multiple testing suggested no significant enrichment (Arl6ip5 FDR = 0.68, Fig 2), the potential power of this burden analysis is highly dependent on the number of possible genes that could result in a viable rescue. If there were 30 such genes in the genome and every one of the 107 rescue mice carried a mutation in one of these 30 genes, each gene would be, on average, represented by ~3.5 mutations (107/30), with >7 genes expected to carry 5 or more mutations, which should have been sufficient to distinguish from the background mutation rate. However, if 500 genes could rescue the phenotype, sequencing close to a thousand mice would be required to achieve sufficient mapping power. The power could be further compromised by modifier genes with incomplete penetrance, imperfect predictions for potentially harmful mutations, and by the previously reported background survival rate for the rescue mice [10]. Due to the uncertainty of the power of these analyses, we proceeded to experimentally test the thrombosupressive effects of loss of function mutations in the genes identified by mutation burden analysis.
Independent null alleles were generated with CRISPR/Cas9 for the top candidate genes (Arl6ip5, C6, Itgb6, Cpn1, Sntg1 and Ces3b; Fig 2) to test for thrombosuppression. From 294 microinjected zygotes with pooled guide RNAs targeting these 6 genes, we obtained 39 progeny. CRISPR/Cas9 genome editing was assessed by Sanger sequencing of the sgRNA target sites. Approximately 190 independent targeting events were observed across the 6 genes in 36 of the 39 mice including small INDELs, single nucleotide changes, and several large (>30bp) deletions or inversions. Targeted alleles were either homozygous, heterozygous, or mosaic, with the number of editing events varying greatly for different sgRNAs (2.5–85%). Two or more different CRISPR/Cas9-induced alleles for each of the candidate genes (S5 Table) were bred to isolation but maintained on the F5L background for subsequent test crossing. The progeny of F5L/L Tfpi+/+ mice crossed with F5L/+ Tfpi+/- mice (one of these parental mice also carrying the CRISPR/Cas9-induced allele) were monitored for survival of F5L/L Tfpi+/- offspring (Table 1, S6 Table).
Over 100 progeny were generated for each of the candidate genes with no obvious rescue effect. A slight increase in rescues carrying the F5L/L Tfpi+/- Arl6ip5+/- genotype was noted, although it remained non-significant after surveying 205 offspring (p = 0.21, Table 1).
Our sensitized suppressor screening strategy is highly dependent on the underlying thrombosis model. Modifier genes rescuing the F5L/L Tfpi+/- synthetic lethal phenotype are potentially relevant to the common human FVL variant and our previous observations that mutations in F8 and F3 can rescue F5L/L Tfpi+/- demonstrate the sensitivity of this model to genetic alterations in coagulation system balance. However, rescue of F5L/L Tfpi+/- lethality by haploinsufficiency for F3 (the target of TFPI) only exhibits penetrance of ~33% [5], a level of rescue which current observations cannot exclude for Arl6ip5 and Sntg1. For most of the other candidate genes, the number of observed F5L/L Tfpi+/- mice did not differ from the expected background survival rate for this genotype (~2%) [10]. Though higher numbers of rescues were observed for offspring from the Sntg1 cross, these were equally distributed between mice with and without the Sntg1 loss-of-function allele.
The number of G1 rescues produced from each ENU-treated G0 male is shown in Fig 3A. Though most of the 182 G0 males yielded few or no G1 rescue offspring, a single G0 produced 6 rescues out of a total of 39 offspring (Fig 3A), including the founder G1 rescue for the largest pedigree (number 13). This observation suggested a potential shared rescue variant rather than 6 independent rescue mutations from the same G0 founder. Similarly, another previously reported ENU screen identified 7 independent ENU pedigrees with an identical phenotype mapping to the same genetic locus, also hypothesized to result from a single shared mutation [11]. While rescue siblings could theoretically originate from the same mutagenized spermatogonial stem cell and share ~50% of their induced mutations [28], such a common stem cell origin was excluded by exome sequence analysis in the rescue G1 sibs identified here (see Materials and methods).
Analysis of WES for 3 of the G1 rescues originating from this common G0 founder male (Fig 3B, S2 Table) identified 3 protein-altering variants (Plcb4R335Q, Pyhin1G157T, and Fignl2G82S) shared among 2 or more of the 6 G1 rescues (S7 Table). Plcb4R335Q was detected as a de novo mutation in one of the non-mutagenized G0 females in phase with the Tfpi null allele (Fig 3B) and was present in 3 out of 6 G1 rescue siblings. Plcb4 is located approximately 50 megabases upstream of the Tfpi locus on chromosome 2 (predicted recombination between Plcb4 and Tfpi ~14.1%) (Fig 3C) [29, 30]. While non-rescue littermates exhibited the expected rate of recombination between the Plcb4R335Q and Tfpi loci (20.2%), all 43 rescue mice (3 G1s and their 40 ≥G2 progeny) were non-recombinant and carried the Plcb4R335Q variant. This co-segregation between the Plcb4R335Q variant and the rescue phenotype is statistically significant (p = 0.003; Fig 3C). Plcb4R335Q lies within a highly conserved region of Plcb4 (Fig 3D) and is predicted to be deleterious by Polyphen-2 [31]. The other identified non-ENU variants (Pyhin1G157T and Fignl2G82S) did not segregate with the rescue phenotype (S6 Fig).
Although the estimated de novo mutation rate for inbred mice (~5.4 x 10−9 bp/generation) is 200X lower than our ENU mutation rate, other de novo variants have coincidentally been identified in ENU screens [32]. Of note, the Plcb4R335Q variant was originally removed from the candidate list by a filtering step based on the assumption that each ENU-induced mutation should be unique to a single G1 offspring. Although this algorithm was very efficient for removing false positive variants in our screen and others [21], our findings illustrate the risk for potential false negative results that this approach confers.
An independent Plcb4 null allele was generated by CRISPR/Cas9. Three distinct INDELs were identified by Sanger sequencing in the 25 progeny obtained from the CRISPR/Cas9-injected oocytes. One of these alleles introduced a single nucleotide insertion at amino acid 328, resulting in a frameshift in the protein coding sequence and a marked decrease in the steady state mRNA level from the mutant allele (~2% compared to wildtype), consistent with nonsense-mediated decay (Plcb4ins1, Fig 4A and 4B). A total of 169 progeny from a F5L/L Plcb4+/ins1 X F5L/+ Tfpi+/- cross yielded 11 F5L/L Tfpi+/- rescue progeny surviving to weaning (Fig 4C, S8 Table). Ten of these 11 rescues carried the Plcb4ins1 allele, consistent with significant rescue (p = 0.01, Fig 4C) with reduced penetrance (~40%). Plcb4 encodes phospholipase C, beta 4 and has been recently associated with auriculocondylar syndrome in humans [33]. No role for PLCB4 in the regulation of hemostasis has been previously reported, and the underlying mechanism for suppression of the lethal F5L/L Tfpi+/- phenotype is unknown.
The above rescue of the F5L/L Tfpi+/- phenotype by an independent Plcb4 mutant allele, strongly supports the identification of the de novo Plcb4R355Q mutation as the causal suppressor variant for Pedigree 13. These findings are also most consistent with a loss-of-function mechanism of action for the Plcb4R355Q mutation. The lack of a positive signal from this genomic region by the linkage analysis described above (S5 Fig) is likely explained by the absence of a nearby genetically informative ENU variant (the closest, Abca2 is located >50 Mb downstream from both Tfpi and Plcb4 (S3 Table, S5 Fig)). Of note, 4 of the 107 rescue mice in the WES mutation burden analysis also carried a Plcb4 mutation consistent with its suppressor function, though below the level of statistical significance. Nonetheless, these findings highlight the feasibility of our approach, given sufficient power.
In conclusion, we performed a dominant, sensitized ENU mutagenesis screen for modifiers of thrombosis. Analysis of extended pedigrees identified Plcb4 as a novel thrombosis modifier. Though mutation burden analysis suggested several other potential modifier loci, including Arl6ip5, incomplete penetrance and the background phenocopy rate significantly limited the power to detect additional thrombosis suppressor genes. Future applications of this approach will likely require significantly larger sample sizes and/or a more stringent sensitized genotype for screening. Nonetheless, our findings demonstrate the power of a sensitized ENU screen and mutation burden analysis to identify novel loci contributing to the regulation of hemostatic balance and candidate modifier genes for thrombosis and bleeding risk in humans.
Mice carrying the murine homolog of the FVL mutation [34] (F5L; also available from Jackson Laboratories stock #004080) or the TFPI Kunitz domain deletion (Tfpi-) [35] were genotyped using PCR assays with primers and conditions as previously described [34, 35], and maintained on the C57BL/6J background (Jackson Laboratories stock #000664). All animal care and procedures were performed in accordance with the Principles of Laboratory and Animal Care established by the National Society for Medical Research. The Institutional Animal Care and Use Committee at the University of Michigan has approved protocols PRO00005191 and PRO00007879 used for the current study and conforms to the standards of “The Guide for the Care and Use of Laboratory Animals” (Revised 2011).
ENU mutagenesis was performed as previously described [5], with all mice on the C57BL/6J genetic background. Briefly, 189 F5L/L male mice (6–8 weeks old) were administered three weekly intraperitoneal injections of 90 mg/kg of ENU (N-ethyl-N-nitrosourea, Sigma-Aldrich). Eight weeks later, 182 surviving males were mated to F5L/+ Tfpi+/- females and their G1 progeny were genotyped at age 2–3 weeks to identify viable F5L/L Tfpi+/- offspring (‘rescues’). F5L/L Tfpi+/- G1 rescues were crossed to F5L/L mice on the C57BL/6J genetic background (backcrossed >20 generations) and transmission was considered positive with the presence of one or more rescue progeny. Theoretical mapping power in rescue pedigrees was estimated by 10,000 simulations using SIMLINK software [36].
Gender, age, WES details, and other characteristics for 108 rescue mice are provided in S2 Table. Genomic DNA (gDNA) extracted from tail biopsies of 56 G1 offspring from the current ENU screen and from an additional 50 F5L/L Tfpi+/- mice on the C57BL/6J background from the previous screen [5] were subjected to WES at the Northwest Genomics Center, University of Washington. Sequencing libraries were prepared using the Roche NimbleGen exome capture system. DNA from an additional two rescue offspring was subjected to WES at Beijing Genomics Institute or Centrillion Genomics Technologies, respectively (S2 Table). These two libraries were prepared using the Agilent SureSelect capture system. 100 bp paired-end sequencing was performed for all 108 exome libraries using Illumina HiSeq 2000 or 4000 sequencing instruments. Two WES mice represented rescue pedigree 1: the G1 founder and a G2 rescue offspring. The latter was used for linkage analysis, but excluded from the burden analysis (S2 Table).
Average sequencing coverage, estimated by QualiMap software [37], was 77X, and >96% of the captured area was covered by at least 6 independent reads (S2 Table). All generated fastq files have been deposited to the NCBI Sequence Read Archive (Project accession number #PRJNA397141). A detailed description of variant calling as well as in-house developed scripts for variant filtration are online as a GitHub repository (github.com/tombergk/FVL_mod). In short, Burrows-Wheeler Aligner [38] was used to align reads to the Mus Musculus GRCm38 reference genome, Picard [39] to remove duplicates, and GATK [40] to call and filter the variants. Annovar software [41] was applied to annotate the variants using the Refseq database. All variants within our mouse cohort present in more than one rescue were declared non-ENU induced and therefore removed. Unique heterozygous variants with a minimum of 6X coverage were considered as potential ENU mutations. Among 107 whole exome sequenced G1 mice, 38 were siblings (13 sib-pairs and 4 trios, S2 Table). 190 heterozygous variants present in 2 or 3 mice (representing sibpairs or trios) out of 107 rescues were examined, with 15 found to be shared by siblings (S7 Table). Of the 7 sibs/trios sharing an otherwise novel variant, none shared >10% of their identified variants–inconsistent with the expected 50% for progeny originating from the same ENU-treated spermatogonial stem cell.
All ENU-induced variants predicted to be potentially harmful within protein coding sequences including missense, nonsense, splice site altering SNVs, and out-of-frame insertions-deletions (INDELs), were summed for every gene. The number of potentially damaging variants per gene was compared to a probability distribution of each gene being targeted by chance. Probability distributions were obtained by running 10 million random permutations using probabilities adjusted to the length of the protein coding region. A detailed pipeline for the permutation analysis is available online (github.com/tombergk/FVL_mod). Genes that harbored more potentially damaging ENU-induced variants than expected by chance were considered as candidate modifier genes. FDR statistical correction for multiple testing was applied as previously described [42].
All coding variants in pedigrees 1, 6, and 13 as well as all variants in candidate modifier genes from the burden analysis were assessed using Sanger sequencing. Variants were considered ENU-induced if identified in the G1 rescue but not its parents. All primers were designed using Primer3 software [43] and purchased from Integrated DNA Technologies. PCR was performed using GoTaq Green PCR Master Mix (Promega), visualized on 2% agarose gel, and purified using QIAquick Gel Extraction Kit (Qiagen). Sanger sequencing of purified PCR products was performed by the University of Michigan Sequencing Core. Outer primers were used to generate the PCR product which was then sequenced using the internal sequencing primers. All outer PCR primers (named: gene name+’_OF/OR’) and internal sequencing primers (named: gene name+’_IF/IR’) are listed in S9 Table.
Guide RNA target sequences were designed with computational tools [44, 45] (http://www.broadinstitute.org/rnai/public/analysis-tools/sgrna-design or http://genome-engineering.org) and top predictions per each candidate gene were selected for functional testing (S10 Table). Single guide RNAs (sgRNA) for C6, Ces3b, Itgb6, and Sntg1 were in vitro synthesized (MAXIscript T7 Kit, Thermo Fisher) from double stranded DNA templates by GeneArt gene synthesis service (Thermo Fisher) while the 4 sgRNAs for Arl6ip5 were in vitro synthesized using the Guide-it sgRNA In Vitro Transcription Kit (Clontech). The sgRNAs were purified prior to activity testing (MEGAclear Transcription Clean-Up Kit, Thermo Fisher). Both the Wash and Elution Solutions of the MEGAclear Kit were pre-filtered with 0.02 μm size exclusion membrane filters (Anotop syringe filters, Whatman) to remove particulates from zygote microinjection solutions, thus preventing microinjection needle blockages.
Target DNA for the in vitro cleavage assays was PCR amplified from genomic DNA isolated from JM8.A3 C57BL/6N mouse embryonic stem (ES) cells [46] with candidate gene specific primers (S10 Table). In vitro digestion of target DNA was carried out by complexes of synthetic sgRNA and S. pyogenes Cas9 Nuclease (New England BioLabs) according to manufacturer's recommendations. Agarose gel electrophoresis of the reaction products was used to identify sgRNA molecules that mediated template cleavage by Cas9 protein (S7 Fig). Arl6ip5 was assayed separately, with one out-of-four tested sgRNAs successfully cleaving the PCR template.
Synthetic sgRNAs that targeted Cpn1 were not identified by the in vitro Cas9 DNA cleavage assay. As an alternative assay, sgRNA target sequences (Cpn1-g1, Cpn1-g2) were cloned into plasmid pX330-U6-Chimeric_BB-CBh-hSpCas9 (Addgene.org Plasmid #42230) [47] and co-electroporated into JM8.A3 ES cells as previously described [48]. Briefly, 15 μg of a Cas9 plasmid and 5 μg of a PGK1-puro expression plasmid [49] were co-electroporated into 0.8x107 ES cells. On days two and three after electroporation media containing 2 μg/ml puromycin was applied to the cells; then selection free media was applied for four days. Genomic DNA was purified from surviving ES cells. The Cpn1 region targeted by the sgRNA was PCR amplified and tested for the presence of indel formation with a T7 endonuclease I assay according to the manufacturer’s instructions (New England Biolabs).
CRISPR/Cas9 gene edited mice were generated in collaboration with the University of Michigan Transgenic Animal Model Core. A premixed solution containing 2.5 ng/μl of each sgRNA for Arl6ip5, C6, Ces3b, Itgb6, Sntg1, and 5 ng/μl of Cas9 mRNA (GeneArt CRISPR Nuclease mRNA, Thermo Fisher) was prepared in RNAse free microinjection buffer (10 mM Tris-Hcl, pH 7.4, 0.25 mM EDTA). The mixture also included 2.5 ng/μl of pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid containing guide Cpn1-g1 and a 2.5 ng/μl of pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid containing guide Cpn1-g2 targeting Cpn1 (S10 Table). The mixture of sgRNAs, Cas9 mRNA, and plasmids was microinjected into the male pronucleus of fertilized mouse eggs obtained from the mating of stud males carrying the F5L/+ Tfpi+/- genotype on the C57BL/6J background with superovulated C57BL/6J female mice. Microinjected eggs were transferred to pseudopregnant B6DF1 female mice (Jackson Laboratories stock #100006). DNA extracted from tail biopsies of offspring was genotyped for the presence of gene editing. The Plcb4 allele was targeted in a separate experiment in collaboration with the University of Michigan Transgenic Animal Model Core using a pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid that contained guide Plcb4 (5 ng/μl).
Initially, sgRNA targeted loci were tested using PCR and Sanger sequencing (primer sequences provided in S10 Table). Small INDELs were deconvoluted from Sanger sequencing reads using TIDE software [50]. A selection of null alleles from >190 editing events were maintained for validation (S5 Table). Large (>30 bp) deletions were genotyped using PCR reactions that resulted in two visibly distinct PCR product sizes for the deletion and wildtype alleles. Expected product sizes and genotyping primers for each deletion are listed in S5 Table. All genotyping strategies were initially validated using Sanger sequencing.
A qPCR approach was applied to exclude large on-target CRISPR/Cas9-induced deletions. All DNA samples were quantified using the Quant-iT™ PicoGreen® dsDNA Assay Kit (Life Technologies) and analyzed on the Molecular Devices SpectraMax® M3 multi-mode microplate reader using SoftMax Pro software and diluted to 5ng/μl. Primer pairs were designed for each gene using Primer Express 3.0 software (S9 Table) and samples were measured in triplicate using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) on a 7900 HT Fast Real-Time PCR System (Applied Biosystems) with DNA from wildtype C57BL/6J mice used as a reference. While large CRISPR/Cas9 induced deletions extending the borders of the PCR primers have been reported [51, 52], qPCR did not detect evidence for a large deletion in any of the CRISPR targeted genes.
The ratio of WT to Plcb4+ins1 mRNA levels was determined as previously described [18]. In short, a whole brain tissue sample was snap frozen in liquid nitrogen from a Plcb4+/ins1 mouse. Total RNA was extracted using an RNA extraction kit (Nucleospin RNA from Macherey-Nagel) and 250 ng of total RNA was converted to complementary DNA (cDNA) using SuperScript IV One-Step RT-PCR (Invitrogen) following the manufacturer’s instructions. Genotyping primers spanning the nearest intron (primers Plcb4_cDNA_OF1 and Plcb4_cDNA_OR, S9 Table) were used to amplify a segment of Plcb4 containing the +1 insertion from the cDNA samples. PCR products were extracted from agarose gels using a QIAquick Gel Purification Kit (Qiagen) and submitted for Sanger sequencing (primer Plcb4_cDNA_OF2, S9 Table). The differential allelic expression was estimated from the ratio between the wildtype and Plcb4ins1 sequence peak areas in the cDNA sample compared to gDNA using Phred software [53]. This ratio was calculated for 10 consecutive positions within the PCR product where the wildtype and Plcb4ins1 alleles contain a different nucleotide.
Kaplan-Meier survival curves and a log-rank test to estimate significant differences in mouse survival were performed using the ‘survival’ package in R [54]. A paired two-tailed Student’s t-test was applied to estimate differences in weights between rescue mice and their littermates. Fisher’s exact tests were applied to estimate deviations from expected proportions in mouse crosses. Mendelian segregation for CRISPR/Cas9-induced alleles among non-rescue littermates was assessed in a subset of mice by Sanger sequencing and then assumed for the rest of the littermates in the Fisher’s exact tests. Benjamini and Hochberg FDR for ENU burden analysis, Student’s t-tests, and Fisher’s exact tests were all performed using the ‘stats’ package in R software [55]. Linkage analysis was performed on the Mendel platform version 14.0 [56] and LOD scores ≥3.3 were considered genome-wide significant [57].
|
10.1371/journal.ppat.1004149 | Dusp3 and Psme3 Are Associated with Murine Susceptibility to Staphylococcus aureus Infection and Human Sepsis | Using A/J mice, which are susceptible to Staphylococcus aureus, we sought to identify genetic determinants of susceptibility to S. aureus, and evaluate their function with regard to S. aureus infection. One QTL region on chromosome 11 containing 422 genes was found to be significantly associated with susceptibility to S. aureus infection. Of these 422 genes, whole genome transcription profiling identified five genes (Dcaf7, Dusp3, Fam134c, Psme3, and Slc4a1) that were significantly differentially expressed in a) S. aureus –infected susceptible (A/J) vs. resistant (C57BL/6J) mice and b) humans with S. aureus blood stream infection vs. healthy subjects. Three of these genes (Dcaf7, Dusp3, and Psme3) were down-regulated in susceptible vs. resistant mice at both pre- and post-infection time points by qPCR. siRNA-mediated knockdown of Dusp3 and Psme3 induced significant increases of cytokine production in S. aureus-challenged RAW264.7 macrophages and bone marrow derived macrophages (BMDMs) through enhancing NF-κB signaling activity. Similar increases in cytokine production and NF-κB activity were also seen in BMDMs from CSS11 (C57BL/6J background with chromosome 11 from A/J), but not C57BL/6J. These findings suggest that Dusp3 and Psme3 contribute to S. aureus infection susceptibility in A/J mice and play a role in human S. aureus infection.
| Staphylococcus aureus causes life-threatening infections in humans. Host genetic determinants influence the outcome of S. aureus infection, yet are poorly understood. Susceptible A/J and resistant C57BL/6J mice provide a unique platform to study the genetic difference responsible for variable host response to S. aureus infection. We showed that chromosome 11 in A/J was responsible for susceptibility to S. aureus. We further identified a QTL locus on Chromosome 11 significantly associated with S. aureus susceptibility. Five genes in the QTL (Dcaf7, Dusp3, Fam134c, Psme3, and Slc4a1) were significantly differently expressed in a) susceptible vs. resistant mice, and b) humans with S. aureus blood stream infection vs. healthy human subjects. Three genes (Dusp3, Psme3, and Dcaf7) were down-regulated in susceptible A/J mice. siRNA-mediated knockdown of Dusp3 and Psme3 in bone marrow derived macrophage (BMDMs) significantly enhanced cytokine responses through NF-κB activity upon S. aureus challenge in a pattern that was also present in S. aureus-challenged BMDMs from susceptible CSS11 (chr. 11 from A/J but otherwise C57BL/6J) mice, but not resistant C57BL/6J mice. These findings suggest that Dusp3 and Psme3 contribute to S. aureus infection susceptibility in A/J mice and play a role in human S. aureus infection.
| Staphylococcus aureus is an important cause of potentially lethal human infections [1]–[3]. It is generally accepted that host genetic variation influences susceptibility to S. aureus colonization and infection [4], [5].
A significant body of evidence supports the importance of human genetic variation on host susceptibility to a variety of infectious diseases. For example, TNF gene SNP rs1800629 is strongly associated with susceptibility to severe sepsis in the Chinese Han population [6], while genetic variants in TRAF6 are significantly associated with susceptibility to sepsis-induced acute lung injury [7]. In addition, a genetic variant of β2-adrenocepter gene increases susceptibility to bacterial meningitis [8], while genetic variations in Toll-like receptors have been linked with both infectious and autoimmune diseases [9]. More interestingly, genetic variation of IL17A gene is associated with altered susceptibility to Gram-positive infection and mortality of severe sepsis [10].
Far less is known about the specific genes associated with host susceptibility to S. aureus. Our group [11], [12] and others [5], [13] have shown that different inbred mouse strains exhibit variable susceptibility to S. aureus infection. For example, A/J is highly susceptible to S. aureus infection, whereas C57BL/6J is resistant [5]. These susceptible and resistant strains provide an attractive approach to investigate the host genetic determinants of susceptibility to S. aureus infection.
Using A/J donor to C57BL/6J host chromosomal substitution strains (CSS) we recently discovered that chromosomes 8, 11, and 18 from A/J account for its high susceptibility to S. aureus infection [11]. However, the genes on chromosome 11 that influence susceptibility to S. aureus remain unknown.
In the present investigation, we used a multi-step selection process to identify genes on A/J chromosome 11 contributing to susceptibility to S. aureus infection. Because human and murine response to sepsis can differ significantly [14], we used whole blood gene expression data from a cohort of patients with S. aureus blood stream infection (BSI) to verify the potential biological relevance of all candidate genes identified in mice. Genes shown to be involved in host response to S. aureus in both mice and humans were evaluated for biological function. Using this cross-species validation approach, we identified Dusp3 and Psme3 as relevant in both human and murine inflammatory response to S. aureus infection, and demonstrated these genes were involved in NF-κB signaling.
In the peritoneal S. aureus sepsis experiment, C57BL/6J, A/J and Chromosomal Substitution Strain 11 (CSS11) (A/J chromosome 11 in C57BL6/J background) mice were infected with S. aureus by an intraperitoneal (IP) route. Survival was observed for five days. C57BL/6J mice were resistant to S. aureus sepsis (median survival >5 days), while CSS11 mice demonstrated a susceptible phenotype (median survival <2 days) (Figure 1A) (p<0.05). In the intravenous S. aureus sepsis experiment, C57BL/6J, A/J and CSS11 mice were infected by direct inoculation of S. aureus by tail-vein route. Survival was observed every 6 hours, both A/J and CSS11 were susceptible to S. aureus infection (median survival <24 hours, p<0.05) as compared with C57BL/6J (Figure 1B).
The kidney bacterial load in CSS11 was also significantly higher after S. aureus challenge as compared with C57BL/6J (2.0±1.32×106 CFU/ml vs 200±158 CFU/ml, p<0.0001) (Figure 1C), indicating the higher mortality of CSS11 strain was associated with higher S. aureus burden in tissue.
C57BL/6J, A/J, and CSS11 mice were also infected with Escherichia coli by IP injection. Survival was observed every 6 hours. Both A/J and CSS11 mice were susceptible to E. coli sepsis as compared with C57BL/6J (median survival <2 days; p<0.05) (Figure S1).
To localize the determinants on A/J chromosome 11 that are responsible for susceptibility to S. aureus infection, QTL analysis was performed. Previously, we established that the traits conferring susceptibility to S. aureus on A/J chromosome 11 were transmitted in an autosomal recessive fashion [11]. Thus, a total of 208 F2 intercross mice were generated by mating F1 (C11A) to F1 (C11A) and infected with methicillin susceptible S. aureus strain Sanger 476. Survival times were analyzed by J/qtl software (Jackson Labs). One significant QTL region containing 422 genes and located between 97006867 and 110713165 was significantly linked to susceptibility to S. aureus infection (Figure 1D).
Next, we employed the previous murine microarray expression data to further identify candidate genes located within the identified QTL region [11] (Figure 2). Genes within the identified QTL region that were differentially expressed between susceptible A/J and resistant C57BL/6J at all pre-infection and post-infection time points were considered to be putative candidate genes. A total of 11 genes met these criteria: Eif1, Cnp, Fam134c, Cntd1, Psme3, Dusp3, Mpp2, Slc4a1, Slc25a39, Dcaf7, and Gna13 (Figure 1D). The accession numbers for each gene was listed (Table 1).
To further evaluate the relevance of our candidate genes in human staphylococcal disease, we next used expression data from patients with S. aureus BSI (n = 32) or Escherichia coli BSI (n = 19) to evaluate the potential clinical relevance of our identified candidate genes. Among the 11 putative genes identified in the murine model, six were found to have human orthologues that exhibited significantly different levels of expression in patients with S. aureus BSI as compared to healthy subjects with no infection (n = 43): Dcaf7 (0.85fold; p = 0.003), Dusp3 (1.73 fold; p<0.0001), Fam134c (0.75fold; p<0.0001), Psme3 (0.78fold; p<0.0001), Mpp2 (1.21fold; p = 0.004), and Slc4a1 (0.81fold; p = 0.012) (Figure 3A). The gene expression patterns for the six genes were also significantly different among the patients with E. coli BSI vs. healthy subjects (Figure 3A).
Using qPCR on the murine samples, five of the 6 genes identified by both human and murine gene expression also exhibited consistent expression changes under infectious vs. non-infectious conditions : Dcaf7 (0.83fold), Dusp3 (1.31fold), Fam134c (0.65fold), Psme3 (0.86fold), and Slc4a1(0.85fold) (Figure 3B). The expression patterns for these five genes were highly consistent between S. aureus-infected mice and S. aureus-infected humans (Figure 3C). The expression patterns of these five genes also remained consistent when mice were infected with E. coli instead of S. aureus (Figure S2). qPCR primers for candidate genes were listed (Table S1).
Using qPCR in susceptible (A/J) vs resistant (C57BL/6J) mice at baseline (0 hr), three of the five genes demonstrated significantly lower expression: Dcaf7 (0.81 fold; p<0.05), Dusp3 (0.27 fold; p<0.01), and Psme3 (0.83 fold; p<0.05), while two genes exhibited significantly higher expression: Fam134c (1.82 fold; p<0.01), Slc4a1 (1.31fold; p<0.05) (Figure 4A). The baseline difference in expression between susceptible (A/J) and resistant (C57BL/6J) mice for all the five genes remained unchanged at 2 hr post-infection of S. aureus (Figure 4A).
Since NF-κB signaling plays a crucial rule in host defense to various pathogens [15]–[17] including S. aureus [18], [19], the impact of the five putative genes on NF-κB signaling was analyzed by co-transfection of NF-κB-luciferase reporter plasmid and siRNA of each gene into RAW264.7 murine macrophage cell line. The siRNAs used in this study were listed (Table S2). The siRNA of each candidate gene efficiently reduced gene expression in RAW 264.7 cells (Figure S3). Knockdown of two genes, Dusp3 (p<0.01) and Psme3 (p<0.01), significantly enhanced NF-κB signaling upon stimulation with either lipoteichoic acid (LTA) or S. aureus particles (Figure 4B). Because both of these genes (Dusp3 and Psme3) were down-regulated in A/J at baseline and S. aureus infection (Figure 4A), siRNA-mediated knockdown mimicked the status of susceptible A/J in RAW264.7 murine macrophages. These results suggest that lower expression of Dusp3 and Psme3 in A/J mice could explain the observed susceptibility to S. aureus.
To better understand how Dusp3 and Psme3 affect NF-κB signaling activity, phosphorylation of p65 at Ser536 and degradation of IκBα were analyzed by western blot. BMDMs from C57BL/6J were transfected with either scrambled, Dusp3, or Psme3 siRNA one day before and then subjected to S. aureus challenge for 15 minutes. The knock-down efficiency was tested in a parallel experiment (Figure S4). Western blot results showed that knockdown of Dusp3 or Psme3 dramatically increased phosphorylation of p65 at Ser536 as compared with scrambled siRNA (Figure 4C, top panel). The degradation of IκBα was also increased in either Dusp3 or Psme3 knockdown cells (Figure 4C, second bottom panel). The antibodies used in this study were listed (Table S3).
Similarly, BMDMs from CSS11 exhibited increased phosphorylation of p65 (Ser536) and degradation of IκBα after S. aureus stimulation as compared with BMDMs from C57BL/6J (Figure 4D). These data indicate that the suppressive function of Dusp3 and Psme3 on NF-κB signaling happens prior to the inhibitory cytosolic complex of IκBα-p65-p50. The signaling event should happen within the cytosol or at the cell membrane rather than in the nucleus, and probably in the proximal signaling stage before IκBα degradation.
Next, we analyzed cytokine and chemokine production after inhibiting NF-κB signaling. Either Bay 11-7085 or DMSO was applied to RAW264.7 cells before S. aureus challenge. qPCR illustrated that Bay inhibition of NF-κB signaling significantly suppressed inflammatory cytokine and chemokine production upon S. aureus stimulation in RAW264.7 macrophages, including IL-1β (p<0.01), IL-6 (p<0.05), and TNF-α (p<0.05) (Figure 4E). Inhibition of NF-κB signaling also enhanced the expression of Dusp3 (p<0.05) and Psme3 (p<0.05), suggesting a negative feedback regulatory loop between NF-κB and both Dusp3 and Psme3 (Figure 4F). qPCR primers for cytokines and chemokines applied in this study were listed (Table S4).
Neutrophils and macrophages from A/J and C57BL/6J exhibit similar bacterial killing capacity [11], suggesting that other host characteristics account for differences in the S. aureus susceptibility phenotype. To evaluate these factors, we used a well-established macrophage differentiation system [15] to differentiate the bone-marrow progenitor cells from A/J and C57BL/6J into macrophages and analyzed macrophage markers by flow cytometry. No significant differences in either CD11b or F4/80 expression were observed (Figure S5). Using flow cytometry, bone marrow derived macrophages (BMDMs) from A/J and C57BL/6J exhibited a similar phagocytosis capacity for S. aureus (Figure 5A).
Luminex cytokine profiling showed that BMDMs from CSS11 produced more inflammatory cytokines as compared with C57BL/6J in response to stimulation with S. aureus, including GM-CSF (p = 0.022), IL-1β (p = 0.03), IL-6 (p = 0.043), and TNF-α (p = 0.011) (Figure 5B, and S6). This elevation in cytokine production is likely to be due to the down-regulation of Dusp3 and Psme3 in CSS11. siRNA knockdown of these two genes in S. aureus-challenged RAW264.7 macrophages and BMDMs greatly enhanced NF-κB signaling (Figure 4B and 4C), which in turn directly activated cytokine and chemokine production (Figure 4E). Indeed, in Dusp3 siRNA transfected S. aureus-challenged RAW264.7 macrophages, Luminex-multiplex profiling detected significantly enhanced cytokine production as compared with scrambled siRNA, including GM-CSF (p = 0.02), IL-1β (p = 0.041), IL-6 (p = 0.04), and TNF-α (p = 0.032) (Figure 5C and S7). Similarly, Psme3 siRNA transfected, S. aureus-challenged RAW264.7 cells also produced significantly more GM-CSF (p = 0.041) and IL-6 (p = 0.01) than control (Figure 5C and S7). Similarly, in S. aureus challenged BMDMs of C57BL/6J, knockdown of Dusp3 dramatically enhanced the production of IL-6 (p = 0.015) and TNF-α (p = 0.018) (Figure 5D and S8). Likewise, knockdown of Psme3 enhanced the production of TNF-α (p = 0.026) as compared with scrambled siRNA (Figure 5D and S8).
Patterns of increased cytokine production were also consistent between mRNA production from Dusp3 or Psme3 siRNA transfected RAW 264.7 macrophages or BMDMs and primary CSS11 BMDMs. Knockdown of Dusp3 in RAW264.7 macrophages significantly increased cytokine mRNA production upon S. aureus challenge as compared with scrambled control, including GM-CSF (p = 0.013), IL-1β (p = 0.048), and TNF-α (p = 0.031) (Figure 6A and S9). Knockdown of Psme3 in RAW264.7 macrophages also significantly increased IL-6 (p = 0.032) mRNA production (Figure 6A and S9). Primary BMDMs from CSS11, which contains chr. 11 from A/J in an otherwise C57BL/6J background, expressed more cytokine mRNA upon S. aureus challenge as compared with C57BL/6J, including GM-CSF (p = 0.019), IL-1β (p = 0.039), IL-6 (p = 0.027), and TNF-α (p = 0.043). This observation is likely to reflect the down-regulation of both Dusp3 and Psme3 in the A/J chromosome 11 contained by CSS11 mouse lineage (Figure 6B and 4A). Similarly, in S. aureus challenged BMDMs of C57BL/6J, knockdown of Dusp3 dramatically increased IL-6 RNA (p = 0.0078) and TNF-α RNA (p = 0.048) (Figure 6C). Knockdown of Psme3 significantly increased TNF-α RNA (p = 0.046) as compared with scrambled siRNA (Figure 6C).
Since NF-κB signaling mediates host defenses generally in a positive way, we hypothesized that persistent, unbated stimulation of production of antimicrobial effectors from immune cells, such as antimicrobial peptides, would eventually lead to “immune paralysis” or “immune exhaustion” that impeded further defense against S. aureus challenge. To test this hypothesis, BMDMs from both C57BL/6J and A/J were pre-exposed to either TNF-α (100 ng/ml) or S. aureus particles (10 µg/ml), then subjected to phagocytosis analysis. Pre-exposure to TNF-α enhanced the phagocytosis capacity of BMDMs from both strains (Figure S10). However, pre-exposure to S. aureus particles reduced phagocytosis ability of BMDMs from both strains, and the reduction was more extensively in BMDMs from A/J (Figure S10). These results suggest that prolonged simultaneous activation of all pathways in the host by S. aureus, and not isolated stimulation of the TNFα-TNF receptor pathway alone, impaired the immune system's ability to further respond to S. aureus infection.
Next, we analyzed the role of Dusp3 and Psme3 upon S. aureus challenge in the immortalized human monocyte cell line U-937. Despite efficient knock-down of Dusp3 and Psme3, S. aureus stimulation of U-937 led to cell death rather than activation of NF-κB signaling and cytokine production (data not shown). We therefore concluded that U-937 was not suitable for our current analysis, and instead evaluated function of these two genes in human samples. To do this, we used publically available datasets of various human immune cells challenged by S. aureus, including human neutrophils (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE16837) and human macrophages (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE13670). In the human neutrophil dataset (GEO:GSE16837), Dusp3 increased to 9.88 fold at 3 hr (p<0.001) and 7.95 fold at 6 hr (p<0.05) as compared with 0 hr after S. aureus stimulation; and Psme3 decreased to 0.68 fold at 3 hr (p<0.0001) and 0.45 fold at 6 hr (p<0.0001) as compared with 0 hr (Figure S11). In the human macrophage dataset (GEO:GSE13670), Dusp3 increased to 1.62 fold at 8 hr following S. aureus challenge compared with controls (p<0.005) and Psme3 decreased to 0.73 fold at 8 hr following S. aureus challenge as compared with each control (p<0.001) (Figure S11). Collectively, these data and our human and murine data together strongly support that Dusp3 and Psme3 are candidate genes highly associated with S. aureus susceptibility.
The genetic basis for host susceptibility to S. aureus is largely unknown. In the current report, we have identified two genes, Dusp3 and Psme3, which are strongly associated with S. aureus sepsis in mice and humans, and have proposed a potential biological mechanism as negative feedback components in NF-κB-mediated signaling (Figure 7). These factors are likely to contribute to host susceptibility to S. aureus sepsis in mouse, and potentially in humans.
We identified a QTL locus on chromosome 11 that was significantly linked to S. aureus susceptibility in A/J mice. We found five candidate genes (Dcaf7, Dusp3, Fam134c, Psme3, and Slc4a1) within the QTL locus that were differentially expressed between susceptible and resistant mouse strains and had human orthologues with the same significant expression patterns in patients with both S. aureus BSI and E. coli BSI. The results suggest that one or more of these five genes are involved in a common host response to S. aureus infection in both humans and mice. qPCR demonstrated that two of these five genes, Dusp3 and Psme3, were down-regulated in susceptible vs. resistant mice, and exhibited a significantly enhanced inflammatory cytokine response when knocked down with siRNA in both RAW264.7 macrophages and BMDMs. The over-production of inflammatory cytokines encountered with siRNA-mediated knockdown of these two genes was consistent with that encountered in S. aureus-challenged bone marrow derived macrophages from CSS 11 (C57BL/6J background with A/J chromosome 11) mice. Taken together, these data provide strong evidence that Dusp3 and Psme3 may be key genes involved in the host susceptibility to S. aureus.
Previous investigations have shown inconsistencies between murine and human response to sepsis [20]–[23]. For example, a recent study by Seok and colleagues demonstrated significant differences between murine and human genomic responses to several acute inflammatory diseases, including burns, trauma, and endotoxemia [14]. Our candidate gene selection approach ensured that all murine candidate genes were also relevant in humans with S. aureus BSI. In this way, almost two-thirds of the more than 1000 genes on A/J chromosome 11 were excluded from contributing to susceptibility to S. aureus. Using this trans-species comparative genomic approach, only Dusp3 and Psme3 were identified as putative contributors to S. aureus susceptibility and both exhibited a significant biological effect on NF-κB signaling. Because NF-κB is centrally involved in the inflammatory response in both mouse and human [24], [25], these two genes are likely to be involved in the host inflammatory response to S. aureus in both humans and mice.
The diverse function of Dusp3 requires further detailed investigation on its biological relevance to S. aureus susceptibility. First, it is greatly involved in signaling transduction pathways regulating protein de-phosphorylation. For example, Dusp3 has been reported to be the main protein tyrosine phosphatase in macrophage mediating cellular processes, including immune response [26]. It is a redox-sensitive and ERK-specific phosphatase. Bacterial colonization results in oxidative inactivation of Dusp3 and consequently stimulation of ERK [27]. Dusp3 also regulates cell death and cell proliferation, exhibiting anti-apoptotic ability in prostate cancer cells and promoting cell cycle progression in carcinoma of the cervix [28], [29]. Perhaps most interestingly, Dusp3 has been shown to affect the expression of vascular endothelial growth factor (VEGF) [30]–[32]. VEGF has been reported to increase cardiovascular collapse and vascular permeability during S. aureus sepsis pathogenesis and has been proposed as a major determinant of vascular hyperpermeability in MRSA sepsis [30], [31]. In our experiment, VEGF was also significantly elevated in Dusp3 siRNA transfected RAW264.7 macrophages in both non-infection and S. aureus infection condition. Finally, Dusp3 homozygous mutant mice exhibited susceptibility to Gram negative bacterial infection in screening results from the Knockout Mouse Consortium program [33], [34]. Collectively, this evidence suggests the Dusp3 pathway may be an important candidate pathway in resolving S. aureus related pathogenesis.
Psme3 is a subunit of a proteasome responsible for the generation of peptides loaded onto MHC class I molecules [35], [36]. MHC class I molecules are crucial components of host response to infection, contributing to host recognition of viral and bacterial infected cells by host cytotoxic CD8-T cells for the final killing and degradation of pathogens [37], [38]. The down-regulation of Psme3 in A/J but not C57BL/6J may result in less degradation of phagocytosed S. aureus, less S. aureus antigen presented on MHC class I molecules, and ultimately less degradation of infected host cells by cytotoxic CD8-T cells. Interestingly, siRNA-mediated down-regulation of Psme3 also dramatically increased VEGF production by RAW264.7, further suggesting that VEGF enrichment affected by down-regulation of Dusp3 and Psme3 in murine chromosome 11 contributes to the S. aureus susceptible phenotype of A/J.
In the process of sepsis, harmful molecular mechanisms contribute to the high mortality seen in severely ill patients [39], [40]. In these cases, a pathogenic role of excessive immunity, also known as “cytokine storm” and reduced immunity through immune paralysis are highly associated with death induced by acute bacterial infection. Referred to as “compensatory anti-inflammatory response syndrome” (CARS), this “immune-exhaustion” phenomenon is the consequence of counter-regulatory mechanisms initiated to limit the over-activated inflammatory response in sepsis patients [41], [42]. In patients with CARS, over-activation of inflammatory response ultimately leads to changes in expression of genes associated with phagocytosis, antigen presentation, cell migration and apoptosis [43]. In the current investigation, downregulation of Dusp3 and Psme3 in A/J increased the production of pro-inflammatory cytokines, which may partially account for the cytokine storm observed in that mouse lineage when infected with S. aureus. These two genes, combined with other factors from chromosome 8 and 18, would together contribute to susceptibility to S. aureus in A/J mice, and potentially in humans.
Further, our studies showed that pre-exposure of macrophages to S. aureus particles compromised their ability to further take up the bacteria. This finding suggests that persistent, unabated stimulation of the immune system by S. aureus infection can eventually lead to immune paralysis or exhaustion of antimicrobial peptides. Since the overproduction of cytokines is the major phenotype of CSS11, we hypothesize that cytokine storm could account for the increased susceptibility of A/J to dying of S. aureus sepsis. Given these findings, we hypothesize that down-regulation of Dusp3 and Psme3 in A/J result in hyper-responsiveness of host immune system, which in turn leads to “immune paralysis” of the host to further defense against prolonged S. aureus challenge. Collectively, these finding also indicate that immediate bacterial clearance is important for host defense against S. aureus.
Several of the other genes identified in our experiments were also promising candidates. Dcaf7 is a scaffold protein for activating MEKK1 kinase [44] and is involved in the human TNFα/NF-κB signal transduction pathway [45]. A dominant negative mutant of MEKK1 was reported to abolish T-cell receptor activation by super-antigen staphylococcus enterotoxin E [46] while NF-κB mediated innate immune defense against S. aureus through TLR2 and NOD2 [16], [47], [48]. Fam134c is a family with sequence similarity to 134, member C. Little is known of the putative function of Fam134c or related family members [49]. However, mice with Fam134c homozygous mutations have demonstrated bacterial susceptible phenotype based on the screening results from the Knockout Mouse Consortium program [33], [34]. Slc4a1 is a membrane protein or protein of membrane related organelles mediating small molecule transporting and intracellular metabolism [50]–[52].
The current study has limitations. First, our cohort of patients did not include subjects who were colonized, but not infected, with S. aureus. Second, the two model systems used in this manuscript (intraperitoneal sepsis, tail vein sepsis) fail to fully represent the diversity of human infections caused by S. aureus (e.g., endocarditis, osteomyelitis, visceral abscesses, pneumonia, soft tissue infection). Our approach does not consider the impact of post-translational modification [53], [54] and single nucleotide polymorphisms on genes and their products. Thus, there may be additional candidate genes on chromosome 11 beyond Dusp3 and Psme3. Moreover, our approach may have missed genes that contribute to the susceptible phenotype by way of a joint or additive effect. In support of this possibility is the fact that our original discovery found that three chromosomes, 8, 11, and 18, were each independently associated with susceptibility to S. aureus [11]. Thus, additional experiments are underway in our lab, including QTL mapping analysis for A/J chromosome 8, defining the pathogenesis of Dusp3 and Psme3 using knockout mice, and evaluating the impact of Dusp3 and Psme3 on host susceptibility to different pathogens.
Despite these limitations, this study makes several key observations. First, we have identified one QTL on chromosome 11 that is significantly linked to survival time after infection with S. aureus. Eleven differentially expressed genes mapped to the significant- or suggestive- threshold of this QTL. Five of these 11 genes exhibited significant evidence of involvement in patients with S. aureus BSI that was consistent with the pattern encountered in the murine model. Two of these five genes, Dusp3 and Psme3, responded to S. aureus challenge by negatively regulating NF-κB signaling, leading to enhanced cytokine response (GM-CSF, IL-1β, IL-6, and TNFα). Consistent with the hypothesis of enhanced A/J susceptibility caused by unchecked inflammatory response, Dusp3 and Psme3 were less expressed in susceptible A/J as compared with resistant C57BL/6J. All of our results support a potential role of these two genes in host response to S. aureus. Dusp3 and Psme3 represent promising candidates for the genetic basis of host susceptibility to S. aureus.
All animal experiments were carried out in strict accordance with the recommendations of NIH guidelines, the Animal Welfare Act, and US federal law. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC Protocol A191-12-07) of Duke University which has been accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International. All animals were housed in a centralized and AAALAC accredited research animal facility that is fully staffed with trained husbandry, technical, and veterinary personnel. The Institutional Review Boards from all involved hospitals approved the human studies referenced in this work. Written informed consent was obtained for all subjects after the nature and possible consequences of the studies were explained.
Subjects were enrolled at Duke University Medical Center (DUMC; Durham, NC), Durham VAMC (Durham, NC), and Henry Ford Hospital (Detroit, Michigan) as part of a prospective, NIH-sponsored study to develop novel diagnostic tests for severe sepsis and community-acquired pneumonia as mentioned before [55]–[57]. All participants were adults. RNA was obtained from blood drawn at the time patients initially presented to the Emergency Department with sepsis. RNA expression data from patients who were ultimately found to have BSI with either S. aureus (n = 32) or E. coli (n = 19) were used in this study. Healthy controls were defined as uninfected human (n = 43), enrolled as part of a study on the effect of aspirin on platelet function among healthy volunteers. Subjects were recruited through advertisements posted on the Duke campus. Blood used to derive gene expression data in these healthy controls was drawn prior to aspirin challenge. Human orthologs of murine genes were identified by Chip comparer (http://chipcomparer.genome.duke.edu/) as reported before [58]. When there were multiple orthologs, we preferentially used the anti-sense target probes that shared the fewest probes with other genes as identified by the probe label.
C57BL/6J, A/J, and CSS11 mice were purchased from the Jackson Laboratory (Bar Harbor, ME). All the mice were allowed to acclimate for more than 7 days before experiments. For generation of F1 progeny, CSS11 mice were mated with C57BL/6J in reciprocal crosses [C57BL/6J male×CSS11 female and C57BL/6J female×CSS11 male] to generate an F1 population with heterozygous chromosome 11 due to homologous recombination. To generate F2 intercross mice for QTL linkage analysis, F1 (C11A) mice were intercrossed with F1 (C11A) to produce more than 200 progeny.
S. aureus clinical strain, Sanger 476 was used in the mortality and infection studies. For preparation of S. aureus for injection, overnight culture of S. aureus was diluted 100 folds with fresh tryptic soy broth (TSB) and shake at 37°C with aeration to log-phase (OD600≈0.8). S. aureus was harvested by centrifugation at 3000 rpm for 10 minutes at 4°C, washed once in DPBS and re-suspended in DPBS.
For murine peritonitis-sepsis experiments, 8-week-old male mice (n = 10) in each strain of C57BL/6J, A/J, and CSS11 were i.p. injected with 107 CFU/g S. aureus (Sanger 476) or 2×105 CFU/g E. coli (K1H7) and observed every 6 hours for morbidity continuously for 5 days. For murine intravenous sepsis, 8-week-old male mice (n = 10) in each strain of C57BL/6J, A/J, and CSS11 were i.v. injected with 2×106 CFU/g S. aureus (Sanger 476) and observed every 6 hours for morbidity continuously for 3 days. For bacterial load quantification, kidneys were collected from euthanized mice at 24 hours post-infection, then homogenized in DPBS and serially diluted (10 fold). The dilutions were plated in Tryptic Soy Agar (TSA) plates and incubated at 37°C overnight for counting colony forming units (CFU).
Polymorphic microsatellite markers on chromosome 11 between C57BL/6J and A/J were chosen from a database maintained by Mouse Genomic Informatics (http://www.informatics.jax.org/). Twelve microsatellite markers were selected with an average inter-marker distance of 3.1 cM covering chromosome 11. A total of 208 F2 intercross were generated, all of which were genotyped for each microsatellite marker by PCR amplification and gel electrophoresis. J/qtl software was used to analyze phenotype and genotype data for linkage of survival time after infection with S. aureus Sanger 476 and marker location. Phenotypes were defined as either sensitive or resistant based on the dichotomization of survival data (survival of less than 2 day is “0” and survival of longer than 2 days is “1”, respectively). All linkage analysis results were expressed as LOD scores. LOD score was considered “suggestive” if > = 1.6 (p = 0.63) and “significant” if > = 3.55 (p = 0.05). Threshold values for linkage were determined by a 1,000 permutation test by using J/qtl.
To generate bone marrow-derived macrophages (BMDMs), bone marrow progenitor cells were harvested from mice and cultured for 7 days in 70% (vol/vol) D10 (DMEM containing 10% (vol/vol) FBS, 2 mM glutamine, 100 µg/ml streptomycin, and 100 units/ml penicillin) and 30% (vol/vol) L-929 cell culture supernatant. Mature BMDMs were washed twice with cold DPBS, collected with 5 mM EDTA in DPBS, and re-plated on tissue-culture plates as reported before [15]. The murine RAW 264.7 macrophage cells (ATCC) were cultured in D10 in tissue culture plates before downstream experiments.
The day before treatment, 2×106 BMDMs from either C57BL/6J or A/J were seeded to 6-well plate for analysis of S. aureus phagocytosis ability. Next day BMDMs were either treated with fresh medium, TNF-α at 100 ng/ml, or S. aureus particles at 10 µg/ml for 24 hours. On the day of the phagocytosis experiment, S. aureus Sanger 476 were grown to exponential period (OD600≈0.8) and harvested by centrifugation. After washing in DPBS, S. aureus were stained by Hexidium Iodide (100 µg/ml) for 15 minutes at room temperature followed by washing once in DPBS and re-suspended in DPBS on ice [59]. Then multiplicity of infection (MOI) 10 was applied for S. aureus infection. Briefly, old medium were removed and replaced with fresh medium containing 2×107 S. aureus to each well of 6-well plate with BMDMs, followed by quick spin at 500 rpm for 5 minutes at room temperature. Cells were then incubated at 37°C for 30 min in CO2 incubator to allow bacterial uptake by macrophages. After introduction of S. aureus, trypsin was added at a final concentration of 0.25% for 10 minutes at room temperature to remove any residual bacteria at the macrophage surface. Macrophages were next washed three times with DPBS to remove remaining bacteria and floated by 5 mM EDTA in DPBS. Macrophages were then stained by FITC-F4/80 and analyzed by Flow cytometer [60]. The fluorescence produced from hexidium iodide staining falls into FL2 channel (Excitation 488/Emmision 575) in FACSCanto [59]. Experiments were repeated at least three times.
To test the role of each candidate gene on cytokine production by host defense cells, we transfected siRNAs into the mouse macrophages. All siRNAs were purchased from Invitrogen. RAW264.7 cells or bone-marrow derived macrophages from C57BL/6J were transfected with 50 nM siRNA by Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions. Twenty-four hours post-transfection, cells were treated with S. aureus Bioparticles (Invitrogen) to a final concentration of 10 µg/ml. At 24 hours post-infection, supernatants were collected and stored at −80°C for Luminex-multiplex cytokine assay. In parallel experiments, cells at 3 hours post-infection were harvested for RNA extraction by RNeasy (QIAGEN), RT-PCR by SuperScript II (Invitrogen) and SYBR Green qPCR analysis (ABI) respectively. Experiments were repeated at least three times. A full list of gene names and siRNA ID numbers were in Table S1.
The day before transfection, 5×106 RAW264.7 macrophages were seeded into 10 cm dishes. On the second day, 10 µg NF-κB-Luc (Clontech) and 0.5 µg pRL-TK (Promega) plasmids were co-transfected by Lipofectamine LTX (Invitrogen) according to manufacturer's instruction. At 6 hours post-transfection, RAW264.7 were split into 6-well plate (1×106/well) and cultured overnight in CO2 incubator. On the third day, scrambled siRNA or siRNA for each candidate gene was transfected into 6-well plate by Lipofectamine RNAiMAX according to manufacturer's instruction. At 6 hours post-transfection, cells were split into 24-well plate (4×105/well). On the fourth day, cells were stimulated by either D10 or D10 containing S. aureus lipoteichoic acid (LTA) (10 µg/ml) or S. aureus particles (10 µg/ml) for seven hours. Then cells were lysed by 1×Passive Lysis Buffer (Promega) and luciferase activities were analyzed by dual-luciferase reporter assay system (Promega). All of Firefly luciferase activity was normalized by the Renilla luciferase activity and relative fold changes were compared. Experiments were repeated at least three times.
4×105 RAW264.7 macrophages were seeded into each well in 24-well plate and cultured overnight in CO2 incubator. Then cells were treated by D10 with DMSO or 4 µM Bay 11-7085 (EMD Millipore) for one hour. Afterwards, the cells were stimulated by four different ways including D10+DMSO, D10+DMSO+S. aureus particles (10 µg/ml), D10+2 µM Bay and D10+2 µM Bay+S. aureus particles (10 µg/ml) for 3 hours. Then RNA was extracted, and reverse-transcription PCR and qPCR were applied. Experiments were repeated at least three times.
Cytokine production was assayed from the collected supernatant of both S. aureus-challenged siRNA transfected RAW cells and the S. aureus-challenged BMDM from C57BL/6J and CSS11 using multiplex cytokine assay kit (Invitrogen) and Luminex technology available at Duke Human Vaccine Institute.
Total RNA was isolated using RNeasy kits (Qiagen) primed with random hexamer oligonucleotides and reversely transcribed using Invitrogen SuperScript II. Real-time quantitative PCR was performed using SYBR Green Mastermix (ABI). All data were normalized to 18s rRNA.
Cells were lysed in RIPA buffer with cocktail of proteinase and protein phosphatase inhibitors. Then 20 µg whole cell lysate was loaded to SDS-PAGE and transferred to PVDF membrane. Blotting was followed according to manufacturer's instruction.
The differences in candidate gene expression, mRNA and protein of cytokines and chemokines and luciferase activities were analyzed by two-tailed Student's t test. The difference in mice survival rate was analyzed by Mann-Whitney u test. P-values smaller than 0.05 were considered to be statistically significant.
|
10.1371/journal.ppat.1000683 | The T3SS Effector EspT Defines a New Category of Invasive Enteropathogenic E. coli (EPEC) Which Form Intracellular Actin Pedestals | Enteropathogenic Escherichia coli (EPEC) strains are defined as extracellular pathogens which nucleate actin rich pedestal-like membrane extensions on intestinal enterocytes to which they intimately adhere. EPEC infection is mediated by type III secretion system effectors, which modulate host cell signaling. Recently we have shown that the WxxxE effector EspT activates Rac1 and Cdc42 leading to formation of membrane ruffles and lamellipodia. Here we report that EspT-induced membrane ruffles facilitate EPEC invasion into non-phagocytic cells in a process involving Rac1 and Wave2. Internalized EPEC resides within a vacuole and Tir is localized to the vacuolar membrane, resulting in actin polymerization and formation of intracellular pedestals. To the best of our knowledge this is the first time a pathogen has been shown to induce formation of actin comets across a vacuole membrane. Moreover, our data breaks the dogma of EPEC as an extracellular pathogen and defines a new category of invasive EPEC.
| Enteropathogenic E. coli (EPEC) is an important diarrheal pathogen responsible for significant infant mortality in the developing world and is increasingly associated with sporadic outbreaks in the developed world. The virulence strategy of EPEC revolves around a conserved Type 3 secretion system (T3SS) which translocates bacterial effector proteins directly into host cells. EPEC is considered to be a non-invasive pathogen which intimately adheres to host cells and polymerizes actin rich pedestals on which extracellular bacteria rest. Recently we have identified the T3SS effector EspT which activates the mammalian Rho GTPases Rac1 and Cdc42, resulting in the formation of membrane ruffles and lamellipodia. In this study we dissect the signaling pathway utilized by EspT to nucleate membrane ruffles and demonstrate that these ruffles can promote EPEC invasion of host cells. Furthermore, we show that internalized EPEC are bound within a vacuole. We also report for the first time the ability of a bacterial pathogen to form actin comet tails across a vacuole membrane. In addition to providing novel insights into the subversion of cellular signaling by invasive pathogens, our study also breaks the long held dogma of EPEC as an extracellular pathogen and will have implications on how future EPEC infections are diagnosed and treated.
| The human pathogens enteropathogenic Escherichia coli (EPEC) and enterohemorrhagic E. coli (EHEC) [1] and the mouse pathogen Citrobacter rodentium [2] are closely related extra-cellular diarrhoeal agents characterized by their ability to colonize the gut epithelium via attaching and effacing (A/E) lesion formation (reviewed in [3]) [4]. Similarly to other Gram-negative bacteria EPEC, EHEC and C. rodentium encode a type III secretion system (T3SS), which is central to their infection strategy (reviewed in [5]) [6]. This complex machinery translocate dozens of effector proteins [7],[8] directly from the bacteria to the eukaryotic cell cytoplasm (reviewed in [9]). The translocated effectors are targeted to various sub-cellular compartments where they subvert a plethora of cell signaling pathways via interactions with a range of host cell proteins.
The host cell cytoskeleton is a common target of T3SS effectors [10]. EPEC, EHEC and C. rodentium translocate the effector Tir into the plasma membrane where it functions as a receptor for the bacterial outer membrane protein intimin [11]. Intimin:Tir interaction leads to activation of N-WASP and formation of actin rich pedestals on which the extracellular bacteria rest [12]. In addition to Tir, A/E pathogens translocate a variety of other effectors which also modulate the host cell cytoskeleton including EspG/EspG2, which induce depolymerization of the microtubule network [13], Map, which induces formation of transient filopodia early in infection [14] and EspM which directs formation of actin stress fibers [15]. Map and EspM are members of the WxxxE family [15],[16],[17], which was first grouped together based on conserved peptide motif consisting of an invariant tryptophan and glutamic acid residues separated by three variable amino acids and their shared ability to subvert host cell small GTPase signaling.
Small GTPases cycle between an inactive GDP bound and an active GTP bound form, allowing them to function as molecular switches in response to a variety of stimuli. The switch from inactive to active forms results in a conformational change, which allows the GTPase to bind downstream mammalian effectors. Small GTPases are regulated by guanine exchange factors (GEFs), GTPase activating proteins (GAPs) and guanine dissociation inhibitor (GDI) proteins (reviewed in [18],[19]). The three best characterized Rho GTPases are RhoA, Rac1 and Cdc42 which are implicated in formation of stress fibers, lamellipodia and filopodia respectively (reviewed in [20]).
The WxxxE effectors were originally proposed to be functional mimics of mammalian small GTPases [16]. However, we have recently shown that EspM activates RhoA [15] whereas Map induces filopodia via activation of Cdc42 and RhoA [17]. In addition to Map and EspM we have recently discovered the novel WxxxE effector EspT, which is encoded by C. rodentium and a subset of EPEC strains [21], including EPEC E110019 which caused a sever outbreak in Finland in 1987 that affected children and adults alike [22]. We have shown that EspT induces formation of lamellipodia and membrane ruffles in epithelial cells via activation of Rac1 and Cdc42 [23].
Membrane ruffles are sheet like structures which are induced by mammalian cells in order to facilitate crawling movement, macro-pinocitosis and receptor recycling (reviewed in [24]). These protrusion are regulated through activity of Rho family GTPases and their downstream effectors (reviewed in [25]). Importantly, a subset of invasive bacterial pathogens hijack and subvert mammalian signal transduction pathways which facilitate formation membrane ruffles in order to promote bacterial entry into mammalian cells. Perhaps the best studied of these pathogens are Salmonella and Shigella which induce extensive membrane ruffles at the site of bacterial attachment (reviewed in [26],[27]). Salmonella invasion is dependent upon the activity of several T3SS effector proteins including SopE/E2 which act as GEFs for Rac1 and Cdc42 [28] and SopB which activates the RhoG GEF SGEF [29]. Shigella has also evolved several invasive mechanisms. For example the translocator IpaC has been shown to induce ruffles at the site of Shigella entry via the activation of Cdc42 [30], recruitment of Src kinase [31] and activation of Abl kinase [32]. The Shigella WxxxE effector IpgB1 has also been shown to induce membrane ruffles via interaction with the ELMO DOCK180 complex which results in activation of Rac1 [33].
EPEC, EHEC and C. rodentium are generally considered extracellular pathogens and their attachment sites on epithelial cells are normally characterized by the assembly of an actin rich pedestal rather than membrane ruffles (reviewed in [3]). However, in both rabbit and human biopsies EPEC have been visualized inside enterocytes and detected in the sub mucosa, mesenteric lymph nodes and spleen [34] (reviewed in [35]). Recently Hernandes et al has shown that the atypical EPEC strain 1551-2 is capable of invading cultured epithelial cells in an intimin omicron dependent manner [36]. As EspT induces membrane ruffles similar to those triggered by IpgB1 [37] the aim of this study was to investigate if expression of EspT leads to EPEC cell invasion and to define the underlying mechanism.
A large screen of clinical EPEC isolates for the presence of espT, a T3SS effector (Fig. S1), has shown that the gene is present in ca. 1.8% of the tested strains [21]. In order to investigate the role of EspT in cell invasion we selected to use the espT positive strains E110019 and C. rodentum; the espT negative EPEC, strain JPN15 [38], was used as a control. In addition, we generated a JPN15 clone that expresses EspT from the bacterial expression vector pSA10 (pICC461).
We infected serum starved HeLa, Swiss 3T3 and Caco2 cells with E110019, JPN15 and JPN15 expressing EspT; the cells were then fixed and processed for scanning electron microscopy (SEM). The JPN15-infected HeLa and Swiss 3T3 cells displayed characteristic diffuse bacterial adhesion without any noteworthy surface structures. Caco2 cells infected with JPN15 also show a diffuse pattern of bacterial adherence and a concordant localized effacement of microvili (Fig. 1). HeLa cells infected with JPN15 expressing EspT or E110019 displayed extensive membrane ruffling over the entire cell surface (Fig. 1); in the vicinity of adherent bacteria the ruffles surrounded and wrapped individual bacterial cells forming structures which appear permissive for internalization. Swiss 3T3 cells infected with JPN15 expressing EspT or E110019 exhibited extensive dorsal ruffles and lamellipodia in addition to localized membrane ruffles at the site of bacterial attachment (Fig. 1). Caco2 cells infected with JPN15 expressing EspT or E110019 displayed prominent membrane ruffles at the site of bacterial adherence in addition to effacement of micovili (Fig. 1). These results show that EspT can induce actin remodeling and surface structures, similar to those associated with Shigella and Salmonella invasion (reviewed in [26]).
We have recently shown that remodeling of the host cell actin cytoskeleton by EspT is dependent on Rac1 and to a lesser extent Cdc42 [23]. Rac1 and Cdc42 utilize a plethora of downstream effectors in order to regulate cytoskeletal dynamics (reviewed in [19] and [25]). Several GTPase effectors including IRSp53, N-WASP, Pak, Wave2 and Abi1 have been previously been implicated in formation of membrane ruffles [39],[40],[41],[42]. By using immuno-fluorescence microscopy we found that both Wave2 and Abi1 were present and co-localized with actin at membrane ruffles and the leading edge of lamellipodia induced by EspT (Fig. 2), while N-WASP was not (data not shown). The signaling protein IRSp53 has been proposed to participate in Abi1-Wave2-Rac1 complex formation [39],[43]. While we did not detect any significant enrichment of IRSp53 in lamellipodia, IRSp53 was localized to membrane ruffles nucleated by EspT (Fig. S2). Taken together these results show that Abi1 and Wave2 are localized to membrane ruffles and lamellipodia induced by EspT but IRSp53 is only recruited to EspT-induced membrane ruffles.
Wave2 is a ubiquitously expressed member of the WASP super family of actin regulators which potently activates the Arp2/3 complex [44]. The Wave family of proteins have a modular structure consisting of a N terminal Wave homology domain (WHD), a central proline rich region (PRR) and a C terminal Arp2/3 binding domain (VCA module) (reviewed in [45]). The WHD domain has been shown to bind Abi1 [42] and the PRR has been shown to interact with the SH3 domain of IRSp53 [39]. We utilized siRNA in order to determine if Wave2 is essential for formation of the EspT-dependent membrane ruffles. Depletion of endogenous Wave2 from Swiss 3T3 cells, confirmed by Western blotting (Fig. 3A), resulted in a marked decrease in formation of membrane ruffles and lamellipodia induced by JPN15 expressing EspT or E110019, compared with cells treated with scrambled siRNA (Fig. 3B). In order to determine which of the Wave2 domains is required for formation of lamellipodia and membrane ruffles by EspT, we transfected Swiss 3T3 and HeLa cells with full length Wave2 or dominant negative forms of Wave2 lacking the WHD (ΔBP) or the acidic Arp2/3 interacting domain (ΔA). Transfected cell were infected for 1.5 h with JPN15 expressing EspT and the presence of lamellipodia or membrane ruffles was assessed. Mock transfected cells or cells transfected with full length Wave2 had lamellipodia and membrane ruffles on 80 to 90% of infected cells (Fig. S3). In contrast, transfection with either the ΔBP or the ΔA Wave2 dominant negative constructs resulted in significant reduction in lamellipodia and membrane ruffle formation (Fig. S3). This result demonstrates that binding of Arp2/3 to Wave2 is essential for EspT-mediated formation of lamellipodia and membrane ruffles. Furthermore, the observation that the N terminal truncated Wave2 ΔBP construct has a dominant negative effect suggests that the WHD motif is also required for EspT mediated cytoskeletal rearrangements. The fact that the Wave2 ΔBP construct, which is capable of binding IRSp53 but not Abi1, is not sufficient to induce EspT dependent actin remodeling further indicates that IRSp53 does not play a prominent role in EspT mediated signaling.
Induction of membrane ruffles is a mechanism employed by a range of pathogenic bacteria in order to facilitate cell invasion. This method of bacterial invasion is referred to as the trigger mechanism and relies upon induction of actin polymerization to form an entry foci and a macropinocytic pocket (reviewed in [26]). JPN15 expressing EspT and E110019 induce host cell membrane remodeling which is reminiscent of entry foci and membrane ruffles induced by Shigella and Salmonella (reviewed in [26]) (Fig. 1).
We used differential staining to visualize invasion of Swiss 3T3 cells by JPN15, JPN15 expressing EspT, and E110019; Salmonella enterica serovar Typhimurium strain SL1344 was used as a control. In addition, we conducted gentamycin protection assays to quantify cell invasion of Swiss 3T3, HeLa and Caco2 cells after 3 h infection. Differential immuno-fluorescence staining and gentamycin protection assays were also performed in HeLa and Swiss 3T3 cells infected for 6 h with wild type C. rodentium, C. rodentium ΔespT and complemented C. rodentium ΔespT. For immuno-fluorescence extracellular bacteria were stained pre cell permeabilzation with primary anti O127 (JPN15), anti O111 (E110019), anti O152 (C. rodentium) or anti LPS (S. Typhimurium) antibodies and a secondary antibody coupled to a Cy3 fluorophore (red). The cells were then permeabilized and total bacteria were stained with the same primary antibodies and a secondary antibody coupled to a Cy2 fluorophore (green), Alexafluor 633 phalloidin and Dapi were used to visualize actin and DNA respectively.
Adherent JPN15 bacteria were homogenously stained by both the extracellular and total bacterial probes, indicating that this strain is not significantly invasive (Fig. 4A). In cells infected with JPN15 expressing EspT or E110019 a significant proportion of the bacteria were labeled with the total bacterial stain but not by the extracellular probe (Fig. 4A). S. Typhimurium-infected cells exhibited characteristic membrane ruffling at the entry foci and a high proportion of bacteria were labeled only with the total bacterial probe (Fig. 4A).
The quantitative gentamycin protection assay revealed that JPN15 does not efficiently invade HeLa, Swiss 3T3 or Caco2 cells, exhibiting an invasion rate of less than 1.5% (Fig. 4B). JPN15 expressing EspT was significantly more invasive with an invasion rate of 15.5% in Swiss 3T3, 14.3% in HeLa and 7.2% in Caco2 cells (Fig. 4B). E110019 invaded Swiss 3T3, HeLa and Caco2 cells at a rate of 9.2%, 11.4% and 5.8% respectively. The invasive capacity of EPEC was significantly less than S. Typhimuriumin (Fig. 4B). Infection of HeLa cells with JPN15 expressing EspTW63A for 3 h confirmed that the WxxxE motif plays a major role in membrane ruffling and cell invasion (Fig. S4).
E110019 is multi drug resistant, which limits the ability to genetically modified the isolate. In order to determine if cell invasion is mediated by EspT, we infected Swiss 3T3 cells for 6 h with wild type C. rodentium and C. rodentium ΔespT. Infection with wild type C. rodentium resulted in membrane ruffles and cell invasion, while the espT mutant exhibited neither (Fig. 4B). Complementing the mutant with espT expressed from pACYC184 (pICC489) restored membrane ruffle formation and cell invasion (Fig. 4B and S5).
In order to confirm that EspT can promote EPEC invasion of non-phagocytic cells independently of other T3SS effectors we ectopically expressed EspT in HeLa cells prior to infection with EPEC ΔescN, a T3SS null mutant. Cells ectopically expressing EspT displayed membrane ruffling which facilitated the uptake of ΔescN bacteria (Fig. S6). No membrane ruffles were observed in cells ectopically expressing EspTW63A (data not shown). These results show that EspT induces EPEC cell invasion by a trigger mechanism, analogous to that of Shigella and Salmonella.
As actin remodeling by EspT is dependent on activation of Rac1, Cdc42 [23] and Wave2 (Fig. 3), we utilized dominant negative constructs of these signaling proteins and Wave2 siRNA to monitor the effect on invasion of JPN15 expressing EspT and E110019. Swiss 3T3 cells transfected with dominant negative Rac1 (Rac1N17), Cdc42 (Cdc42N17), Wave2ΔA truncated in the acidic Arp2/3 interacting region and Wave2ΔBP lacking the WHD were infected for 3 h. The cells were fixed and stained for bacterial invasion as described above. Cells transfected with Cdc42N17 were still permissive of bacterial invasion while cells transfected with the Rac1N17, Wave2ΔA or Wave2ΔBP dominant negative constructs were not (Fig. 5B). Depletion of Wave2 using siRNA in Swiss 3T3 cells significantly reduced the invasive capacity of both JPN15 expressing EspT and E110019 compared to cells treated with non-targeting siRNA (Fig. 3A and 5B). Thus, Rac1, Wave2 and Abi1 are essential mediators of EspT-induced bacterial invasion.
After the initial invasion of host cells internalized bacteria are often bound within a vacuole which resembles early endosomes (reviewed in [46]). Intracellular bacteria either remain within the vacuole or rapidly escape to the cytoplasm [26]. In order to determine whether invasive EPEC are bound within a vacuole or free in the cytoplasm HeLa cells were infected with JPN15, JPN15 over expressing EspT (pICC461) and E110019 for 5 min up to 24 h and stained with various vacuolar markers including Early Endosome Antigen 1 (EEA1), Vacuolar ATPase (VATPase) and Lamp1. Internalized JPN15 expressing EspT and E110019 were labeled with EEA1 while external bacteria and JPN15 lacking EspT were not (Fig. 6 shows staining at 45 min post infection). EEA1 staining was apparent after 5 min and persists up to 1 h post infection (data not shown). At 3 h and up to 12 h post infection the EPEC containing vacuole (ECV) was labeled with VATPase whilst external bacteria were not (Fig. 7B). Similarly to the Salmonella containing vacuole (SCV), a subset of ECVs became enriched with the lysosomal glycoprotein Lamp1 after 16 h (Fig. S7 and S8) and appear to adopt a perinuclear localization (Fig. 6B).
In order to determine if EPEC bacteria can multiply intracellularly we infected Swiss 3T3 cells with E110019 for 30 min before extracellular bacteria were killed by gentamycin. Infected cells were fixed for immuno-fluorescence microscopy at 2, 8, 16 and 24 h post infection. We observed a time dependent increase in the level of intracellular bacteria suggesting that internalized EPEC can multiply within host cells (Fig. S8).
After escaping from the vacuole many intracellular pathogens such as Shigella, Burkholderia and Listeria utilize specialized outer membrane proteins to recruit actin nucleating factors in order to produce a propulsive force (reviewed in [26]). EPEC is synonymous with actin nucleation which leads to formation of Tir-dependent actin rich pedestals [12]. During the course of this study we observed that invasive EPEC were associated with filamentous actin comets reminiscent of pedestals. Confocal X-stacks confirmed that the intracellular EPEC bacteria were associated with pedestal-like filamentous actin structures (Fig. 7A). In order to determine if Tir was localized at the actin nucleation sites, we infected HeLa cells for 1 h with JPN15, JPN15 expressing EspT and E110019; following washes the cells were treated with gentamycin for a further 8 h. The cells were then stained with anti-VATPase and anti-Tir antisera in conjunction with phalloidin and Dapi staining. HeLa cells infected with JPN15 exhibited extracellular, pedestal-associated, bacteria which were associated with Tir but not with VATPase (Fig. 7B). In contrast, internalized JPN15 expressing EspT and E110019 bacteria were co-labeled with anti-VATPase, actin and Tir (Fig. 7B). Similarly, invasive C. rodentium also formed intracellular pedestals (Fig. S5), while C. rodentium Δtir was invasive but failed to trigger actin polymerization (Fig. S8). In addition, the intracellular EPEC ΔescN, internalized by ectopically expressing EspT, were not associated with actin pedestals (Fig. S6). These results suggest that the actin filaments associated with EPEC contained within the ECV is nucleated in a Tir-dependent mechanism analogous to pedestal formation by extracellular bacteria.
In order to confirm this assertion we infected HeLa cells with E110019 for 2 h and processed the cells for transmission electron microscopy (TEM). The TEM confirmed that E110019 bacteria are internalized via ruffle formation (Fig. 7C). E110019 can also be seen forming multiple pedestals with the membrane on opposing surfaces during ruffle formation and closure (Fig. 7D). Moreover, internalized EPEC bacteria contained within the ECV are associated actin pedestals, which are strikingly similar to those normally associated with extracellular EPEC (Fig. 7C–E). Interestingly, bacteria bound within ECVs can form multiple pedestals around their circumference (Fig. 7E).
In order to determine if the formation of intracellular pedestals by A/E pathogens plays a role in bacterial replication and survival within host cells we infected Swiss cells with wild type C. rodentium and C. rodentium Δtir for 1.5 h and with E110019 for 30 min. Extracellular bacteria were then killed by a gentamycin wash and the cells incubated for a further 6, 12 or 24 h. We observed that both wild type C. rodentium and E110019 were capable of intracellular replication whereas the C. rodentium Δtir mutant failed to replicate and instead exhibited a slow decline in bacterial numbers over time (Fig. S8). These results demonstrate that formation of pedestals by invasive A/E pathogens may play a functional role during intracellular survival.
A/E pathogens have been long considered to be extracellular bacteria which do not invade mammalian cells [47]. However, sporadic reports have shown that atypical EPEC strains can invade non-phagocytic cells [36],[48]. The invasive ability has been linked to intimin-Tir mediated tight association of EPEC with the host cell membrane which is hypothesized to produce immature phagocytosis cups leading to a passive push effect and inefficient internalization [35],[36]. In this study we demonstrated for the first time that EPEC can actively invade non-phagocytic cells by inducing formation of membrane ruffles, defining a new category of invasive EPEC. Furthermore we demonstrate that this phenomenon is dependent on the T3SS effector EspT which has previously shown activate Rac1 and Cdc42 [23]. We also show that both actin remodeling and invasion is dependent upon a functional EspT as JPN15 expressing a EspTW63A failed to induce membrane ruffles or to invade. Importantly, in a previous report we indicated that expression of EspT might not confer bacterial invasion of epithelial cells [23]. However, these experiments were conducted using EPEC E2348/69, which in contrast to JPN15, C. rodentium and E110019, forms tight microclonies that mask the invasion phenotype (data not shown).
Intracellular pathogens have evolved a variety of mechanisms to promote invasion of mammalian cells, including the trigger (employed by Salmonella and Shigella) and zipper (employed by Yersinia and Listeria) mechanisms (reviewed in [26]). The trigger invasion mechanism is characterized by formation of actin rich membrane ruffles at the site of bacterial attachment, which are regulated by Rho GTPases, particularly Rac1 and Cdc42 and other cytoskeletal regulators such as PI3K [49]. Shigella and Salmonella utilize T3SS effector and translocator proteins such as IpgB1 and IpaC and SopB and SopE/2 to hijack host cell GTPase and phospho-inositol signaling to modulate membrane ruffling and formation of the macropinocytic pocket [28],[29],[30],[37]. Importantly, although both IpgB1 and EspT belong to the WxxxE family of effectors and play a prominent role in bacterial invasion by inducing membrane ruffles, we have recently shown that they activate Rac1 by distinct mechanisms [23].
Downstream of Rho GTPase signaling, membrane ruffle formation is nucleated by the WASP super family proteins including N-WASP and Wave2. Salmonella invasion has been demonstrated to be at least in part dependent upon the Arp2/3 binding activity of Wave2 and also the association of Wave2 with Abi1 [50]. Wave2 cannot bind Rac1 directly; two different mechanisms have been proposed to describe how a Wave2-Rac1 complex is formed. Innocenti et al and Steffen et al demonstrate that Wave2 binds to Abi1 and two accessory proteins PIR121 and Nap1 which mediate Rac1 binding [42],[51]. A report by Miki et al proposed that IRSp53 is the protein which links Rac1 to Wave2 [39]. In this study we demonstrate that EspT activation of Rac1 leads to a downstream recruitment of Wave2, Abi1 and IRSp53 to membrane ruffles. Depletion of endogenous Wave2 using siRNA resulted in a significant reduction in both the level of membrane ruffles induced by strains expressing EspT and their associated invasive capacity. We also show that the Arp2/3 and Abi1 binding regions of Wave2, but not N-WASP, are required for EspT-induced membrane ruffles and invasion. Furthermore, a construct of Wave2 which retained the IRSp53 and Arp2/3 binding regions but lacked the Abi1 interacting domain had a dominant negative effect on membrane ruffle formation, suggesting that IRSp53 is not required for, but may play an accessory role in, EspT-mediated actin rearrangements.
Once internalized Shigella and Listeria quickly escape the vacuole (reviewed in [52]). In contrast, Salmonella remains vacuole bound and utilizes different virulence factors to modify the vacuolar environment, position and interaction with the host endomembrane system in order to create an intracellular replicative SCV (reviewed in [53]). In this study we demonstrated that after invasion EPEC is bound within a vacuole (ECV) and remains vacuolated until at least 16 h post infection. We found that the ECV is EEA1 positive for up to 1 h post infection and progresses to being VATPase positive from 3 h to 12 h post infection. Furthermore, 12 h after infection the ECV appears to adopt a peri-nuclear position, which resembles the properties of the SCV. Similarly to the SCV (reviewed in [53]) we found that a subset of ECVs become enriched in the lysosomal glycoprotein Lamp1 (data not shown) indicating lysosomal fusion with the ECV. We also demonstrate that internalized EPEC bacteria can survive and replicate within host cells in a time dependent manner. Importantly and uniquely, we found that the ECV is associated with filamentous actin tails, which are reminiscent of the extracellular pedestals normally nucleated by EPEC stains. Formation of intracellular actin pedestal were essential for bacterial survival, as the intracellular population of invasive tir mutant declined over time.
Formation of extracellular pedestals is dependent upon the T3SS effector Tir [11]. The interaction of Tir with intimin triggers recruitment of the mammalian adaptor Nck which in turn recruits and activates N-WASP leading to Arp2/3 recruitment and actin polymerization [11],[54],[55],[56]. In this study we found that internalized EPEC can localize Tir to the vacuolar membrane in a T3SS dependent manner and that the localization of Tir can promote actin nucleation. Furthermore, we found that a C. rodentium Δtir mutant is still invasive but does not form intracellular pedestals, demonstrating that pedestal formation by internalized bacteria is a Tir-dependent process analogous to that of extracellular bacteria. Additionally using TEM we found that invasive EPEC bound within a vacuole are associated with intracellular pedestals around the circumference of the bacteria. Interestingly, membrane ruffles seen engulfing invading EPEC were occasionally associated with pedestals, suggesting the pedestals can be formed during or after internalization.
Canonically actin is recruited to the surface of intracellular pathogens which are non-vacuolated and this recruitment is mediated by outer-membrane proteins which are free to interact with host cell signaling molecules present in the cytoplasm. For example following escape from the vacuole Shigella and Listeria utilize IcsA/VirG and ActA, respectively, to trigger actin polymerization and motility (reviewed in [57]). The Vaccinia virus uses the viral membrane protein A36R in order to generate actin based motility in a similar manner to the extracellular EPEC pedestals [58]. Importantly, the SCV is also associated with an actin nest which is required to maintain the integrity of the vacuole and support the intracellular replication of Salmonella (reviewed in [59]).
Due to the positioning of the actin extensions around the entire circumference of EPEC it is unlikely these intracellular pedestals are involved in classical actin-based motility. However, there are reports suggesting that actin polymerization and depolymerization around the periphery of E-cadherin-coated beads can lead to directional movement in process referred to as flashing [60]; for this reason at this stage we cannot rule out the possibility that intracellular pedestals confer actin based motility. Furthermore, formation of intracellular pedestals by invasive EPEC may play a role in maintaining the vacuole integrity in a similar way to that described for other vacuolated pathogens [59]. To the best of our knowledge the current study demonstrates for the first time that an intracellular bacteria is able to recruit filamentous actin comets to the pathogen cell surface whilst encapsulated in a vacuole.
In order to survive within intracellular niche vacuolated bacteria must evade host cell lysosome mediated degradation. Interestingly, during the course of this study we observed that internalized EPEC, which were enclosed in ECVs, displaying strong actin staining around their circumference were rarely Lamp1 positive, whereas ECVs which had little or no actin polymerization associated with them were homogenously labeled with Lamp1 (data not shown). Furthermore, we observed that a C. rodentium Δtir mutant was attenuated for intracellular replication. We propose that formation of actin rich intracellular pedestals around the circumference of the ECV by invasive EPEC may constitute a physical barrier to lysosome fusion protecting the enclosed bacteria from lysosomal degradation; however this hypothesis requires further testing. A similar phenomenon has been described for the trafficking of endosomes and lysosomes to wounded sites of plasma membrane. At sites of plasma membrane disruption lysosomes and endosomes are recruited to seal the breach, this process is inhibited if the cortical actin meshwork is stabilized and enhanced when it is disrupted [61]. Similarly the lysosome dependent internalization of Trypanosoma cruzi requires a depolymerization of the cortical actin network to allow lysosome transit to the plasma membrane [62].
Recently, while screen ca. 1000 clinical EPEC and EHEC isolates we found that none of the EHEC strains and only 1.8% of the EPEC strains contain espT [21]. Interestingly, espT was found in EPEC E110019 which was linked to a particularly sever outbreak of gastroenteritis in Finland [22]. E110019 was found to be particularly infectious and unusually for EPEC was associated with person to person spread and adult disease [22]. Although we have no clinical data of the other espT positive isolates it is tempting to speculate that the expression of EspT could be at least in part responsible for the hyper virulence of the E110019 strain. Further studies of the invasive EPEC category are needed to assess the risk they pose to human health.
Bacterial strains used in this study are listed in Table 1. The C. rodentium ΔespT were constructed using the using the one-step PCR λ-red-mediated mutation protocol [63] The O111:H2 E110019 strain was isolated from an outbreak in Finland [22]. All the strains were maintained on Luria–Bertani (LB) broth or agar supplemented with ampicilin (100µg/ml) or Kanamycin (50 µg/ml).
Plasmids used in this study are listed in Table 2; primers are listed in Table 3. espT was amplified by PCR using E110019 genomic DNA as template and cloned into pSA10 [64] (primer pair 1 and 2). All constructs were verified by DNA sequencing. Site directed mutagenesis of EspT was carried out using a Quickchange® II kit (Stratagene) and primers 3 and 4 according to the manufacturers instructions. Plasmids pSA10::espT was used as template for the mutagenic reactions. The pCX340 vector encoding EspT-TEM fusion was constructed after amplification of espT from C. rodentium using primer pair 5 and 6.
The mammalian expression vector pRK5 containing one of RacN17 or Cdc42N17 dominant negatives used in the transfection assays were a gift from Nathalie Lamarche-Vane. The pRK5 encoding Wave2, and the pDSRED Wave2ΔA and Wave2ΔBP were a kind gift from Laura Machesky via Ray Carabeo.
48 h prior to infection cells were seeded onto glass coverslips at a density of 5×105 cells per well and maintained in DMEM supplemented with 10% FCS at 37°C in 5% CO2. Caco2 cells were grown in DMEM supplemented with 20% FCS at 37°C in 5% CO2. The cells were washed in PBS and the media changed every 24 h for 12 days until the cells polarized. 3 h before infection, the cells were washed 3 times with PBS, the media replaced with fresh DMEM without FCS supplemented with 1% mannose and 500 µl of primed bacteria were added to each well, the plates were then centrifuged at 1000 rpm for 5 min at room temperature and infections were carried out at 37°C in 5% CO2.
HeLa cells were seeded on to glass coverslips for infection as previously described above. Wild type EPEC E2348/69 and ΔescN T3SS null mutant were transformed with the pCX340 vector encoding EspT-TEM-1 fusion; an NleD-TEM fusion was used as a positive control. Translocation assay was performed as described previously [65].
Coverslips were washed 3 times in PBS and fixed with 3% Paraformaldehyde (PFA) for 15 min before washing 3 more times in PBS. For immuno-staining, the cells were permeabilized for 5 min in PBS 0.5% Triton X100, washed 3 times in PBS and quenched for 30 min with 50 mM NH4Cl. Pre prermeabilized samples were not treated with triton X100. The coverslips were then blocked for 1 h with PBS 0.5% BSA before incubation with primary and secondary antibodies. The primary antibody mouse anti EEA1 (BD biosciences) and mouse anti VATPase (Gifted by Prof. D. Holden) were used at a dilution of 1∶100, while rabbit anti O127, anti O111, anti O152, anti Tir and goat anti CSA-1 (salmonella LPS, gifted by Prof D. Holden) were used at a dilution of 1∶500. Rabbit anti Wave2 (SantaCruz Biotechnology) and Mouse anti Abi1 (Abcam) were used at 1∶200 dilutions. Coverslips were incubated with the primary antibody for 1 h, washed three times in PBS and incubated with the secondary antibodies. Donkey anti-rabbit IgG conjugated to a Cy2 or Cy3 fluorophore, donkey anti-mouse IgG conjugated to a Cy5 or Cy5 fluorophore, donkey anti goat IgG conjugated to a Cy2 or Cy3 fluorophore (Jackson laboratories) were used at a 1∶200. Actin was stained using AlexaFluor 633 phalloidin, Oregon Green phalloidin or Rhodamine phalliodin (Invitrogen) at a 1∶100 dilution. All dilutions were in PBS/0.5% BSA. Coverslips were mounted on slides using ProLong® Gold antifade reagent (Invitrogen) and visualized by Zeiss Axioimager immunofluorescence microscope using the following excitation wavelengths: Cy3 – 550nm, Cy5 – 650nm and Oregon Green – 488nm. All images were analyzed using the Axiovision Rel 4.5 software. Confocal X stacks were taken using a Leica Sp2 microscope. Cell boundaries were determined using actin staining and Abobe photoshop software.
Swiss 3T3 cells or HeLa cells were transfected with pRK5 encoding EspT, RhoAN19, RacN17, Cdc42N17 dominant negatives fused to a Myc tag, pDSRED encoding Wave2, Wave2ΔA or Wave2ΔBP by lipofectamine 2000 (Invitrogen) according to the manufacturer's recommendations. The cells were incubated at 37°C in a humidified incubator with 5% CO2 for 16 h, washed twice in PBS before having their media replaced with DMEM as described previously. Transfected cells were infected with the appropriate strain as described above.
HeLa cells were seeded at a density of approximatetly 5×106 cells per well 24 h prior to transfection of either Wave2 siRNA pool or a non-targeting pool supplied by Dharmacon using HiPerFect (Qiagen) according to the manufacturers instructions. The media was changed 16 h after transfection and the cells were allowed to recover for 12 h before being trypsinated and seeded at a density of 5×106 cells. The siRNA procedure was repeated for a total of 3 rounds before the cells were used. Levels of Wave2 and tubulin were then detected by western blotting using anti wave2 (Santa Cruz) and anti tubulin (Sigma) antibodies. Cells were then infected with the appropriate strain and processed for immuno-fluorescence microscopy as previously described.
Cells seeded into the wells of a 24 well plate were infected as described above for 6 h at 37°C in 5% CO2. The pre-gentamycin plates were washed 5 times in PBS and then permeabilzed for 15 minutes with 1% saponin in sterile water before plating in triplicate on LB plates in dilutions ranging from 100 to 10−7. The post gentamycin samples were washed 5 times with PBS after the final wash the PBS was replaced with serum free DMEM containing 200µg/ml of gentmaycin and the cells incubated for 1 h at 37°C in 5% CO2. The plates were then washed a further 5 times in PBS before permeabilization and plating as described above. The pre and post gentamycin plates were then incubated for 15 h in a static 37°C incubator and the colony forming units (cfu) were counted. The percentage of invasion was calculated based on the ratio of cfu on the pre and post gentamycin plates.
Glass coverslips were seeded and infected for 2 h with the appropriate strains as described above. The cells were washed 3 times in phosphate buffer pH7.2 and then fixed with 2.5% Gluteraldehyde (Agar) in phosphate buffer pH7.2 for 15 min. The coverslips were then washed with phosphate buffer pH7.2 a further 3 times before being post fixed in 1% Osmium Tetroxide for 1 h. The cells were then washed 3 times in phosphate buffer before being washed for 15 min in graded ethanol solutions from 50% to 100% to dehydrate the samples. The cells were then transferred to an Emitech K850 Critical Point drier and processed according to the manufacturer's instructions. The coverslips were coated in gold/palladium mix using a Emitech Sc7620 minisputter to a thickness of approximately 370Å. Samples for scanning electron microscopy (SEM) were then examined blindly at an accelerating voltage of 25 kV using a Jeol JSM-6390.
6 well plates were seeded and infected for 2 h with the appropriate strains. The cells were washed 3 times in phosphate buffer pH7.2 and then fixed with 2.5% Gluteraldehyde in phosphate buffer pH7.2 for 15 min. The plates were then washed with phosphate buffer pH7.2 a further 3 times before being removed from the plate using a Teflon scraper and subsequently harvested into in eppendorf tube. The eppendorfs were then centrifuged at 10,000 RPM to pellet the cells. The cell pellets were post fixed in 1% Osmium Tetroxide for 1 h, followed by 1% buffered tannic acid for 30 min and then a 1% aqueous sodium sulfate rinse for 10 min. The sample was dehydrated using an ethanol-propylene oxide series (with 2% uranyl acetate added at the 30% step) and embedded in Epon-araldite for 24 h at 60°C. Ultrathin sections (60 nm) were cut with a Leica EMUC6 ultramicrotome, contrasted with uranyl acetate and lead citrate, and viewed with an FEI 120-kV Spirit Biotwin TEM. Images were obtained with a Tietz F415 digital TemCam.
|
10.1371/journal.pbio.1002135 | Extracellular Adenosine Mediates a Systemic Metabolic Switch during Immune Response | Immune defense is energetically costly, and thus an effective response requires metabolic adaptation of the organism to reallocate energy from storage, growth, and development towards the immune system. We employ the natural infection of Drosophila with a parasitoid wasp to study energy regulation during immune response. To combat the invasion, the host must produce specialized immune cells (lamellocytes) that destroy the parasitoid egg. We show that a significant portion of nutrients are allocated to differentiating lamellocytes when they would otherwise be used for development. This systemic metabolic switch is mediated by extracellular adenosine released from immune cells. The switch is crucial for an effective immune response. Preventing adenosine transport from immune cells or blocking adenosine receptor precludes the metabolic switch and the deceleration of development, dramatically reducing host resistance. Adenosine thus serves as a signal that the “selfish” immune cells send during infection to secure more energy at the expense of other tissues.
| The immune response is energetically costly and often requires adaption of the whole organism to ensure it receives enough energy. It is not well understood how distribution of energy resources within the organism is regulated during an immune response. To understand this better, we used parasitoid wasp infection of fruit fly larvae—the host larvae have 48 h before they pupate to destroy the infecting “alien” or face destruction by the parasitoid that will consume the developing pupa. Here we find a signal, generated by the host immune cells, which mediates a systemic energy switch. This signal—adenosine—suppresses processes driving larval to pupal development of the host, thereby freeing up energy for the immune system. We show that the resulting developmental delay in the fruit fly larvae is crucial for an efficient immune response; without the adenosine signal, resistance to the parasitoid drops drastically. Generation of this signal by immune cells demonstrates that in response to external stressors, the immune system can mobilize reallocation to itself of energy and nutrients from the rest of the organism.
| Immune response is energetically costly [1,2]. Immune cells, upon activation, favor glycolysis over oxidative phosphorylation for fast, albeit inefficient, energy generation and macromolecule synthesis [3,4]. This metabolic shift requires extra glucose as glycolysis produces much less ATP than does oxidative phosphorylation [5]. Therefore, at the organismal level, the energy shifts from storage and nonimmune processes towards the needs of the immune system [6–9].
Regulation of energy during the immune response is critical—full response requires a significant amount of energy, and inability to provide it with nutrients can lead to immune system suppression and reduced resistance [10–12]. In mammalian systems, the inflammatory cytokines TNF-α, IFN-γ, IL-1, and IL-6 are released upon recognition of the pathogen and, besides modulating immune functions, they also stimulate energy release [2,13–16]. Immune cells must respond rapidly to the activating signals, and thus they change their metabolism, which involves, at least in mammalian systems, the preferential use of aerobic glycolysis, known as the Warburg effect [3,4,17]. The increased demand for energy by the immune system requires, both in vertebrates and invertebrates, adaptation of the whole organism, which is associated with an overall metabolic suppression and a systemic insulin resistance in all tissues except the immune cells [2,12,18]. The importance of the systemic regulation of energy is demonstrated by examples of certain infections leading to depletion of energy reserves (wasting) and eventually death of the organism [15,19]. Despite the importance of the systemic regulation of energy, we have only fragmentary knowledge about the molecular mechanisms involved in the regulation of energy during immune response at the organismal level and about the communication between different parts of the organism mediating the shift of energy from storage and growth towards immunity [12,20,21].
Extracellular adenosine (e-Ado) is a signal originating from damaged or stressed tissues. Acting as an energy sensor, e-Ado is released from metabolically stressed cells with depleted ATP [22,23] or made from extracellular ATP leaking from damaged tissues [24]. e-Ado then works as a local or systemic hormone, adjusting metabolism by acting either via adenosine receptors or by the uptake into the cells and conversion to AMP activating AMP-activated protein kinase (AMPK) [24,25]. These actions lead to a suppression of energy consuming processes [22,26–29] and to a release of energy from stores [30].
Damaged tissues and metabolically stressed cells are very likely to occur during immune response and thus it is not surprising that elevated levels of e-Ado are also detected, for example, during sepsis in humans [31]. The capacity of e-Ado to regulate energy metabolism, to “measure” the level of tissue and organismal stress, and to adapt the energy use to the actual situation all make e-Ado a perfect candidate for an energy regulator during immune response. However, the mode of e-Ado action under immune challenge is unclear, as the role of e-Ado in energy regulation has mainly been studied in relation to anoxia in anoxia-tolerant organisms such as turtles and hypoxia and ischemia in rodent models and human patients [22,30], while e-Ado has thus far been associated with mammalian immune response only through its immunomodulatory and anti-inflammatory function [24,32].
We, and others, have previously shown that adenosine regulatory and signaling network in Drosophila is similar to mammalian systems [33–37]. In addition, we have shown that e-Ado regulates energy metabolism in Drosophila. Increase of e-Ado levels caused by a deficiency of adenosine deaminase-related growth factor A (ADGF-A) leads to hyperglycemia and reduced energy storage [38]. We have also found that the regulation of e-Ado by ADGF-A is particularly important during parasitoid wasp infection in Drosophila larvae; ADGF-A is strongly expressed in immune cells that encapsulate the invading wasp egg [39]. These findings further support a potential role of e-Ado in energy regulation during immune response.
Here, we use the parasitoid wasp infection as a model to study the energy regulation during immune reaction. Parasitoid wasps inject their eggs into Drosophila larvae, and if the fly larva does not destroy the egg in time, the hatched wasp larva will consume the host [40]. The fly larva recognizes the egg and mounts a robust immune response that involves proliferation and differentiation of specialized immune cells, lamellocytes, which eventually encapsulate the parasitoid egg. Using this immune response as a model, we traced the dietary glucose destinations, measured selected metabolites and gene expressions, and analyzed host resistance and the impact of the immune response on its development.
We describe here the systemic changes in energy metabolism during the immune challenge and the role of e-Ado in the regulation of these changes. We have found that e-Ado, released from the immune cells, mediates a metabolic switch characterized by the suppression of nutrient storage and developmental growth in favor of the immune defense. This metabolic switch—a tradeoff between development and defense—is crucial for the resistance to infection. In Drosophila larvae lacking adenosine signaling, development is not suppressed, and the resistance dramatically drops.
The endoparasitoid wasp Leptopilina boulardi injects its egg in early third-instar Drosophila larva. The egg, usually hiding in gut folds, is first recognized by the host-circulating hemocytes (Fig 1A) and the recognition triggers immune response [40]. This involves production of specialized cells called lamellocytes (Fig 1A and 1B) within the first 24 h postinfection (hpi; 0 hpi is the time of infection and corresponds to 72 h after egg laying; the time in hpi is also used for the uninfected control). Lamellocytes are then released into circulation, and the egg gets encapsulated with subsequent melanization by 48 hpi (Fig 1A). Production of lamellocytes involves a transient proliferation of prohemocytes in the lymph gland and their terminal differentiation into lamellocytes [41]. The efficiency of egg encapsulation depends on the ability to produce lamellocytes and thus varies among different genetic strains of Drosophila [42,43]. Our model was based on the Canton S strain of Drosophila melanogaster bearing the w1118 mutation (hereafter w), which served as a control genotype in all our experiments (the term “control” is reserved hereafter for uninfected situations, i.e., control w means uninfected w larvae). On average, 42% of these w host larvae succeeded to destroy the wasp egg and 38% survived to adulthood while 42% parasitoids developed to adult wasps (Fig 1C).
Parasitoid-infected third-instar larvae experienced a 15% developmental delay, pupating on average 7 h later than uninfected controls (Fig 1D). Such a delay might result from redistribution of energy from development towards immune defense. We therefore examined various energy aspects during infection.
Without infection, circulating glucose was kept below 0.04 μg per μg protein (Fig 2A). Both the glycogen and triacylglycerol (TAG) stores kept increasing, while circulating and tissue trehalose levels remained steady (Fig 2A). Trehalose is a nonreducing disaccharide source of glucose, which is liberated by the action of trehalase [44]. To trace the fate of glucose, we employed dietary radiolabeled D[U-14C]-glucose. The glucose-derived 14C became evenly distributed in the larvae among saccharides, proteins, and lipids (Fig 2B). About 84% of 14C was found in developing tissues (Fig 2C). We divided the organism here in a simplified way into the immune system (represented by cellular immunity, the most important defense against parasitoids, including circulating hemocytes and lymph gland), the circulation (hemolymph), and the rest of the tissues representing mainly development, growth, and energy storage.
In infected larvae, the accumulation of TAG and glycogen reserves ceased (Fig 2A). This was accompanied by down-regulation of glycogen synthase (CG6904; FlyBase ID: FBgn0266064) and up-regulation of glycogen phosphorylase expression (CG7254; FlyBase ID: FBgn0004507)(Fig 3A). The amount of tissue trehalose decreased (Fig 2A), and less dietary glucose was incorporated into lipids and proteins (Fig 2B and S2 Fig). These hallmarks of suppressed energy storage and growth were corroborated by reduced incorporation of 14C into developing tissues from 84% in uninfected larvae to 77% at 6 hpi and 63% at 18 hpi (Fig 2C and S2 Fig).
The above effects were associated with hyperglycemia as indicated by elevated hemolymph glucose and 14C at the expense of developing tissues (Fig 2A and 2C). Incorporation of 14C into lipids and proteins (at the whole organism level) was also suppressed during infection (Fig 2B), which was accompanied by down-regulation of specific glycolytic enzyme genes in the fat body (Fig 3B and S3 Fig). The diversion of metabolism from building energy reserves and from fat body glycolysis was thus in agreement with extra 14C in the carbohydrate form and with the increase of circulating glucose and trehalose. Circulating trehalose peaked at 6 hpi (Fig 2A) concomitantly with increased expression of a trehalose transporter in the fat body, the organ where trehalose is produced (Fig 3C).
At the same time, the immune cells changed their behavior during infection in the opposite direction, leading to increased energy consumption. Around one-tenth of 14C is normally allocated to immune cells, leaving almost 90% to the rest of the organism, but immune cells demanded up to one-third of nutrients during immune response (Fig 2C). Expression of several glycolytic genes including lactate dehydrogenase Impl3 (CG10160; FlyBase ID: FBgn0001258) increased both in the circulating hemocytes and in the lymph gland (Figs 3B, S4, and S5). This resembled the glucose-demanding aerobic glycolysis, the Warburg effect, in activated mammalian immune cells. Both the lymph gland and the circulating hemocytes expressed elevated amounts of glucose transporter Glut1 (CG43946; FlyBase ID: FBgn0264574) and trehalose transporter Tret1-1 (CG30035; FlyBase ID: FBgn0050035) mRNAs (Fig 3D). Interestingly, later during infection (12–18 hpi), the circulating hemocytes together with already differentiated lamellocytes strongly increased expression of both Tret1-1 and trehalase (CG9364; FlyBase ID: FBgn0003748) (Fig 3D). This suggests that differentiated immune cells preferentially uptake energy in the form of trehalose, which may be linked to the decline of circulating trehalose after 6 hpi (Fig 2A). These results demonstrate a shift of energy distribution away from storage and growth, first towards circulating glucose and trehalose, and then towards the immune cells (Fig 2).
We have previously shown that e-Ado increases circulating glucose via adenosine receptor (AdoR; CG9753; FlyBase ID: FBgn0039747) signaling [38]. Here, we tested if e-Ado was involved in the observed effects of infection on the metabolic shift. While the circulating glucose increased more than 2-fold during infection in w larvae, this increase was suppressed in adoR (FlyBase ID: FBal0191589) mutant larvae (Fig 4A), indicating that AdoR was indeed necessary for the energy redistribution during infection. Therefore, we compared the number of lamellocytes as a measure of immune response. While w larvae produced 5–6 thousand lamellocytes by 24 hpi, the adoR mutants contained less than a third of this amount (Fig 4B). Yet the adoR mutants were clearly capable of differentiating functional lamellocytes that displayed normal morphology, expressed a lamellocyte-specific MSNF9>GFP marker (FlyBase ID:FBtp0064497), and were capable of encapsulating the wasp egg (Fig 4C and S16 Fig). Therefore, adoR larvae were impaired in efficiency or speed of lamellocyte production, and this corresponded with their reduced resistance against the parasitoid invasion relative to w larvae. Indeed, the adoR mutants were three times less successful at neutralizing the wasp eggs and surviving to adult flies (Fig 4C). Thus, AdoR signaling is crucial for effective immune defense against the parasitoid.
The impaired defense in the adoR mutants was not due to affected recognition of the wasp egg, as the number of plasmatocytes attached to the egg surface within the first few hpi was similar in w and adoR larvae (S7 Fig). Therefore, we tested if shortage of energy could be the problem as suggested by failure to increase circulating sugar levels in adoR larvae (Fig 4A). When we fed these larvae a high-glucose diet (12% instead of the regular 5%), the hemolymph glucose significantly increased even without infection in both w and adoR larvae (Fig 4D). This dietary treatment significantly increased the number of lamellocytes in the infected adoR larvae (Fig 4B), suggesting that it was the lack of energy causing inefficient differentiation of lamellocytes in the absence of AdoR.
Interestingly, adding glucose to the diet did not further increase the level of circulating glucose during infection. In fact, the increase induced by infection was greater than that achieved with dietary glucose (Fig 4D), and consistently the number of lamellocytes in infected w larvae was the same on both diets (Fig 4B). Since the glucose increase induced by the dietary treatment was not as high as the one induced by the infection, the number of lamellocytes in adoR did not reach, even on the high-glucose diet, the levels observed in w (Fig 4B). This suggests that the glucose available in circulation is the limiting factor for the lamellocyte differentiation.
Upon infection, more glucose was retained in the saccharide fraction in the w larvae (Fig 2B), indicating that this glucose was available for energy needs of the immune response and less used for storage and growth. Little (at 6 hpi) or no (18 hpi) such retention was observed in adoR mutants (Fig 5A and S2 Fig), suggesting that storage and/or growth were not suppressed during infection in the absence of AdoR. This notion was supported by the relative distribution of 14C among individual tissues (Fig 5B). The distribution was the same in uninfected w and adoR animals. The incorporation of 14C did not change at 6 hpi in infected adoR (as opposed to w), and the shift from storage and growth (red part) towards immune cells (blue part) was much smaller in infected adoR compared to w at 18 hpi (Fig 5B). Importantly, the comparison of relative distribution of 14C into tissues was allowed by equal total uptake of 14C-glucose from diet in w and adoR larvae (S8 Fig). Interestingly, the comparison of absolute numbers of 14C entering the system also revealed anorexia during infection (lower uptake of 14C; S8 Fig), supporting a common observation during immune responses [45]. This anorexia did not seem to depend on AdoR. Besides the lymph gland with slightly lower 14C in adoR mutants, the tissue distribution of 14C was similar in uninfected w and adoR larvae at both time points (Fig 5B and S10 Fig).
Upon infection, only the brain and imaginal disc complex and fat body of w larvae contained significantly less 14C while hemolymph contained significantly more 14C at 6 hpi (Fig 5B and S9 Fig). While 14C incorporation into brain+discs significantly decreased in w larvae, it did not change in the adoR mutant upon infection (Fig 5B and S9 Fig), demonstrating that the suppression of developmental growth, which occurred during infection, was missing in adoR. This is supported by the measurement of the wing imaginal disc growth. While the growth of discs was significantly delayed in w control upon infection, the delay did not occur in the adoR mutant (Fig 5C). Similarly, the delay in development observed in infected w (as measured by pupation rate) did not occur in adoR, which pupated as there would be no infection (Fig 5D).
At 18 hpi, all tissues were affected by infection, significantly increasing 14C in immune cells and hemolymph and decreasing in the rest (Fig 5B and S9 Fig). In all cases but gut, the changes were significantly smaller in adoR than in w (S10 Fig), indicating that the AdoR signaling was involved in the overall suppression of nonimmune processes.
The missing suppression of development in adoR larvae resulted in shortage of energy available for the immune system as documented first by almost no increase of 14C in the hemolymph at 6 hpi and then by much lower 14C incorporation into the immune cells at 18 hpi compared to infected w larvae (Fig 5B). Weak suppression of nonimmune processes in the absence of AdoR may also be linked to the missing peak of circulating trehalose at 6 hpi (Fig 5E and S2 Fig). Functional AdoR signaling seems to lower glucose transport and to increase trehalose transport in the fat body (suggested by expression levels of the respective transporter genes; S11 Fig), leading to increased trehalose at 6 hpi. The trehalose peak probably serves as a reservoir for fast glucose production, which will be increasingly needed for immune defense. The rapid lamellocyte differentiation is lagging in adoR larvae, likely reflecting lower consumption of trehalose relative to w larvae (Fig 5E).
The AdoR signaling reallocates energy towards immune defense, suggesting that e-Ado is released upon immune challenge. Therefore, we next wanted to determine the source of e-Ado during wasp invasion. We individually knocked down the Equilibrative nucleoside transporters, ENT1 (CG11907; FlyBase ID: FBgn0031250) and ENT2 (CG31911; FlyBase ID: FBgn0263916), which are expressed in Drosophila larvae [36,46]. We delivered RNAi to various tissues utilizing the Gal4-UAS system [47], and as a simple readout we used lamellocyte count at 24 hpi (S12 Fig). Among the tested combinations, only ENT2 knockdown driven by Srp-Gal4 (FlyBase ID: FBtp0020112) in cells of the hematopoietic lineage achieved a reduction in the number of lamellocytes that resembled the effect of adoR deficiency (Figs 4B, 6A, and S12). Srp-Gal4 was expressed in all hematopoietic cells, including the circulating hemocytes and all cells of the lymph gland that also contained precursors of lamellocytes (S13 Fig). In contrast, knocking down ENT2 in already differentiated hemocytes (by Hml-Gal4 and Upd3-Gal4 drivers; FlyBase ID: FBtp0040877 and FBtp0020110) did not affect the lamellocyte number (S12 Fig).
ENT2 mRNA was abundant in the lymph gland and brain but weakly expressed in circulating hemocytes and virtually undetected in the fat body (Fig 6B). During infection, ENT2 expression increased in all these tissues except the fat body (Fig 6B) and, consistently, ENT2 RNAi delivered using a fat body-specific C7-Gal4 driver did not affect the number of lamellocytes (S12 Fig). The increasing expression of ENT2 during infection in the brain leaves a possibility that the nervous system contributes e-Ado; however, undetectable expression of Srp-Gal4 in the brain, except for minor signal in some nerve cords (S13 Fig), makes the observed effects of ENT2 removal attributable to the immune cells.
The results above suggest that Ado transport from immune cells, including the differentiating ones, is important for efficient lamellocyte differentiation. As in the case of adoR mutation (Fig 4B), the loss of lamellocytes was rescued by increasing dietary glucose in the Srp>ENT2-RNAi larvae (Fig 6A). Similarly to adoR mutation, ENT2 knockdown in immune cells also cancelled changes in nutrient distribution that normally take place in infected w larvae; there was no peak of circulating trehalose at 6 hpi and no increase in circulating glucose (Fig 6C and S2 Fig). The partition of 14C into saccharides, proteins, and lipids also resembled the pattern seen in adoR mutant larvae (compare Fig 6D with Fig 5A and S2 Fig with S14 Fig).
Together, the above data indicate that deficiency in e-Ado release and in its receptor, AdoR, consistently lead to the same failure of energy reallocation during immune challenge. Indeed, like loss of AdoR, knocking down ENT2 also reduced the host resistance against wasp invasion (Fig 6E), while the normal developmental delay observed in w controls upon infection did not occur in Srp>ENT2-RNAi larvae (Fig 6F). Interestingly, pupation occurred earlier in Srp>ENT2-RNAi compared to w or adoR animals even without infection (Fig 6F); the size of pupae was unaffected implying faster growth instead of precocious pupation of Srp>ENT2-RNAi.
While glycogen storage was suppressed similarly upon infection in adoR mutant and w larvae (Fig 5E), there was no significant difference in glycogen content between infected and uninfected Srp>ENT2-RNAi larvae (Fig 6C). Even more apparent was the effect on lipid storage where the accumulation of TAG in the fat body was suppressed both in w and adoR but not at all in Srp>ENT2-RNAi larvae (Fig 6G). Blocking Ado transport thus led to continued nutrient storage even upon immune challenge, suggesting that energy storage during infection might be regulated by e-Ado independently of AdoR.
An overall metabolic suppression is a common host response to infection [18,12,6]. A likely purpose for the suppression is to conserve energy for the immune response that is energetically costly [2,12]. The defense of the Drosophila larva against the parasitoid wasp requires a rapid production of specialized immune cells (lamellocytes) that encapsulate the parasitoid egg. This has provided us with a unique in vivo model to study the metabolic changes and their regulation during immune response. We show here that the production of lamellocytes is an energetically demanding process, and that a systemic metabolic switch is required for their effective differentiation. This switch includes (1) suppression of energy storage and developmental growth, (2) retaining more energy in circulation, and (3) increased consumption of energy by the immune system (Fig 7).
Suppression of energy storage (glycogen and lipids) and suppression of growth, as documented by slower growth of imaginal discs, lead to a developmental delay. We show here that e-Ado is a signal mediating this metabolic switch. Blocking this signal then demonstrates that the metabolic switch is crucial for an effective immune response. Without this signal, development and growth proceed at a normal speed, thus reducing energy available to the immune cells. Insufficiency of immune cells due to the shortage of energy then leads to a drastically reduced resistance against the parasitoid. Experimental interference with e-Ado or its receptor, AdoR, thus demonstrates the importance of tradeoff between development and immune response, and identifies e-Ado as a signal responsible for the switch.
Blocking Ado transport from immune cells by knocking down the equilibrative nucleoside transporter ENT2 identified the differentiating immune cells as an important source of the signal for the metabolic switch. This suggests that the immune cells could autonomously regulate energy influx based on their acute needs. Ado is a fine sensor of the cellular energy state, as it becomes produced when the ATP:AMP ratio decreases [23]. This scenario is appealing mainly because immune cells dramatically change their metabolism upon activation, leading to increased aerobic glycolysis akin to the Warburg effect [3,4]. Our expression analysis of glycolytic genes, glucose and trehalose transporters, and 14C uptake by immune cells suggested a similar behavior for the differentiating immune cells upon wasp attack. The ability to rapidly react to a metabolic stress could be why ENT2 is strongly expressed in the lymph gland and the brain, both privileged organs from the energy point of view.
AdoR signaling is important for the suppression of developmental growth. Normally, infection leads to lower consumption of energy by the brain and imaginal discs (later also by other tissues), but the consumption continues in adoR-deficient larvae as if they were uninfected. At the same time, AdoR signaling seems to lower glucose transport and to increase trehalose transport in the fat body as inferred from expression levels of the respective transporter genes. The fat body is the site where trehalose is produced from glucose [44]; trehalose is then released back to the hemolymph, and more so during infection. The adoR mutation causes a misbalance of glucose and trehalose transport in the fat body, causing more nutrients to be retained there. The effect of AdoR signaling on the fat body combined with the suppression of developmental growth leads to hyperglycemia that in turn ensures enough energy to supply the immune cells. If the growth suppression fails to occur, as in the adoR mutant, the immune cells are unable to compete with developing tissues that consume the majority of energy. By analogy to the selfish brain theory [48], “selfish” immune cells may usurp energy to themselves by way of AdoR-mediated silencing of nonimmune processes. Our work thus brings experimental evidence and explains the molecular mechanism for recently published theoretical concept of selfish immune system [49].
Interestingly, the AdoR signaling does not mediate the suppression of energy storage (glycogen and TAG) during infection. However, increasing glycogen and TAG stores in infected Srp>ENT2-RNAi larvae with blocked Ado transport from immune cells indicates that the storage suppression is also under e-Ado control but through an AdoR-independent mechanism. Such a mechanism, which needs to be further studied, may involve e-Ado uptake, conversion to AMP by adenosine kinase, and activation of AMPK [25]. The Srp>ENT2-RNAi larvae proceeded faster through development not only during infection but even without infection when compared to control larvae. This suggests that the regulation of energy storage by e-Ado may play a role even during normal development.
e-Ado signaling was previously associated with regulation of hemocyte differentiation, and blocking the AdoR signaling was suggested to lower the differentiation in the lymph gland under noninfectious conditions [50]. The hallmark of lamellocyte differentiation upon parasitoid wasp infection is the turning off the Jak-Stat signaling in the medullary zone of the lymph gland containing the prohemocytes [51]. Expression of cytokine Upd3 (CG33542; FlyBase ID: FBgn0053542) is down-regulated, and the ratio of Jak-Stat receptor Domeless (CG14226; FlyBase ID: FBgn0043903) and its negative coreceptor Latran (CG14225; FlyBase ID: FBgn0031055) is switched upon wasp infection leading to turning off the Jak-Stat and to induction of lamellocyte differentiation [52]. The expression patterns of Upd3, Domeless and Latran mRNAs normally and during infection are unaffected both in adoR and Srp>ENT2-RNAi (S15 Fig), indicating that the induction of lamellocyte differentiation is functional in these lines. In addition, the lymph glands develop normally in both adoR and Srp>ENT2-RNAi ([50] and S17 Fig). Our results demonstrate that the adoR and Srp>ENT2-RNAi larvae are capable of lamellocyte differentiation; they are just less effective, and the reason is most likely the lack of energy as indicated by the rescue of this phenotype with extra dietary glucose.
An important part of the global energy switch observed upon parasitoid invasion is the AdoR-mediated suppression of developmental growth. Although AdoR is relatively strongly expressed in imaginal discs [34], we do not know if it is the tissue-autonomous signaling of AdoR, or whether AdoR acts systemically on metabolism as AdoR is also strongly expressed in the larval endocrine glands and brain; both scenarios may apply simultaneously. It is known that the activation of adenosine receptor leads to metabolic suppression—at the individual cell level, the activation can inhibit growth of tumor cells [26], but it can also cause a systemic suppression during anoxia [28,29] or torpor [27]. Our work demonstrates that the AdoR-mediated suppression plays an important role also during immune response. It will be important to identify the target cells and signaling cascades mediating the observed suppression in future studies.
We show here that the metabolic switch is mediated by e-Ado and that the switch is crucial for an effective immune response. It is of interest to see if this e-Ado role is common to other organisms including humans. e-Ado plays the same role in energy regulation in flies and mammalian systems [30,38]. For example, sepsis is associated with hyperglycemia and insulin resistance as well as with increased e-Ado [31,53], suggesting that e-Ado could indeed mediate the systemic metabolic switch in higher organisms. However, analyzing this role of e-Ado in mammals will be complicated by the existence of multiple adenosine receptors with partly contradicting functions [54,55] and by diverse roles of e-Ado in immunomodulation [24,56,57].
In conclusion, our study demonstrates that extracellular adenosine, released from immune cells, mediates a systemic metabolic switch leading to suppression of energy storage and developmental growth, thus leaving more energy to the immune cells. This switch is crucial for the effective immune response and blocking adenosine signaling drastically reduces host resistance to the pathogen. This may resemble a selfish brain theory in a way that the immune system, like the brain, is a privileged part of the organism, capable of suppressing energy consumption by other tissues in its own interest. Such a selfish immune system [49] would use e-Ado as a signal to appropriate extra energy resources during immune challenge.
All strains were backcrossed at least ten times to w1118 genetic background; w1118 was used as a control in all experiments. adoR mutant was homozygous for adoR1 mutation (FBal0191589). RNAi lines originated from VDRC: UAS-Ent1-RNAi (ID 109885) and UAS-Ent2-RNAi (ID 100464). SrpD-Gal4, Upd3-Gal4, and MSNF9-GFP were obtained from Michele Crozatier, HmlΔ-Gal4 from Bruno Lemaitre and C7-Gal4 from Marek Jindra. Flies were grown on cornmeal medium (8% cornmeal, 5% glucose, 4% yeast, 1% agar) at 25°C. For dietary treatment, larvae were transferred upon infection to cornmeal diet with 12% instead of 5% glucose. Early 3rd instar larvae were infected by parasitoid wasp L. boulardi. Weak infection (1–2 eggs per larva) was used for resistance and pupation analysis; strong infection (4–7 eggs per larva) was used in all other cases.
To determine pupation rate and resistance to parasitoids, infected and control larvae were placed into fresh vials (1 experiment = 30 larvae per vial, 3 vials per genotype; 4 independent experiments). Pupation rate was determined by counting newly appeared pupae every 6 h and incremental percentage of number of pupae per total number of infected and control larvae at a particular time point postinfection was plotted; Log-rank survival analysis was used for comparison. For resistance, we first dissected 20 larvae per experiment from each genotype to count fully melanized wasp eggs (winning host) or surviving wasp larvae (winning parasitoid). Second, we counted all emerged adult flies as surviving the infection and flies without any egg (i.e., uninfected individuals) were excluded from the total number in the experiment. Adult wasps emerged from the vial were counted as adult parasitoid winners.
Expression was analyzed by quantitative real-time PCR. Samples were collected from three independent infection experiments with three technical replicates for each experiment. Expression was normalized to Ribosomal protein Rp49.
Larvae were fed either 73 h AEL or 91 h AEL for 20 min a diet containing D[U-14C]-glucose (10.6 Gbq/mmol; Amersham Biosciences) in yeast. Samples were collected 5 h later. Each sample contained tissues from 30 larvae—all hemolymph was collected by ripping larvae in PBS, centrifuging them, and dividing them into pelleted hemocytes and hemolymph fractions; brains with attached discs and wing discs, whole guts, whole fat bodies, and lymph glands were separated by dissection, and the rest were used as carcass. Macromolecular fractions were separated from tissue homogenates according to [58] for saccharides and lipids and by TCA treatment for proteins. Part of the homogenate was used for measurement of total absorbed amount of 14C molecules. Number of 14C disintegrations per minute was detected by liquid scintillator.
Glucose, trehalose, and glycogen were measured as described [59], using GAGO-20 kit (Sigma). Lipids extracted with chlorophorm:methanol were quantified by HPLC and mass spectrometry.
Wing discs were dissected from larvae at 90 h AEL (18 hpi), and their size was determined from micrographs by FIJI software.
Data were analyzed by GraphPad Prism 6 (GraphPad Software, Inc.).
Extended Materials and Methods are available in S1 Text.
|
10.1371/journal.pgen.1000969 | Affecting Rhomboid-3 Function Causes a Dilated Heart in Adult Drosophila | Drosophila is a well recognized model of several human diseases, and recent investigations have demonstrated that Drosophila can be used as a model of human heart failure. Previously, we described that optical coherence tomography (OCT) can be used to rapidly examine the cardiac function in adult, awake flies. This technique provides images that are similar to echocardiography in humans, and therefore we postulated that this approach could be combined with the vast resources that are available in the fly community to identify new mutants that have abnormal heart function, a hallmark of certain cardiovascular diseases. Using OCT to examine the cardiac function in adult Drosophila from a set of molecularly-defined genomic deficiencies from the DrosDel and Exelixis collections, we identified an abnormally enlarged cardiac chamber in a series of deficiency mutants spanning the rhomboid 3 locus. Rhomboid 3 is a member of a highly conserved family of intramembrane serine proteases and processes Spitz, an epidermal growth factor (EGF)–like ligand. Using multiple approaches based on the examination of deficiency stocks, a series of mutants in the rhomboid-Spitz–EGF receptor pathway, and cardiac-specific transgenic rescue or dominant-negative repression of EGFR, we demonstrate that rhomboid 3 mediated activation of the EGF receptor pathway is necessary for proper adult cardiac function. The importance of EGF receptor signaling in the adult Drosophila heart underscores the concept that evolutionarily conserved signaling mechanisms are required to maintain normal myocardial function. Interestingly, prior work showing the inhibition of ErbB2, a member of the EGF receptor family, in transgenic knock-out mice or individuals that received herceptin chemotherapy is associated with the development of dilated cardiomyopathy. Our results, in conjunction with the demonstration that altered ErbB2 signaling underlies certain forms of mammalian cardiomyopathy, suggest that an evolutionarily conserved signaling mechanism may be necessary to maintain post-developmental cardiac function.
| Heart failure is a common cardiovascular disease that is characterized by problems with the ability of the heart muscle to contract, called impaired systolic function, and an enlarged heart chamber. Model systems of human heart failure are necessary to facilitate the screening and identification of genes and genetic variations that either cause or influence the development and progression of this disease. To better understand these genes, we conducted a genetic screen employing molecularly defined deletions throughout the genome of the adult fruit fly, Drosophila melanogaster. We used an optical coherence tomography imaging technique that provided images similar to echocardiography in humans to measure the cardiac function in adult flies. We identified mutants in members of the rhomboid protease family and epidermal growth factor receptor that cause an enlarged cardiac chamber. Interestingly, abnormalities in the function of members of the epidermal growth factor receptor family in humans that undergo certain chemotherapies are associated with the development of dilated cardiomyopathy and heart failure. Our results suggest that epidermal growth factor receptor signaling may be an evolutionarily conserved pathway that is necessary to maintain normal adult cardiac function.
| The identification of genes that cause or modify cardiac dysfunction is required to understand the complex biology that is responsible for cardiomyopathies and heart failure in humans. The screening and identification of genetic mutants that affect cardiac function are facilitated by model systems of human heart failure. Drosophila has been used as a model of several human diseases and recent investigations have demonstrated that Drosophila can be used as a model of human heart failure, defined as an abnormally enlarged cardiac chamber when the heart is fully relaxed at end-diastole and an impaired systolic function [1]–[5]. Previously, we described that optical coherence tomography (OCT) can be used to examine the cardiac function in adult, awake flies [5]. OCT is a non-destructive, non-invasive imaging modality based on reflectivity of near-infrared light and provides detailed functional imaging of the adult fly heart in a manner similar to echocardiography in humans.
To identify new gene mutations that cause an enlarged cardiac chamber, we employed OCT to examine the cardiac chamber size in adult Drosophila from a set of the DrosDel and Exelixis collections that have molecularly-defined genomic deficiencies [6], [7]. During the course of our genetic screen, we identified abnormalities in cardiac chamber dimensions in a series of deficiency mutants spanning the rhomboid 3 locus. Drosophila rhomboid 3 (rho3), also known as roughoid/ru, is a member of a highly conserved family of intramembrane serine proteases and processes Spitz, an epidermal growth factor (EGF)-like ligand [8]–[14]. Initially, isolated in a genetic screen of embryonic developmental defects, rhomboids are essential for proper EGF receptor (EGFR) signaling in Drosophila. In fact, rhomboid and EGFR signaling is necessary for the development of the fly eye, trachea, and embryonic somatic musculature and many other places in the fly [7], [8], [15]–[17].
Spitz and EGFR signaling are required for the specification and diversification of Drosophila embryonic muscle progenitors [16]. During development, individual muscle groups are differentially sensitive to the level of EGFR signaling that results from the spatial restriction of Spitz and other ligands [14]. Rhomboids have been implicated in this process and rhomboid-1 (rho) has been shown to be required for dorsal acute 1 (DA1) muscle formation in the embryo [16].
The embryonic dorsal vessel that becomes the adult fly heart develops through an orchestrated series of spatially restricted mesodermal and ectodermal signals involving dpp, Wnt, and hedgehog [18]–[22]. During morphogenesis the adult heart, also known as the conical chamber, arises from the embryonic aorta while the embryonic dorsal vessel degenerates to become the terminal chamber of the adult circulatory system [23], [24]. Recently, Perrin et al. have begun to elucidate the signaling pathways in the transition from the fly embryonic to adult heart through an examination of gene profiling studies [25]. This work has identified roles for FGF, Wnt, and PDGF-VEGF signaling in the morphogenesis of the adult fly heart. However, the signaling pathways that are necessary for maintenance of cardiac function after establishment of the adult Drosophila circulatory system are only recently beginning to be identified [2], [26].
Our results demonstrate that a partial inhibition of rho3 causes an enlargement in the cardiac chamber in adult Drosophila through inhibition of the EGFR pathway. We show that the mutant roughoid-1, designated ru1, encodes a missense mutation in rho3 corresponding to a premature stop codon and has a recessive cardiac phenotype. While the ru1 homozygote embryonic dorsal vessel and pupal heart are not significantly different from w1118, the adult heart in ru1 homozygote is larger than w1118, consistent with the functional measurements obtained by OCT. Furthermore, the restoration of rho3 by cardiac expression of transgenic wild-type rho3 rescues cardiac function in the context of genomic deficiencies for endogenous rho3 or the ru1 mutant.
Since rho3 is important in processing of Spitz and subsequent EGFR signaling, we examined the effects of alteration in Spitz and EGFR on cardiac function in adult Drosophila. Cardiac specific expression of a processed form of activated Spitz rescued the abnormality in cardiac phenotype that was observed in rho3 mutants. Additionally, expression of EGFR also restores cardiac function in the context of genomic deficiencies in rho3 or the ru1 mutant. Finally, a temperature-sensitive, cardiac specific expression of dominant-negative EGFR in adult Drosophila causes a progressive enlargement of the cardiac chamber that is observed by serial OCT measurements. Collectively, these results suggest that deficiencies in rho3 and its downstream components in the EGFR signaling pathway lead to an enlargement of the cardiac chamber in adult Drosophila.
As part of a systematic, genome wide screen in Drosophila to identify candidate gene mutations that cause abnormalities in adult cardiac function, we employed optical coherence tomography to examine the cardiac function in awake, adult Drosophila [5]. We examined flies with molecularly-defined genomic deficiencies along the 3L chromosome from the DrosDel and Exelixis collections and identified a set of mutants that had a dilated heart manifest as enlarged end-diastolic dimension (EDD) and end-systolic dimension (ESD). We also calculated fractional shortening (FS), a parameter that correlates with contractile function and has been used to describe the cardiac function in a variety of models of cardiovascular disease [27]–[32] (Figure S1). All deficiency mutants were bred into a w1118 background to remove possible influences that balancer chromosomes may have on the cardiac phenotype. Therefore, all the genomic deficiency mutants were examined as heterozygotes. In an initial screen of 32 deficiency stocks (Table S1), we identified a markedly enlarged cardiac chamber in mutant Df(3L)ED4238/+ (Figure 1A–1D). The severity of the cardiac chamber enlargement observed in Df(3L)ED4238/+ was indistinguishable from the cardiac dysfunction in the troponin-I mutant, hdp2, that has been previously well-characterized [5]. An examination of the cardiac chamber size in deficiency mutants that surrounded the region corresponding to Df(3L)ED4238/+ revealed three additional genomic deficiencies, Df(3L)ED4196/+, Df(3L)ED207/+, and Df(3L)ED4191/+, that had abnormal cardiac dimensions thereby narrowing the initial genomic interval from ∼800Kb to a candidate region encompassing ∼120 Kb that encoded 9 genes (Figure 2 and Table 1). Since the resolution of OCT is limited to 8 microns, we also examined the data in a dichotomized manner. We defined “normal” diastolic cardiac size as an EDD <90 microns and an “enlarged heart” as an EDD > 90 microns. Additionally, “normal” systolic cardiac dimension was defined as <20 microns and “abnormal” systolic dimension as > 20 microns. The examination of dichotomized data also supported the observation that Df(3L)4238/+, Df(3L)ED4196/+, Df(3L)ED207/+, and Df(3L)ED4191/+ had abnormal cardiac chamber sizes.
To identify the candidate gene that was responsible for the abnormal cardiac phenotype within the region, mutants from the Exelixis P-element insertion collection were used to engineer mutant Drosophila that had molecularly-defined deficiencies spanning the genomic region of interest [6]. The deficiency mutant, f05423-d04958/+, spanned the genomic region that was inferred from the initial screen and had an enlarged cardiac chamber compared to w1118 (EDD 102 +/- 7 microns and ESD 27 +/- 6 microns for f05423-d04958/+ versus EDD of 66 +/- 8 microns, and ESD of <10 microns for w1118) (Table 1). Interestingly, a smaller genomic deficiency, designated d07829-f07223/+, that encompassed the gene encoding rhomboid 3, (herein referred to as rho3) and CG32320 also demonstrated an enlarged cardiac chamber (EDD 105 +/- 9 microns and ESD 35 +/- 7 microns for d07829-f07223/+) while a genomic deficiency, f05432-d01050/+, encoding several genes upstream of rho3 had normal cardiac function (Figure 2 and Table 1). rho3 mRNA levels were significantly decreased (∼25% in d07829-f07223/+ compared to w1118), while no different in f05432-d01050/+ mutant that has a genomic deficiency outside the rho3 locus (Figure 3B). Furthermore, we observed normal cardiac function in the mutant Df(3L)Exel6087/+ that was heterozygous for a deficiency across the rho1 and rho2 loci but did not disrupt rho3 (Table 1). Since the candidate region also contained the predicted gene CG32320, we examine the cardiac phenotype in two mutants that had the insertion of P-elements into the GC32320 genomic regions. Mi{Et1}CG32320MB04181, inserted into the coding region of the 4th predicted exon of CG32320, and Mi{ET1}CG32320MB08823, inserted in the 5′ UTR region of CG32320, had normal cardiac phenotypes (Table 1). Based on these observations, we focused our attention on rho3 as the candidate gene that was responsible for the abnormal cardiac phenotype observed in the deficiency mutants.
Next, we examined available mutants that spanned the candidate region and identified a cardiac abnormality in roughoid (ru1). ru1 was initially isolated by Sturtevant in 1918, has been shown to be a hypomorphic allele that results in a rough eye phenotype, and was subsequently mapped to the rho3 locus [13], [33]. The ru1 mutant had a C to T mutation at cDNA nucleotide position 163 that predicted a premature stop codon in rho3 at amino acid position 55 (Figure S2); however, expression of recombinant Flag-tag fusion protein from cDNA isolated from ru1 flies did not encode a truncated protein in S2 cells (Figure S2). Only homozygote ru1 mutant flies had enlarged cardiac phenotype (EDD 107 +/− 3 microns and ESD 47 +/− 2microns (n = 12) for homozygote ru1 mutants versus EDD 81 +/− 5 microns and ESD <10 microns (n = 12) for heterozygote ru1 mutants) (Figure 3A). The abnormal recessive cardiac phenotype persisted after several rounds of backcrossing into the w1118 genetic background (data not shown).
Interestingly, the initial identification of an enlarged cardiac chamber was made in heterozygous genomic deficiency mutants; however, abnormalities in cardiac chamber size were only found in homozygous ru1 mutants. This observation is consistent with prior studies that identified ru1 as a hypomorphic allele based on abnormalities in fly eye phenotypes [13]. We also examined the cardiac chamber size in rho-3PLLb, a mutant that has a break in the first intron of rhomboid-3 and removes a portion of the rhomboid-3 coding sequence [13]. The rho-3PLLb allele has been described as a null allele, and therefore should lead, when heterozygous, to the same phenotype as the heterozygous genomic deficiencies Df(3L)ED4328/+. We used QRT-PCR to confirm that heterozygotes for rho-3PLLb had a 50% reduction in rho3 mRNA, similar to rho3 mRNA levels in heterozygous Df(3L)ED4238/+, while homozygotes for rho-3PLLb had non-detectable levels of rho3 mRNA (Figure 3B). Mutants homozygous or heterozygous for the rho-3PLLb allele had enlarged cardiac chambers (Figure 3A and Table 2). The ESD in mutants for homozygous rho-3PLLb was slightly larger than that for heterozygous rho-3PLLb but did not achieve statistical significance.
Additionally, we examined the contribution of rho4 and rho6 abnormalities on cardiac function using mutants that were deficient for rho4, rho6 or both rho4 and rho6 (Freeman, M. unpublished data). Mutants that lacked rho4, rho6 or both rho4 and rho6 had normal cardiac function (Table 2). We also examined the potential contribution of rho1 and rho3 by examining the cardiac function in ru1/+, rho7M43/+ flies that were heterozygous for the rho3/ru allele and an amorphic allele of rho1. The ru1/+, rho7M43/+ mutants had normal cardiac function (Table 2), suggesting that cardiac abnormalities observed in the homozygous ru1 mutants were not phenocopied by a combination of defects in rho3 and rho1. These data suggested that rho-3 inactivation is responsible for the observed enlarged heart phenotype; however, we could not definitely rule out the contribution from rho-1.
Rhomboid-3 is expressed in embryos, pupae, and adult tissues including the heart (Table S2). Since alterations in embryonic dorsal vessel development can lead to abnormal cardiac phenotypes in adult Drosophila, the dorsal vessel morphology was evaluated in Drosophila embryos from stage 13 to 16. Transgenic Drosophila that express tinC-GFP in the context of w1118 or ru1 homozygotes demonstrated similar dorsal vessel morphology (data not shown). The heart in transgenic Drosophila that expressed tinC-GFP in the context of w1118 or ru1 homozygotes did not show a significant difference in overall morphology at different stages of pupal development (Figure S3).
Since adult homozygous ru1 mutants had an enlarged cardiac chamber by OCT, the morphology of the adult cardiac chamber in ru1 homozygous mutants was evaluated by histological sectioning. A comparison between w1118 and ru1 homozygous mutants was performed by examining 8 micron serial sections in the posterior portion of the A1 segment (see details in Materials and Methods section). Although the cardiac chamber size in ru1 was larger than w1118, consistent with OCT results, and the cardiac chamber wall thickness in ru1 homozygote mutants appeared thinner compared to w1118, the extra cardiac structures appeared similar between ru1 and w1118 (Figure S4).
Moreover, the actin fibers in the dorsal diaphragm muscle, also known as the ventral longitudinal fibers, and cardiomyocytes in adult Drosophila hearts were examined via using confocal microscopy. Visualization of actin structures by staining with phalloidin demonstrated that the dorsal diaphragm muscle was similar between ru1 homozygotes and w1118. The GFP positive cardiomyocyte nuclei and actin fiber organization were also similar (Figure 4). Overall, the tissue surrounding the adult hearts in dissected specimens was similar in ru1 homozygotes and w1118. The functional abnormalities observed in the ru1 homozygous mutants did not appear to be related to alterations in extra cardiac structures.
In addition, the other cardiac parameters, such as arrhythmia and heart rate, were not significantly different between ru1 homozygotes and w1118 in intact adults or in dissected adult hearts perfused with artificial hemolymph (Figure S5).
To confirm whether rho3 was responsible for the abnormal cardiac phenotype, we engineered transgenic Drosophila that harbored wild-type (wt)-rho3 under the UAS-promoter. In combination with cardiomyocyte specific tinC-Gal4 driver, we examined the effects of cardiac-specific expression of wt-rho3 in the context of Df(3L)ED4238/+ or ru1 homozygotes. The tinC-mediated expression of wt-rho3 rescued the abnormal enlarged cardiac phenotype (Figure 5). Furthermore, the enlarged cardiac chamber phenotype that was observed in other genomic deficiencies spanning rho3 [Df(3L)ED4191/+ or Df(3L)ED4196/+] was also rescued by wt-rho3 (data not shown).
Since rho3 is known to process an inactive membrane-bound Spitz (mSpi) to an active soluble Spitz (sSpi) that can subsequently interact with EGFR, we examined the effects of these components on cardiac function in flies that had enlarged cardiac chambers attributed to genomic deficiencies [10]. The cardiac expression of sSpi, but not mSpi, rescued the abnormal cardiac phenotype observed in the Df(3L)ED4238/+ mutant (Figure 5). Furthermore, the cardiac-specific expression of EGFR in the context Df(3L)ED4238/+ also restored normal cardiac function. We also tried to examine the ability of constitutively active forms of EGFR to rescue the abnormal cardiac phenotype observed in the genomic deficiency mutants [34], [35]. The cardiac-specific expression of EGFRAct(2) resulted in a failure of flies to eclose from the late pupal state (data not shown).
Next, we examined the effects of cardiac-specific expression of mSpi, sSpi, and EGFR in the context of ru1 homozygous mutant [10], [36], [37]. The cardiac-specific expression of sSpi or EGFR by tinC-Gal4 in the context of ru1 homozygotes restored normal cardiac function while expression of mSpi by tinC-Gal4 in the context of ru1 homozygotes still had an enlarged cardiac chamber (Figure 5).
Additionally, we examined spi1, a loss of function allele, and spis3547, a P-element insertion mutant. Both mutants are lethal as homozygotes and therefore were examined in the heterozygous state [38]. spi1/+ and spis3547/+ had normal cardiac function (Table 2). Interestingly, spi-SCP2, a homozygous viable hypomorph of Spi, had an abnormal cardiac phenotype (Table 2). Moreover, the homozygous mutant Kerenexc27-7-B that lacks Keren, an EGFR ligand with similarity to Spitz that is involved in border cells in the Drosophila ovary had normal cardiac function [15], [39], [40] (Table 2).
These results suggest that the enlarged cardiac phenotypes observed in Df(3L)ED4238/+ and ru1 homozygotes result from alterations in EGFR signaling since the abnormal cardiac phenotype can be rescued by cardiac-specific expression of soluble Spi. Furthermore, Spi may have a threshold effect on cardiac function since heterozygote loss of function alleles for Spi had a normal cardiac phenotype; but, the homozygous hypomorph, spi-SCP2, had an abnormal cardiac function [15], [38] (Table 2).
Since restoration of EGFR signaling by transgenic expression of sSpi or EGFR rescued the abnormal cardiac phenotypes observed in rho3 deficiencies and the ru1 homozygotes, the effects of inhibiting EGFR signaling in the adult Drosophila heart were evaluated using a well-characterized dominant-negative EGFR (EGFR.DN) [16].
The cardiac function in flies with the hypomorphic allele, EGFR24f, in the context of a temperature sensitive lethal allele of EGFR (EGFRtsla) was evaluated by performing serial OCT cardiac measurements in individual flies at 18°C followed by 24 hours at 25°C [37]. Interestingly, the cardiac function in EGFRf24/tsla flies was similar to w1118 at 18°C; however, the cardiac function deteriorated after 24 hr at 25°C (Figure 6A and 6C).
Next, the cardiac function was assessed by serial OCT measurements in individual flies that harbored UAS-EGFR.DN, tubulin-Gal80ts, and tinC-Gal4 [41]–[43]. Since the measurement of cardiac function by OCT is non-invasive and non-destructive, we serially measured the cardiac function in individual flies by performing OCT, allowing the fly to recover over a 24 hour period, and then repeating OCT assessments throughout the course of the experiments. Prior work in our laboratory demonstrated that the tubulin-Gal80ts; tinC-Gal4 fly driver line results in no significant transgene mRNA level at 18°C and between a 20 and 40 fold expression of transgene mRNA expression at 25°C by 24 to 48 hours [44]. Serial OCT cardiac evaluation in individual flies demonstrated that cardiac function was normal at 18°C but progressively deteriorated at 25°C as manifest by enlarged EDD and ESD (EDD 68 +/- 3 microns and ESD <10 microns (n = 15) at 18°C verses EDD 118 +/- 4 microns and ESD 51 +/- 2 microns (n = 15) at 25°C for 144 hr for P{UAS-EGFR.DN}/P{tubulin-Gal80ts}; P{UAS-EGFR.DN}/P{tinC-Gal4}) (Figure 6B and 6C). The effects of EGFR.DN expression on cardiac chamber size were similar in flies of 7 or 30 days age (data not shown). Interestingly, the EGFR.DN mediated effects on cardiac chamber size were reversible when the EGFR.DN expression was repressed by temperature shift (Figure 6B and 6C). The changes observed in cardiac chamber size and function depended on the expression of specific transgenes since flies that harbor tubulin-Gal80ts; tinC-Gal4 in the absence of UAS-EGFR.DN did not have abnormal cardiac phenotypes under similar temperature conditions (Figure 6C). Furthermore, flies that express unrelated UAS transgenes do not demonstrate alterations under similar temperature shift conditions [2], [44]. These results suggest that inhibition of EGFR signaling in the adult fly heart results in a progressive deterioration in cardiac function and suggests that proper EGFR signaling is required for the maintenance of normal adult heart function.
Our results suggest that rho3-mediated EGFR signaling is responsible for the enlarged cardiac phenotype observed in Df(3L)ED4238/+. Several lines of evidence support a role for rho3 in adult cardiac function. Analyses of the cardiac phenotypes in mutants that have ru1 or rho3pLLB alleles are consistent with previously described effects in other tissues including the fly eye [13]. The ru1 allele acts as a hypomorphic allele since homozygote ru1, but not heterozygote ru1, mutants have an enlarged cardiac chamber. However, homozygote and heterozygote rho3pLLB mutants have enlarged cardiac chambers consistent with the rho3pLLB allele acting as a null allele. The ru1 allele encodes a premature stop codon however a truncated protein product was not detected and therefore the mechanism underlying the hypomorphic nature of the ru1 allele is not clear. Possible explanations include a dosage effect such that a reduction in functional rho3 transcript/product to below a certain threshold results in a phenotype.
Prior work has demonstrated that rho3 interacts with rho1 and these proteases act together in cell signaling processes [14]. Although our results suggest that rho3 is important in the adult Drosophila heart, we recognize that rho1 and rho2 may also contribute to the complex signals that influence post-developmental cardiac function. Interestingly, Df(3L)Exel6087/+ that was heterozygote for a deficiency across the rho1 and rho2 loci but did not disrupt rho3 had normal cardiac function. Additionally, ru1/+, rho7M43/+ flies that were heterozygous for the ru1 allele and an amorphic allele of rho1 also did not appear to have an enlarged cardiac chamber. QRT-PCR data also showed that rho1 mRNA levels were not significantly different in deficiency lines, Df(3L)ED4238/+ or d07829-f07223/+, that had an abnormal cardiac chamber size. Although the abnormal cardiac phenotype observed in strains that have heterozygote deficiencies in rho3 or homozygous ru1 mutants was rescued by tinC-mediated expression of recombinant rho3, we recognize that over-expression of rho3 may compensate for abnormalities in rho1 function and therefore obfuscate the importance of rho1 in adult cardiac function.
Our results also demonstrated that normal cardiac chamber size was restored in Df(3L)ED4238/+ and homozygous ru1 by cardiac-specific expression of components downstream of rho3 including Spitz and EGFR. The restoration of cardiac function by transgenic expression of EGFR can be explained by two possible processes. First, the over-expression of EGFR monomers may increase the probability ligand-independent dimerization of EGFR resulting in potentially higher basal EGFR signaling and a rescue of cardiac function [10], [45]. Second, an unknown EGFR ligand other that Spitz may be responsible for restoration of cardiac function. Other EGFR ligands, including Keren and Gurkin, have been described to activate EGFR in specific tissues including the eye and ovary [39], [46]–[48]. Our results demonstrate that mutants with a null allele for Keren do not have an abnormal cardiac function although Keren was reported to function similarly to Spitz in eye development. However, other Spitz-like ligands may interact with cardiac EGFR and explain the observation that cardiac chamber size is normal in the presence of cardiac-specific EGFR transgene expression in the context of impaired rho3 function. Additionally, the rescue of the abnormal cardiac phenotype by tinC-Gal4 mediated transgenic expression of rho3 itself or active soluble Spitz suggests that these signals function in an autocrine manner or between the cardiomyocytes that comprise the cardiac tube.
Despite the functional abnormality observed by OCT in adult ru1 homozygous mutants, we observed that the embryonic dorsal vessel and pupal heart morphology in ru1 homozygotes did not differ from w1118. The adult homozygous ru1 mutants had similar extra cardiac structures and dorsal diaphragm muscle fibers compared to w1118. Furthermore, the actin fiber organization did not appear obviously different from w1118 although the heart chamber wall appeared thinner as compared to w1118. These findings are strikingly different from other fly mutants that have enlarged cardiac chambers [5], [31], [49]. For example, hdp2, a fly that has a recessive point mutation in troponin-I, has a markedly enlarged heart with poor systolic function and has abnormalities in the organization of contractile elements. Additionally, flies that have deficiencies in dystrophin also demonstrate poor contractile function, enlarged cardiac chambers, and myofibrillar disorganization [31]. Our findings suggest that additional mechanisms may exist that responsible for an enlargement in the Drosophila heart and are not associated with dramatic abnormalities in myofibrillar organization or cardiomyocyte cell number. While the exact mechanisms through which alterations in rhomboid signaling translate into functional myocardial impairment remain to be elucidated, our results support that further studies of the downstream signaling components of EGFR are necessary.
The post-developmental expression of a dominant-negative EGFR under the control of the temperature-sensitive Gal80 system resulted in a progressive deterioration in cardiac function. These results suggest that rho3-EGFR signaling maintains adult cardiac function. Additionally, the developing cardiac system in Drosophila appears to be sensitive to the level of EGFR activation since tinC-mediated expression of constitutively-active EGFR transgenes resulted in the failure of flies to eclose from pupal cases. We also observed the rescue of cardiac chamber size and function post-developmentally when EGFR-DN was repressed in adult flies after the development of an enlarged heart. The findings suggest that the adult fly heart has a degree of plasticity such that the removal of a deleterious cardiac perturbation may restore normal heart camber size and function. This is quite different from most cardiomyopathies in mammals in which a defect in cardiac function often leads to a progressive and irreversible deterioration of the heart. The mechanisms underlying the plasticity of the adult fly heart and the exact downstream components of EGFR signaling that are responsible for maintaining adult cardiac function will require further investigation.
The identification of EGFR signaling in adult Drosophila heart function underscores the concept that evolutionarily conserved signaling mechanisms are required to maintain normal myocardial function. In mammals the EGF receptor is one member of the larger ErbB receptor family. Recently, Alvarado et al. had shown that human ErbB2 is the closest structural relative to Drosophila EGFR and suggest that ErbB2 shares more similarities with invertebrate EGFR than other isoforms of other human ErbB members [50]. ErbB2 is a target of chemotherapy in the treatment of several cancers and the inhibition of ErbB2 signaling by transgenic knock-out in mice and herceptin-treatment humans is associated with the development of dilated cardiomyopathy [51]–[53]. Our findings support that EGFR signaling is required for cardiac function in Drosophila. Furthermore, our results in conjunction with the demonstration that altered ErbB2 signaling underlies certain forms of mammalian cardiomyopathy suggest that an evolutionarily conserved signaling mechanism may be necessary to maintain post-developmental cardiac function.
Rhomboid proteases are highly conserved and present throughout the animal kingdom including mammals [54]. Although rhomboids do not appear be involved in EGF signaling in mammals, prior work has identified ephrin B and thrombomodulin as potential substrates of rhomboids [55], [56]. Our results demonstrate a new role for rhomboid proteases in post-developmental adult Drosophila cardiac function and suggest that further investigations may provide insight into potential mechanisms of mammalian rhomboid proteases in human dilated cardiomyopathies and heart failure.
Our observations also support an important concept that underlies our screening strategy. Namely, a screen for altered cardiac chamber dimensions and function in adult flies based on functional phenotyping can identify mutations that do not significantly alter the morphology of the dorsal vessel in the embryo. We acknowledge that the consequences of cardiac dysfunction in the fly are not the same as in humans; however, aspects of cardiac abnormalities are shared at some level and may help identify pathways that potentially affect human disease. While this concept may be simple, our observations support the rationale of our ongoing screens to identify genes that cause cardiac abnormalities in adult Drosophila.
DrosDel and Exelixis stocks as well as all mutants were obtained from the Bloomington Drosophila Stock Center unless otherwise stated below. All stocks were maintained on standard yeast protein media at room temperature. The p{tinC-Gal4} was kindly proved by Manfred Frasch [22]. The p{UAS-sSpi} and p{UAS-mSpi} stocks were kindly provided by Andreas Bergmann and Hermann Steller [10], [36]. The p{tubulin-Gal80ts}; p{tinC-Gal4} stocks were engineered based on methods described by McGuire et. al. [42]. rho-3PLLb, spi-SCP2, rho-4, rho-6, and the double rho-4;rho-6 mutants were kindly provided by Matthew Freeman [13]. The Kerenexc 27-7-B mutant was kindly provided by Denise Montell [39].
Drosophila containing specific P-element insertions designated P{XP}Ptp61Fd01050, PBac{WH}Ptp61Ff07223, P{XP}Ptp61Fd07829, and PBac{WH}msd5f05423 were obtained from the Exelixis Collection at Harvard Medical School. Specific genomic deficiencies were generated using P{hsFLP}1, y[1] w[1118]; Dr[Mio]/TM3, ry[*] Sb[1] according to the established methods [6]. The genomic deficiencies were genotyped using iPCR methods and DNA sequencing as described by Parks et. al. [6].
Cardiac function in adult Drosophila was measured using a custom built OCT microscopy system (Bioptigen, Inc. Durham, NC) as previously described [5]. Flies that had genomic deficiencies from the Exelixis or DrosDel collections were bred with w1118 to remove balancer chromosomes and the F1 generation was used to examine cardiac parameters. Additionally, all stocks that had alleles maintained over a balancer chromosomes were bred to w1118 to remove the balancer chromosome prior to evaluation of the cardiac parameters in the F1 offspring.
Briefly, adult female Drosophila between 7 and 10 days post eclosion were briefly subjected to CO2, placed on a soft gel support, and allowed to fully awaken based on body movement. All flies were imaged as B-modes in the longitudinal orientation to identify the cardiac chamber in the A1 segment and then in the transverse orientation to center the heart chamber. Multiple 3 second OCT m-modes were recorded and images were processed using ImageJ software using a 125 micron standard. After m-mode acquisition, the flies were examined in the transverse B-mode orientation to assure consistent measurements from the heart chamber. End-diastolic (EDD), end-systolic (ESD), and heart rate were determined from 3 consecutive heart beats. Fractional shortening (FS) was calculated as [EDD-ESD]/EDD ×100. OCT measurements were binned into 8 micron cut-off values based on the axial resolution of the OCT instrument. Each OCT measurement is represented in the graphs as a closed circle to provide an estimate of the frequency observed for each measurement in the corresponding groups for the experiments described in the text. A summary of the OCT data for each group represented by an open circle and corresponds to the mean +/- SE. Since the resolution of OCT is limited to 8 microns, we also analyzed the data in a dichotomized manner where we defined an “enlarged” heart as an EDD >90 microns and “impaired systolic function” as an ESD >20 microns.
For serial OCT studies, adult flies were bred at 18°C until 7 to 10 days post eclosion prior to initial evaluation of cardiac function by OCT. After each OCT measurement, individual files were gently removed from the soft gel and placed in individual vials at 25°C with food. Serial evaluation of cardiac function by OCT was conducted as described in the figures.
Total RNA samples from ten female flies of 7 days post-eclosion were prepared from w1118, Df(3L)ED4328/+, heterozygous genomic deficiencies d07829-f07223/+ or f05432-d01050/+, ru1 homozygous mutants (ru1/ru1), rho3pLLB heterozygotes (rho3pLLB/+), or rho3pLLB homozygotes (rho3pLLB/rho3pLLB) using RNA-Bee (Tel-Test “B”). Embryos, larvae, pupae, and dissected adult tissues were collected from w1118 to define the transcriptional expression pattern of rho1 and rho3. Two µg of RNA was used for generation of cDNA using SuperScript II reverse transcriptase (Invitrogen, Inc.). Applied Biosystems Taqman Gene expression assays were used to perform quantitative (real time) RT-PCR (rho3: Dm01837284_m1; rho1: Dm01821932_m1, and Ribosomal Protein L32 (Rpl32): Dm 02151827-g1 for endogenous control). The following reaction components were used for each probe: 2 µL cDNA, 12.5 µl 2X TaqMan Universal PCRMaster Mix without Amperase (Applied Biosystems, Inc.), 1.25 µl of probe, and 9.25 µl water in a 25 µl total volume. Reactions were amplified and analyzed in triplicate using an ABI PRISMH 7000 Sequence Detection System. PCR reaction conditions were as follows: Step 1: 95°C for 10 minutes, Step 2: 40 cycles of 95°C for 15 seconds followed by 60°C for 1 minute. Expression relative to Rpl32 was calculated using 2−ΔΔCt and levels were normalized to baseline. We performed three independent experiments in triplicate using different batches of flies each time.
The cDNA encoding wt-rho3 was isolated by RT-PCR from w1118 adult flies, subcloned into pUAST, and verified by sequencing. Transgenic Drosophila harboring wt-rho3 were generated by established methods [57]. The tinC-GFP Drosophila lines were created by isolating and subcloning the 304 bp tinC genomic sequence from w1118 into pGreen-H-Pelican [22], [58].
S2 cells were maintained in Schneider's media containing 10% fetal bovine serum under standard conditions. Total RNA was isolated from adult w1118, Oregon-R, and homozygous ru1 mutant flies using RNAzole-Bee (Tel-Test “B”) and then used to obtain cDNA by a reverse transcriptase mediated reaction using oligo-d(T)16 primers. Next, rho3 cDNA was isolated by PCR with primers corresponding to the 5′ and 3′ ends of rho3 and Platinum-Taq (Invitrogen Inc.). The PCR products were subcloned into pUAST with 5′ primers containing flag-tag epitopes. DNA sequencing was performed at multiple steps throughout the subcloning steps.
S2 cells were transfected with 1 µg of pUAST alone or pUAST containing N-terminally flag-tagged rho3 cloned from w1118, Oregon-R, and ru1 mutant flies, respectively, in the presence of 1 ug of pPTGAL plasmid encoding Gal4 driven by an ubiquitin promoter (gift from H. Amerin) using Cellfectin reagent (Invitrogen, Inc.). After 48 hours of incubation, the cells were washed gently in PBS, lysed in 500 µl of RIPA buffer, placed on ice for 20 minutes, and then subjected to centrifugation at 12,000×g for 15 minutes. Total protein content in each lysate was determined using a Bio-Rad protein assay. 20 µg of total protein was applied to 12% polyacrylamide denaturing gels and proteins were resolved by electrophoresis prior to Immunoblotting. The mouse monoclonal anti-flag M2 antibody (1∶2000) (Sigma, Inc.) was used to detect Flag-fusion proteins.
Adult Drosophila corresponding to homozygous tinC-GFP in w1118 background alone or in the presence of homozygous ru1 at 7 days age post eclosion were used to examine adult cardiac morphology. Flies were briefly anesthetized by administration of CO2, the head and thorax were removed and the abdomen was placed in artificial hemolymph buffer (108 mM Na+, 5 mM K+, 2 mM Ca2+, 8 mM MgCl2, 1 mM NaH2PO4, 4 mM NaHCO3, 10 mM sucrose, 5 mM trehalose, and 5 mM Hepes (pH 7.1) [59]. An incision was made along the ventral aspect of the abdomen and the internal abdominal organs were gently removed. The surrounding fat and tissue was removed using a pulled glass capillary pipette. Then hemolymph buffer that contained 10mM EGTA was added to relax the cardiac muscle as described by Alayari et. al. [60]. Next, samples were fixed in 4% paraformaldehyde for 20 minutes at room temperature prior to staining with anti-GFP-antibody (1∶500) (Invitrogen, Inc.) for detection of cardiomyocytes and phalloidin-TexasRed (1∶1,000) (Invitrogen, Inc.) for actin staining. The stained heart preparations were visualized under a Zeiss LSM510 confocal microscope and 0.4 micron Z-stack images were analyzed.
For evaluation of cardiac morphology during pupal stages, Drosophila corresponding to homozygous tinC-GFP in w1118 background or in the presence of homozygous ru1 were collected between the ∼P6 and P13 stages according to staging described by Bainbridge and Bownes [61]. Pupal hearts were directly visualized using a Leica M165FC fluorescent stereomicroscope.
For heart rate measurements, whole flies corresponding to homozygous tinC-GFP in w1118 background or in the presence of homozygous ru1were examined intact or after dissection as described above and the hearts were visualized using a Leica M165FC fluorescent stereomicroscope equipped with a DFC310Fx camera. Images were captures at a frame rate of ∼38 fps for analysis.
Adult female flies of 7 to 10 days age were collected, immersed in 70% alcohol for 1 minute, and then fixed in 10% buffered formalin overnight at 4°C. Next, the specimens were rinsed in PBS and dehydrated in ethanol through sequential gradients. Then, the samples were washed twice with xylenes before immersion in liquid paraffin. After solidification, paraffin blocks were sectioned serially at 8 µm thickness in longitudinal or transverse orientation (n = 8 flies per orientation per group). Sections were rehydrated and stained with hematoxylin and eosin. The following criteria were used to control for the position of the heart chamber among different flies that were evaluated. First, we identified the transverse section with that contained the en face view of the ventral longitudinal fibers (VLF) corresponding to the anterior conical chamber. Second, we counted three serial 8 micron sections posterior from the section with the en face view of the VLF. Third, we measured the chamber dimension in the next three serial 8 micron sections (denoted MS1, MS2, and MS3 in Figure S4). Sections were analyzed using an Olympus 1X70 microscope at 400× magnification and images captured using a PaxCam digital camera. We measured the diameter of the cardiac chamber in the dorsal to ventral direction in transverse oriented sections since this measurement corresponded to the orientation of the measurements obtained by OCT. Wall thickness was calculated by measuring the cardiac chamber wall width along the lateral walls at two positions in three serial sections to obtain the mean ± SEM.
To determine the degree that fixation may have had on chamber size, we examined the chamber size in flies that were fixed after the 70% alcohol step by performing measurements using OCT. The cardiac chamber sizes in fixed flies were similar to the EDD in awake, adult flies.
Comparisons of EDD chamber dimensions were determined by either a student's t-test for two samples or an analysis of variances (ANOVA) with Bonferroni corrections for multiple comparisons when necessary. Comparisons of ESD chamber dimensions were determined by Mann-Whitney tests for comparisons of two samples or Kruskal-Wallis tests with a Dunn's test for multiple comparisons. GraphPad Prism statistical software (GraphPad Software Inc.) was used for all analyses.
|
10.1371/journal.ppat.1002184 | Activation of HIV Transcription by the Viral Tat Protein Requires a Demethylation Step Mediated by Lysine-specific Demethylase 1 (LSD1/KDM1) | The essential transactivator function of the HIV Tat protein is regulated by multiple posttranslational modifications. Although individual modifications are well characterized, their crosstalk and dynamics of occurrence during the HIV transcription cycle remain unclear.
We examine interactions between two critical modifications within the RNA-binding domain of Tat: monomethylation of lysine 51 (K51) mediated by Set7/9/KMT7, an early event in the Tat transactivation cycle that strengthens the interaction of Tat with TAR RNA, and acetylation of lysine 50 (K50) mediated by p300/KAT3B, a later process that dissociates the complex formed by Tat, TAR RNA and the cyclin T1 subunit of the positive transcription elongation factor b (P-TEFb). We find K51 monomethylation inhibited in synthetic Tat peptides carrying an acetyl group at K50 while acetylation can occur in methylated peptides, albeit at a reduced rate. To examine whether Tat is subject to sequential monomethylation and acetylation in cells, we performed mass spectrometry on immunoprecipitated Tat proteins and generated new modification-specific Tat antibodies against monomethylated/acetylated Tat. No bimodified Tat protein was detected in cells pointing to a demethylation step during the Tat transactivation cycle. We identify lysine-specific demethylase 1 (LSD1/KDM1) as a Tat K51-specific demethylase, which is required for the activation of HIV transcription in latently infected T cells. LSD1/KDM1 and its cofactor CoREST associates with the HIV promoter in vivo and activate Tat transcriptional activity in a K51-dependent manner. In addition, small hairpin RNAs directed against LSD1/KDM1 or inhibition of its activity with the monoamine oxidase inhibitor phenelzine suppresses the activation of HIV transcription in latently infected T cells.
Our data support the model that a LSD1/KDM1/CoREST complex, normally known as a transcriptional suppressor, acts as a novel activator of HIV transcription through demethylation of K51 in Tat. Small molecule inhibitors of LSD1/KDM1 show therapeutic promise by enforcing HIV latency in infected T cells.
| One of the remaining questions in HIV research is how the virus establishes a dormant (latent) stage and thereby escapes eradication by current antiretroviral therapy. Latently infected T cells do not produce significant amounts of viral genomes or viral proteins due to the silencing of a specific step in the viral life cycle called transcription. Viral transcription can be reactivated in latently infected cells, a process that rekindles HIV infection after antiretroviral therapy is discontinued. A key regulator of viral transcription is the viral Tat protein. Here we identify a novel cellular enzyme that regulates HIV transcription through the modification of the Tat protein. This enzyme, LSD1, is generally known as a transcriptional suppressor. In HIV infection, however, it acts as a transcriptional activator because downregulation of LSD1 expression or inhibition of its enzymatic activity suppresses reactivation of HIV from latency. Our findings provide novel insight into the mechanisms of HIV latency and identify a potential new strategy that may help to keep HIV dormant in latently infected cells.
| Epigenetic processes are critical in the regulation of gene expression from the integrated HIV provirus and have become a focal point of research in therapeutics for HIV latency. Latently infected T cells persist in HIV-infected individuals despite highly active antiretroviral therapy (HAART) and rekindle the infection when HAART is discontinued [1], [2]. In the majority of latently infected cells, HIV infection is blocked at the transcriptional level. Therapeutic efforts are aimed at permanently silencing HIV gene expression in latently infected cells or at “flushing out” the viral reservoirs by reverting the transcriptional silencing that lies at the core of HIV proviral latency.
Known epigenetic processes involved in the regulation of HIV gene expression include DNA methylation [3], [4], chromatin remodeling events [5], [6], [7], posttranslational modifications of histones [8], [9] and posttranslational modifications of the HIV Tat protein [10], [11], [12], [13], [14], [15], [16]. Tat is an essential viral gene product that potently activates HIV gene expression through its unique interactions with the TAR element located at the 5′ ends of nascent viral transcripts and the cellular positive transcription elongation factor b (P-TEFb) [17], [18]. Two Tat species naturally exist in HIV-infected cells: a full-length Tat protein of ∼101 aa length encoded by both tat exons and a shorter splice variant of 72 aa length encoded by the first tat exon. Both Tat forms are transcriptionally active and form a trimolecular complex with the cyclin T1 subunit of P-TEFb and TAR RNA to recruit the kinase activity of CDK9 to elongating HIV transcripts. The bulk of Tat is produced after successful integration of the provirus into the host genome where it activates its own production via a feed-forward mechanism [19].
Several posttranslational modifications of Tat have been identified that modulate the interactions of Tat with P-TEFb and TAR RNA [20] (see Table S1). Two of these modifications, acetylation of K50 and monomethylation of K51, occur at adjacent residues within the arginine-rich motif (ARM) in Tat, a region involved in TAR RNA binding, nuclear localization and protein stability [21]. K50 is the preferred target for the acetyltransferase activity of p300/KAT3B in Tat while K51 is monomethylated by the lysine methyltransferase Set7/9/KMT7 [10], [11], [16], [22]. K50 and K51 are also targets of the acetyltransferase activity of hGCN5/KAT2A and the di- or trimethyltransferase activity of SETDB1/KMT1 [15], [22].
K50 acetylation and K51 monomethylation have both important positive regulatory functions in Tat transactivation. Monomethylation of K51 strengthens the interactions of Tat with P-TEFb and TAR RNA while acetylation of K50 dissociates the Tat/TAR/P-TEFb complex and recruits the PCAF/KAT2B histone acetyltransferase to the elongating RNA polymerase II complex [16], [23], [24], [25]. These findings form the basis for a dynamic view of the Tat transactivation cycle in which changes in the modification status of Tat occur sequentially and govern differential cofactor interactions of a single Tat molecule during HIV transcription [26], [27].
We were intrigued by the close proximity of the two modifications in Tat (K50 acetylation and K51 methylation) and speculated that a bimodified protein may exist in cells. Similar studies were previously performed with the p53 tumor suppressor protein and supported the model that lysines in close proximity to each other are sequentially methylated and acetylated [28], [29]. However, we did not detect bimodified Tat in cells using mass spectrometry or newly generated antibodies specific for monomethylated/acetylated Tat. Instead, we identified LSD1/KDM1 as a Tat demethylase and an unexpected new transcriptional coactivator required for activation of HIV gene expression in latently infected T cells.
To examine how acetylation of K50 affects monomethylation of the neighboring K51 residue, we incubated short synthetic Tat peptides (aa 48–58) carrying an acetylated lysine at position 50 with recombinant Set7/9/KMT7 enzyme and radiolabeled S-adenosyl-L-methionine (SAM). Reactions were dissolved on a high percentage Tris-Tricine gel and examined by autoradiography. Acetylation at K50 completely suppressed methylation of the peptide by Set7/9/KMT7 (Figure 1A). The same was observed when a K51-monomethylated peptide was tested in the reaction indicating that K50 is not a target of the Set7/9/KMT7 monomethyltransferase activity (Figure 1A). We also performed the inverse experiment and incubated a Tat peptide carrying a monomethyl group at position 51 with the K50 acetyltransferase p300/KAT3B and observed that acetylation can proceed, albeit with a 40% decrease in efficiency as compared to an unmodified peptide (Figure 1B). Interestingly, a K50-acetylated peptide was further acetylated by p300/KAT3B confirming previous results that K51 also functions as a target of the p300/KAT3B acetyltransferase activity, especially when K50 is not available [10], [11], [30]. Similar results were observed when the reactions were performed with full-length synthetic Tat proteins (aa 1–72) carrying acetylated K50 or monomethylated K51 residues (Figure S1). These data demonstrate that in vitro monomethylation cannot occur efficiently on an acetylated Tat substrate supporting previous data that point to a role of K51 monomethylation early in the Tat transactivation cycle before acetylation of K50 [16]. Because we find that K50 acetylation can occur in vitro when Tat is monomethylated, we examined whether Tat is subject to sequential monomethylation/acetylation in vivo.
To search for monomethylated/acetylated Tat in cells, we performed mass spectrometry of Tat immunoprecipitated from TNFα-activated J-Lat A2 cells. This Jurkat-derived cell line harbors an integrated bicistronic lentiviral vector, which expresses FLAG-tagged Tat and GFP from the integrated HIV LTR (LTR-Tat-IRES-GFP) upon stimulation with TNFα [31]. Immunoprecipitated material was separated by SDS-PAGE and stained with FLAMINGO fluorescence dye. The Tat band was cut from the gel and applied to in-gel digestion with chymotrypsin. Residual digested peptides were analyzed by MALDI-TOF/TOF mass spectrometry. A representative MALDI-TOF MS spectrum of the digested peptides is shown in Figure 2A, and more than 100 peptide ion signals were detected. A peptide encompassing the Tat ARM region without modification was detected at 1084.681 m/z, which was identified as the peptide from glycine 48 to arginine 55 in the Tat-FLAG molecule by MALDI-TOF/TOF MS/MS analysis (Figure 2B). We also detected a mass signal at 1197.724 m/z, which corresponded to the Tat peptide from lysine 50 to arginine 57 carrying a monomethyl group at lysine 51 (Figure 2C). Bimodified Tat (AcK50/Me1K51) was not detected in this experiment. In addition, we did not detect dimethylation at K51, but detected a peptide in which both K50 and K51 carried a mass addition of 42 Da, indicating that these residues could be either acetylated or trimethylated in cells (data not shown). Mass spectrometry cannot differentiate efficiently between these two modifications. We previously confirmed that acetylation of K50 exists in cells using acetylation-specific Tat antibodies [25] but could not detect trimethylation of K51 using trimethyl-Tat-specific antibodies [32] supporting a model where both residues may be acetylated rather than trimethylated in cells. Further experiments using modification-specific antibodies directed against both sites are currently underway.
To independently analyze the existence of bimodified (AcK50/Me1K51) Tat in cells, we generated a polyclonal antiserum specific for doubly modified Tat. ARM peptides carrying an acetyl group at position 50 and a monomethyl group at position 51 were injected into rabbits and affinity purified on a column carrying the bimodified antigen. The resulting antibodies (α-AcK50/Me1K51 Tat) were specific for the bimodified ARM peptides and did not react with singly modified peptides in dot blot analysis (Figure 3A). In contrast, an antiserum that we previously generated against monomethylated K51 in Tat (α-Me1K51 Tat) [16] reacted with ARM peptides monomethylated at K51 as expected but also showed cross-reactivity with bimodified peptides (Figure 3A). The same results were obtained when we tested the antibodies by western blot analysis of synthetic Tat proteins (aa 1–72), which carried either one or both modifications. The α-AcK50/Me1K51 Tat antibodies specifically recognized doubly modified Tat while the α-Me1K51 Tat recognized both methylated and doubly modified Tat (Figure 3B). No cross-reactivity was observed with unmodified or acetylated Tat.
To test whether doubly modified Tat exists in cells, we transfected FLAG-tagged Tat into 293T cells and purified Tat with α-FLAG agarose. K51 methylated Tat was readily detected by western blot analysis using α-Me1K51 Tat antibodies while no signal was detected with the α-AcK50/Me1K51 Tat antibodies (Figure 3C). Of note, both antibodies recognized their cognate antigens with similar sensitivities as shown by western blot analysis of full-length synthetic methylated and acetylated/methylated Tat proteins (Figure 3C). Similar experiments were performed with antibodies against AcK50Me2K51 and AcK50Me3K51 in Tat and showed no reactivity with Tat in cells (data not shown). This result confirms the data obtained by mass spectrometry, which indicate that doubly modified Tat is not a major Tat species in cells.
We speculated that Tat is demethylated at K51 before acetylation occurs. Recombinant LSD1/KDM1 demethylated synthetic monomethylated Tat in a dose-dependent manner as shown by western blot analysis using α-Me1K51 Tat antibodies (Figure 4A). LSD1/KDM1 also demethylated its cognate substrate, dimethyl lysine 4 in histone H3, as expected (Figure 4B). Interestingly, LSD1/KDM1 demethylated monomethylated Tat in vitro regardless of whether the neighboring K50 residue was acetylated or not, suggesting that LSD1/KDM1 may demethylate Tat in cells either before or immediately after acetylation had occurred (Figure 4C).
To test whether LSD1/KDM1 is involved in the demethylation of Tat in cells, we introduced lentiviral vectors carrying shRNAs against LSD1/KDM1 into J-Lat A2 cells and reduced endogenous expression of LSD1/KDM1 (Figure 4D). We then induced expression of Tat with TNFα and monitored monomethylation of Tat K51 using western blotting with α-Me1K51 Tat antibodies. Monomethylation of Tat was 2.6-fold enhanced in cells expressing shRNAs against LSD1/KDM1 as compared to cells expressing control shRNAs although the overall expression of Tat was reduced (Figure 4D). This reduction is explained by the negative effect of LSD1-knockdown on Tat transcriptional activity (see below), which drives Tat expression from the LTR in these cells. Collectively, these results demonstrate that LSD1/KDM1 demethylates Tat K51 in vitro and in cells.
To test whether LSD1/KDM1 interacts with Tat in cells, FLAG-tagged Tat proteins were expressed in 293T cells after transient transfection. Following immunoprecipitation with α-FLAG antibodies, endogenous LSD1/KDM1 was detected by western blotting in the immunoprecipitated material (Figure 5A). Tat proteins carrying point mutations either in K50 (K50A) or K51 (K51A) also efficiently coimmunoprecipitated with LSD1/KDM1 indicating that the interaction was not dependent on demethylation of K51 in Tat. A similar, albeit weaker interaction was observed when we tested Tat's interaction with the LSD1/KDM1 cofactor CoREST [33], [34] suggesting that Tat may recruit a functional LSD1/KDM1/CoREST complex to the HIV promoter. In contrast, no interaction of Tat or Tat mutants was observed with cellular HDAC1, often also described as part of LSD1/KDM1/CoREST corepressor complexes [35], [36], [33], [37]. It is not clear at the moment whether the observed interactions of Tat with LSD1/KDM1 or CoREST are direct or mediated by other cellular proteins.
To test the hypothesis that LSD1/KDM1 and CoREST are recruited to the HIV LTR, we performed chromatin immunoprecipitation assays. Chromatin was prepared from J-Lat A2 cells, in which Tat expression was stimulated by TNFα treatment or which were left nonstimulated. Quantitative PCR analysis of the immunoprecipitated material with primers specific for the HIV LTR indicated that LSD1/KDM1 and CoREST, while only present at low concentrations at the promoter under nonstimulated conditions, were specifically recruited in response to TNFα stimulation (Figure 5B, LSD1 and CoREST). No signal was detected when immunoprecipitation was performed with beads alone (Figure 5B, Control). Overall cellular expression of LSD1/KDM1 and CoREST was unchanged in response to treatment with TNFα in J-Lat A2 cells (Figure 5C). These results demonstrate that LSD1/KDM1 and CoREST are recruited to the HIV LTR in response to Tat. However, recruitment may also occur indirectly via other LTR activators in response to TNFα treatment.
To test the function of LSD1/KDM1 in HIV transcription, A2 cells were transduced with lentiviral vectors expressing two different shRNAs directed against LSD1/KDM1 or control shRNAs directed against firefly luciferase or a scrambled shRNA. All vectors also expressed the mCherry marker to track infection efficiencies. More than 90% of cells expressed mCherry after lentiviral vector infection, and no difference in infection efficiencies was observed between the different lentiviral vectors (not shown). ShRNA-expressing cells were stimulated with TNFα, and expression of GFP was measured by flow cytometry. GFP expression was reduced by 40-60% in LSD1/KDM1 knockdown cells as compared to cell lines expressing luciferase or scrambled shRNAs (Figure 6A). No toxicity of LSD1 knockdown was observed in shRNA-treated cells as measured by dye exclusion in flow cytometry (Figure 6B). ShRNA#1 had a stronger suppressive effect on GFP expression than shRNA#2 mirroring the degree of LSD1/KDM1 knockdown in these cells (Figure 6C). The same result was obtained in 5A8 J-Lat cells harboring a full-length GFP-tagged latent HIV genome (Figure S2). Here, reactivation from latency is achieved by stimulation with α-CD3/CD28 antibodies in ∼40% of cells. Only 5 or 15% of cells reactivate HIV transcription in cells treated with LSD1#1 or LSD1#2 shRNAs, respectively, confirming that LSD1 is important for full transcriptional activity after reactivation from latency after T cell receptor stimulation (Figure S2). A similar suppression of GFP expression in the absence of cell toxicity was observed in A2 cells, in which the expression of CoREST was downregulated (Figure 6D–F). Interestingly, in A72 cells, in which GFP expression is driven by the LTR alone in the absence of Tat, no effect of either downregulation of cellular LSD1 or CoREST expression was observed pointing to a specific effect of LSD1/KDM1 and CoREST in Tat transactivation (Figure S3). Collectively, these results demonstrate that an LSD1/KDM1/CoREST complex, often a suppressor of cellular gene expression, functions as a co-activator of HIV transcription.
To test whether LSD1/KDM1 activates HIV transcription through Tat demethylation, we introduced siRNAs specific for LSD1/KDM1 or control siRNAs into HeLa cells. Cells were then co-transfected with the HIV LTR luciferase reporter gene and an expression construct for Tat. Tat transactivation of the HIV LTR was suppressed by ∼50% when expression of LSD1/KDM1 was reduced in cells indicating that LSD1/KDM1 is a positive cofactor of Tat transactivation (Figure 6G). Expression of the TatK51A mutant resulted in a similar decrease in Tat transactivation (∼50%) as previously reported [16], but no further reduction was observed in LSD1/KDM1 knockdown cells supporting the model that LSD1/KDM1 activates Tat transactivation through K51 demethylation (Figure 6G). The transcriptional activity of the HIV LTR alone was also reduced in LSD1/KDM1 knockdown cells (∼28%) although values did not reach statistical significance indicating that an additional target for LSD1/KDM1 may or may not exist at the HIV LTR in the absence of Tat (Figure 6G). Importantly, LSD1/KDM1 knockdown had no suppressive effect on the EF-1α promoter that was driving Tat expression in these co-transfection experiments excluding the possibility that LSD1/KDM1 controls Tat expression and not Tat function (Figure 6G). Successful knockdown of LSD1/KDM1 expression was confirmed by western blotting (Figure 6H).
Since LSD1/KDM1 belongs to the amine oxidase enzyme superfamily that oxidatively removes methyl groups from di- or monomethylated lysines, some monoamine oxidase (MAO) inhibitors can act as LSD1/KDM1 inhibitors [34], [38], [39], [40]. It was recently reported that the MAO antidepressant agent phenelzine (phenethylhydrazine) is far more potent in inhibiting LSD1/KDM1 activity in cells than previously appreciated [38]. To test the activity of this agent in HIV infection, J-Lat A2 cells were treated with increasing amounts of phenelzine or the CDK inhibitor 5, 6-dichloro-1-β-D-ribofuranosyl-1H-benzimidazole (DRB), a known Tat inhibitor. Phenelzine, similar to DRB, prevented TNFα-mediated activation of gene expression in a dose-dependent manner, albeit at ∼150 fold higher concentrations than DRB (IC50 = 300 µM Figure 7A, white circle). No cell toxicity was observed for both agents at the tested concentrations (Figure 7A, black circle).
The same experiment was performed in a primary T cell model of HIV latency. Quiescent CD4+ T cells were isolated from blood of two healthy donors and were spin-inoculated with an infectious clone of HIV expressing luciferase within the nef open reading frame following a similar protocol as previously described [41]. Infected CD4+ cells were cultured with the integrase inhibitor saquinavir for 3 days to ensure that postintegration latency was measured and then treated with increasing amounts of phenelzine followed by stimulation with α-CD3 and α-CD28 antibodies to activate latent HIV transcription. Activation of luciferase expression was successfully suppressed by phenelzine treatment confirming the effectiveness of the drug in the context of a full-length infectious clone of HIV in primary T cells (Figure 7B). A slight decrease in cell viability was observed in one donor at the highest concentration of phenelzine (Figure 7B). However, no effect of phenelzine on cell viability was observed in additional two donors, in whom reactivation from HIV latency were also successfully inhibited by the drug (Figure S4). Interestingly, in activated primary T cells, phenelzine was more efficient in suppressing HIV gene expression than DRB while in latently infected, but not activated, cells phenelzine, contrary to DRB, had no suppressive effect on luciferase expression (Figure S5). Collectively, these results identify phenelzine as a potent new inhibitor of HIV reactivation from latency and support the model that LSD1/KDM1 is a novel activator of HIV transcription through Tat demethylation.
We investigated whether Tat lysine methylation and acetylation events within the Tat ARM are linked via a demethylation step mediated by LSD1/KDM1. Similar to previous reports on the tumor suppressor p53 [28], [29], we find that methylation and acetylation of Tat can occur sequentially in vitro with methylation at K51 occurring first allowing subsequent acetylation of K50, albeit at diminished efficiency. Detailed in vivo analysis of the Tat ARM reveals that a bimodified Tat form does likely not exist in cells because it was not detected by mass spectrometry and by western blotting using newly generated bispecific Tat antibodies. Instead, we identify LSD1/KDM1 as a K51-specific Tat demethylase and a novel transcriptional activator of HIV transcription. These findings may be clinically relevant because we demonstrate that phenelzine, a MAO inhibitor with activity against LSD1/KDM1, successfully suppresses re-activation of HIV transcription in latently infected T cells.
Until recently, it was unclear whether methylation of lysines is reversible. Today, there exist two types of lysine demethylases, LSD1/KDM1 and Jumoji C domain-containing demethylases [42]. LSD1/KDM1 is a flavin adenine dinucleotide (FAD)-dependent amine oxidase, which can demethylate lysine 4 in histone H3 (an activatory mark) and lysine 9 in histone H3 (a silencing mark). LSD1/KDM1 can remove methyl groups from mono- or di-, but not tri-methylated lysines [43], [44]. Since the FAD-dependent amine oxidase family of enzymes, which includes MAO-A, MAO-B and LSD1/KDM1, share a common mechanism for the oxidative cleavage of the unactivated nitrogen, known MAO inhibitors such as phenelzine used in this study or others have activity against LSD1/KDM1 [34], [38], [39], [40].
It was recently reported that besides histones, LSD1/KDM1 can also demethylate non-histone proteins including the tumor suppressor p53, the DNA methylase Dnmt1, and transcription factor E2F1 [45], [46], [47], [48]. Demethylation of p53 by LSD1 alters the interaction of p53 with its coactivator 53BP1 and represses the proapoptotic function of p53 [45]. Similarly, demethylation of Dnmt1 by LSD1 triggers a loss of protein stability and a loss of global DNA methylation while demethylation of E2F1 is required for E2F1 stabilization and apoptotic function [46], [47]. Our finding that Tat function is activated by LSD1/KDM1-mediated demethylation adds another nonhistone protein to the growing list of LSD1/KDM1 substrates. Like E2F1, Tat is activated by LSD1/KDM1 demethylation, a finding that supports the model that the coordinated occurrence of Tat modifications is essential for efficient Tat transcriptional activity.
Our finding that LSD1/KDM1 and CoREST are both recruited to the activated HIV LTR in vivo points to Tat demethylation as a novel mechanism how HIV may corrupt the function of a known corepressor complex to enhance its own replication. The interaction with CoREST is known to direct the LSD1/KDM1 activity towards lysine 4 in nucleosomal histone H3 and is associated with transcriptional repression [33], [34]. However, LSD1/KDM1 can also act as a transcriptional activator for instance through demethylation of lysine 9 in histone H3 in conjunction with androgen receptor-mediated transcription [44], through demethylation of lysine 9 in histone H3 in α-herpesvirus infections [49] or through demethylation of the E2F1 transcription factor which activates its apoptotic function [46]. Evidence that support a role of Tat demethylation in the LSD1/KDM1 coactivator function in HIV transcription comes from the Tat K51A mutant, which remains unaffected by siRNA-mediated downregulation of LSD1/KDM1 expression. However, other LSD1 substrates may exist at the HIV promoter that activate HIV transcription when Tat is absent. An attractive target is methylated lysine 9 in histone H3 at the HIV provirus, which was previously linked to HIV latency [9], [50], [51] and is the target of the activatory function of LSD1/KDM1 in the transcriptional control of herpes simplex virus- and varicella zoster virus latency [49]. Notably, a recent study shows that the latter function also involves CoREST indicating that a functional LSD1/KDM1/CoREST complex can function as suppressor of cellular gene expression or as activator of viral transcription [52]. Interestingly, HDAC1/2 are generally part of this complex, but we do not observe any interaction of Tat with HDAC1 in our study. This may point to HDAC2 or another HDAC associated with the LSD1/KDM1 subcomplex recruited to the HIV LTR or may indicate that Tat specifically dissociates HDACs from LSD1/KDM1 complexes involved in its demethylation. Notably, binding of CoREST, but not HDAC1, to LSD1/KDM1 restored the ability of recombinant LSD1 to demethylate nucleosomal substrates while HDACs are thought to act upstream of LSD1/KDM1 by providing hypoacetylated substrates for demethylation [33].
We focused here on the interaction between Tat demethylation at K51 and K50 acetylation, but interplay may also exist between K51 demethylation and other known Tat modifications such as arginine methylation within the Tat ARM (Table S1). We have previously shown that both acetylation and deacetylation of K50 in Tat are required for full Tat transactivation. While acetylation of K50 by p300/KAT3B dissociates Tat from TAR RNA and P-TEFb, deacetylation by SIRT1 may be necessary to recycle nonacetylated Tat for reentry into the transactivation cycle [12]. Here, we show that the same “Yang/Yang” principle applies to methylation and demethylation of K51 in Tat. Both, methylation of K51 by Set7/9/KMT7 and demethylation of K51 by LSD1/KDM1 activate Tat transactivation because knockdown or inhibition of each enzyme leads to reduction of Tat transcriptional activity in a K51-dependent manner. We propose a model where demethylation of Tat occurs as a critical step during the Tat transactivation cycle possibly before acetylation of Tat by p300/KAT3B occurs (Figure 8). In support of this model, we have previously shown that monomethylation of Tat is an early event that strengthens the interaction of Tat with TAR RNA and P-TEFb [16]. In addition, we show here that in vitro monomethylation at K51 decreases efficient acetylation of K50 by p300/KAT3B supporting the model that prior demethylation is required to allow full Tat acetylation at K50 and possibly at K51 in cells. We also found in preliminary experiments that LSD1/KDM1 coimmunoprecipitates with p300/KAT3B in cellular extracts pointing to a potential recruiting function of LSD1/KDM1 for p300/KAT3B to Tat (N. Sakane and M.Ott, unpublished data).
Our data provide important first evidence that LSD1 inhibitors may function as therapeutics to suppress reactivation of HIV transcription in latently infected cells. They further support the model that targeting Tat posttranslational modifications may be a valid therapeutic strategy to control HIV transcription and latency; Tat becomes “locked” in one modified state when individual modifying enzymes are blocked and the normal flow of Tat modifications is disturbed. Interestingly, MAO inhibitors with inhibitory functions towards LSD1/KDM1 have suppressive activity in latent infections of α-herpesvirus [49]. Our results indicate that they may have a broader antiviral application that includes HIV-1. The development of more specific LSD1/KDM1 inhibitors will bring further validation to the model that LSD1/KDM1 is an important new drug target in the treatment of latent HIV infection.
HeLa and 293T cells (obtained from the American Type Culture Collection), and the J-Lat clone A2 [31] were maintained under standard cell-culture conditions. The following antibodies were commercially available: α-LSD1/KDM1 (#ab51877, abcam, Cambridge, MA), α-CoREST (#ab24166, abcam), α-FLAG M2 (#F-3165 Sigma-Aldrich, St. Louis MO), α-histone H3K4me2 (#07-030, Millipore, Billerica, MA), α-histone H3 (#07-690, Millipore), α-tubulin (#T6074, Sigma-Aldrich), α-Tat (MMS-116P, Covance, Emeryville, CA), and α-CD28 (#16-0289-85 eBioscience, San Diego CA). α-K51 monomethylated Tat polyclonal antibodies were previously described [16]. α-CD3 (OKT-3) was obtained from the UCSF monoclonal antibodies core facility. The α-HDAC1 polyclonal antibodies were a kind gift of Eric Verdin, Gladstone Institute of Virology and Immunology, San Francisco.
Phenelzine Sulfate was purchased from Spectrum Chemical MHG Corp. (#3032 Gardena, CA) and Enzo Life Sciences (#EI-217, Plymouth meeting, PA). Saquinavir was obtained from the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH. Recombinant human TNFα was purchased from Humanzyme (#HZ-1014, Chicago, IL).
The synthetic Tat proteins (aa 1–72) was synthesized as previously described [12] together with the Tat ARM short peptide (aa 45-58) by Dr. Hans-Richard Rackwitz (Peptide Specialty Laboratories GmbH, Heidelberg, Germany).
The HIV LTR luciferase construct, the EF-1α-Tat/FLAG expression vector, the K51A mutated EF-1α-Tat/FLAG expression vector and the pEF-1α-RL (Renilla luciferase) were described before [16]. The His-tagged LSD1 prokaryotic expression vector was previously described (Department of Pathology, Harvard Medical School, [43]).
A modified version of the pSicoR lentiviral vector that encodes the mCherry reporter gene driven by an EF-1α promoter (pSicoRMS) [53], [54] was kindly provided by Matthew Spindler (Gladstone Institute of Cardiovascular Disease). ShRNAs targeting LSD1 (LSD1 #1:GAAGGCTCTTCTAGCAATA, LSD1 #2: CATGTGCCTGTTTCTGCCATG) were cloned into pSicoRMS. The pSicoRMS containing a non-targeting control sequence (shScramble: GTCAAGTCTCACTTGCGTC) [55] and targeting luciferase (shLuciferase: CTTACGCTGAGTACTTCGA) was kindly provided by Dr. Silke Wissing (Gladstone Institute of Virology and Immunology). ShRNAs against CoREST and control empty pLKO.1 vector were purchased from Thermo Fischer Scientific (Waltham MA).
C-terminal FLAG-tagged Tat protein (Tat/FLAG) purified from J-Lat A2 cells (∼100 ng) was further purified by SDS-PAGE (FLAMINGO gel stain, Bio-Rad Hercules, CA). Tat band was excised and washed with 200 µL of 50 mM ammonium bicarbonate containing 50% (v/v) ethanol followed by 200 µL of ethanol twice. The Tat protein in the gel was reduced with 10 mM DTT in 50 mM ammonium bicarbonate for 1 h at 56°C and alkylated with 55 mM iodoacetamide in 50 mM ammonium bicarbonate for 30 min at room temperature. After reduction and alkylation, the gel was dehydrated with acetonitrile 3 times. The gel was rehydrated by adding 200 µL of 50 mM ammonium bicarbonate with 5 ng/µL of chymotrypsin (Roche, Penzberg, Upper Bavaria, Germany) and incubated at 30°C for 2 h. Digested Tat peptides were extracted from the gel with 1% (v/v) formic acid containing 30% (v/v) acetonitrile followed by 1% (v/v) formic acid containing 60% (v/v) acetonitrile. The extracted peptide solution was dried up by speed vac. Then, residual peptides were reconstituted with 30 µL of 0.1% (v/v) TFA containing 2% (v/v) acetonitrile and desalted by ZipTipC18 (Millipore) according to the manufacturer's description. 2 µL of cleaned peptide solution eluted from ZipTipC18 was deposited on the Bruker metallic MALDI target (MTP384 ground steel, Bruker Daltonics, Billerica, MA) and mixed with 2 µL of saturated matrix solution (α-cyano-4-hydroxycinnamic acid solution in 33% (v/v) acetonitrile, 0.1% (v/v) TFA). Peptide mixture was allowed to dry at room temperature. The peptide mixture was analyzed by ultraflex III TOF/TOF (Bruker Daltonics) MALDI-TOF/TOF mass spectrometer, operated in reflector mode for positive ion detection, and controlled by flexControl 3.0 software. For MS/MS acquisitions, the ions of interest were fragmented by laser-induced decay, and mass of fragments was analyzed using LIFT mode. Monoisotopic mass was determined using flexAnalysis 3.0 software with the SNAP peak picking algorithm. The modifications of peptides were analyzed using UniMod database in the Biotools software.
The strategy to generate bimodified Tat (AcK50/Me1K51) specific antibodies was performed as previously described [16], [32]. Briefly, KLH conjugated bimodified ARM peptides (AcK50/Me1K51) were injected into rabbits. The same peptides were used for affinity purification of the resulting antiserum. Specificity of antibodies was monitored by dot-blot analysis using ARM peptides and western blot analysis using synthetic full-length Tat proteins.
Protein expression and purification of recombinant LSD1 and in vitro demethylation reactions of LSD1 (0.5–2.0 µg) with synthetic Tat proteins (3 µg) or purified total cellular histones (8 µg) were performed as previously described [43]. The reactions were analyzed by western blotting using α-Me1K51 Tat antibodies (1 µg ml). In vitro methylation reactions with synthetic Tat protein (aa 1-72; 1 µg) and ARM peptides (aa 45-58; 100 µM), Set7/9-KMT7 enzyme (2 µg Millipore), and 3H-S-Adenosyl Methionine (Perkin Elmer) were performed as previously described [16]. In vitro acetylation reactions with synthetic Tat proteins (1 µg), ARM peptides (100 µM), GST-p300 HAT enzyme (aa 1195-1810; 5 µg; [56]) and 14C-acetyl CoA (0.1 µCi; Perkin Elmer) were performed as described [10]. Reactions were separated by SDS-PAGE or Tris-Tricine gel electrophoresis and visualized by autoradiography.
siRNA analysis for HeLa cells were performed as described [16]. Briefly, HeLa cells were transfected with pooled LSD1 and control siRNAs (200 pmol, Dharmacon; Lafayette, CO) using Oligofectamine (#58303, Invitrogen, Carlsbad, CA) and were retransfected after 48 h with the HIV LTR luciferase construct (200 ng), Tat-expressing vectors (2 ng), and corresponding amounts of the empty vector using lipofectamine reagent (#50470, Invitrogen). Cells were harvested 24 h later and processed for luciferase assays (Luciferase Assay System, #E1501, Promega, Madison, WI) or western blotting.
J-Lat A2 cells were transduced with pseudotyped pSicoRMS-derived lentiviral vectors expressing shRNAs against LSD1 (shLSD1 #1 and #2), against luciferase (shLuciferase) or a nontargeting shRNA control (ShScramble). These lentiviral vectors also express the mCherry protein under the control of the EF-1α promoter (see cells, reagents and antibodies). 5 to 10 days after infection, cells were treated with 0.08 ng/ml of TNFα for 12 h. Expression of GFP and mCherry was analyzed by flow cytometry (BD LSRII, Beckton Dickinson, Franklin Lakes, NJ). Similar experiments were performed using pLKO.1-derived vectors expressing shRNAs against CoREST (shCoREST #1 and #2) or empty vector controls (shControl). When pLKO.1 vectors were used, puromycin was added one day after shRNA infection (1 ng/ml). Cell viability was determined by propidium iodide staining (#P-3566, Invitrogen) or LIVE/DEAD Fixable Violet Dead Cell Stain Kit (#L34958, Invitrogen) followed by flow cytometry.
Chromatin immunoprecipitations from J-Lat A2 cells were performed as previously described [4], [16]. Chromatin solutions were isolated from A2 cells treated with TNFα (2 ng/ml) and were immunoprecipitated with α-LSD1 antibodies (abcam), α-CoREST antibodies (abcam) or control rabbit pre-immune serum. The immunoprecipitated material was quantified by real-time PCR with primers specific for the HIV LTR using the ABI7700 Sequence Detection System (Applied Biosystems, Foster City, CA) and the 2x Hot Sybr real-time PCR kit (#HSM-400, McLab, South San Francisco, CA). Primer sequences were: HIV LTR upstream: GAGCCCTCAGATCCTGCATA, HIV LTR downstream: AGCTCCTCTGGTTTCCCTTT.
293T cells were transfected with Tat expressing vector using Fugene 6 reagent (Roche). 24 h after transfection, cells were lysed in IP buffer (250 mM NaCl, 0.1% NP40, 20 mM NaH2PO4 (pH 7.5), 5 mM EDTA, 30 mM sodium pyrophosphate, 10 mM NaF and protease inhibitors) and immunoprecipitated with α-FLAG M2 agarose (Sigma-Aldrich) over night at 4°C. Beads were extensively washed and analyzed by western blotting with α-LSD1, α-CoREST, α-HDAC1 or monoclonal α-FLAG antibodies.
The infectious NL4-3-luciferase clone of HIV was generated by cloning the BamHI to XhoI fragment of pNL-Luc-E-R- within the nef coding region into pNL4-3 [48]. This generates a fully infectious clone capable of multiple rounds of infection and producing luciferase driven from the LTR promoter. Infectious particles were produced after transfection of the clone into 293T cells. Two days after transfection, the transfected supernatants were collected and concentrated by ultra-centrifuge (20,000 rpm, 2 h), and virus concentration was determined by analyzing concentration of p24gag (HIV-1 antigen p24 ELISA kit #NEK050A001KT, Perkin Elmer). CD4+ T cells were isolated from human whole blood buffy coats obtained from anonymous donors by centrifugation onto a Histopaque-1077 cushion (#10771, Sigma-Aldrich), enrichment of T cells by rosetting with sheep red blood cells (#CS115 Colorado Serum, Denver, CO) and depletion of non-CD4+ T cells with the CD4+ T cell isolation kit (#130-091-155, Miltenyi Biotec Bergisch Gladbach, Germany) and AutoMACS cell separator (Miltenyi Biotec). Purity of isolated CD4+T cells was confirmed by flow cytometry. For infection, 1 µg of p24gag was used for 5×106 CD4 T cells. The mixture of virus and cells were centrifuged at 2400 rpm for 2 h. After spinoculation, cells were cultured in the presence of 5 µM saquinavir for 3 days and were then stimulated with α-CD3 (2.5 µg/ml, coated) and α-CD28 antibodies (1 µg/ml, soluble) in the presence or absence of phenelzine (100 µM–1 mM). After over night incubation, cells were harvested and processed for luciferase assays (Luciferase Assay System, Promega). Cell viability was determined by propidium iodide staining (#P-3566, Invitrogen).
|
10.1371/journal.pgen.1002572 | Nos2 Inactivation Promotes the Development of Medulloblastoma in Ptch1+/− Mice by Deregulation of Gap43–Dependent Granule Cell Precursor Migration | Medulloblastoma is the most common malignant brain tumor in children. A subset of medulloblastoma originates from granule cell precursors (GCPs) of the developing cerebellum and demonstrates aberrant hedgehog signaling, typically due to inactivating mutations in the receptor PTCH1, a pathomechanism recapitulated in Ptch1+/− mice. As nitric oxide may regulate GCP proliferation and differentiation, we crossed Ptch1+/− mice with mice lacking inducible nitric oxide synthase (Nos2) to investigate a possible influence on tumorigenesis. We observed a two-fold higher medulloblastoma rate in Ptch1+/− Nos2−/− mice compared to Ptch1+/− Nos2+/+ mice. To identify the molecular mechanisms underlying this finding, we performed gene expression profiling of medulloblastomas from both genotypes, as well as normal cerebellar tissue samples of different developmental stages and genotypes. Downregulation of hedgehog target genes was observed in postnatal cerebellum from Ptch1+/+ Nos2−/− mice but not from Ptch1+/− Nos2−/− mice. The most consistent effect of Nos2 deficiency was downregulation of growth-associated protein 43 (Gap43). Functional studies in neuronal progenitor cells demonstrated nitric oxide dependence of Gap43 expression and impaired migration upon Gap43 knock-down. Both effects were confirmed in situ by immunofluorescence analyses on tissue sections of the developing cerebellum. Finally, the number of proliferating GCPs at the cerebellar periphery was decreased in Ptch1+/+ Nos2−/− mice but increased in Ptch1+/− Nos2−/− mice relative to Ptch1+/− Nos2+/+ mice. Taken together, these results indicate that Nos2 deficiency promotes medulloblastoma development in Ptch1+/− mice through retention of proliferating GCPs in the external granular layer due to reduced Gap43 expression. This study illustrates a new role of nitric oxide signaling in cerebellar development and demonstrates that the localization of pre-neoplastic cells during morphogenesis is crucial for their malignant progression.
| Medulloblastoma is a common pediatric brain tumor, a subtype of which is driven by aberrant hedgehog pathway activation in cerebellar granule cell precursors. Although this tumor etiology has been intensively investigated in the well-established Ptch1+/− mouse model, knowledge is still lacking about the molecular interactions between neoplastic transformation and other developmental processes. Nitric oxide (NO) has been reported to be involved in controlling proliferation and differentiation of these cells. Therefore, inactivation of the NO–producing enzyme Nos2 in combination with the mutated Ptch1 gene should provide insight into how developmental regulation influences pathogenesis. Here, we describe a new role for NO in developing neuronal precursors of the cerebellum facilitating physiologically accurate migration via regulation of Gap43. We further demonstrate that disturbance of these processes leads to retention of granule precursor cells to the cerebellar periphery. Together with the sustained proliferation of these cells in combined Ptch1+/− Nos2−/− mice, this effect results in an increased medulloblastoma incidence relative to Ptch1+/− mice and demonstrates a new disease-promoting mechanism in this tumor entity.
| Medulloblastoma (MB) is a highly malignant tumor of the cerebellum that preferentially develops in children and adolescents. Although the survival rate for standard risk MB is around 70% [1] surviving patients often suffer from neurodevelopmental and cognitive side effects of the aggressive therapy [2]. Therefore, improved understanding of the molecular pathomechanisms driving MB growth is necessary to develop less toxic and more effective treatments. Recent molecular profiling studies suggested at least four MB subtypes that are associated with distinct expression profiles, genomic aberrations and clinical features [3], [4]. One of these MB subtypes is characterized by aberrant activation of the hedgehog (Hh) pathway and typically corresponds to the desmoplastic (nodular) MB variant. This subtype is supposed to develop from granule cell precursors (GCPs) of the external granular layer (EGL) [5].
The EGL is a transient germinal zone at the subpial cerebellar surface consisting of rhombic lip-derived progenitor cells that have migrated tangentially to the emerging cerebellar cortex at late stages of embryonal brain development [6]. During the early postnatal period in mouse, the morphogenic factor sonic hedgehog (Shh) is secreted by subjacent Purkinje cells and binds to patched receptors (Ptch1 and Ptch2) expressed on the GCP surface [7]. Ligand binding to Ptch1 then leads to functional de-repression of Smoh (Drosophila smoothened homolog) and subsequent activation of Gli (Glioma-associated oncogene family zinc finger) transcription factors [8]. This launches a temporally concerted gene expression pattern causing a proliferation burst and massive expansion of the GCP population during the first two postnatal weeks [7]. In particular, the direct Gli-target N-myc [9], [10] and D type cyclins [11] were shown to be crucial for the growth and neoplastic transformation of GCPs [12]. In addition, the set of genes targeted by activated Gli transcription factors also include components of the canonical Hh pathway for feedback-loop regulation, such as the receptors Ptch1 and Ptch2 as well as the hedgehog-interacting protein (Hip) [10], [13]. After several rounds of cell division, GCPs normally exit cell cycle and accumulate at the inner site of the EGL [14], where they start to migrate through the molecular layer (ML) and the Purkinje cell layer to form the internal granular layer (IGL) [15]. The mechanisms underlying the attenuation of the mitotic response and eventually the stop of GCP proliferation are not well understood. The most evident concepts describe extrinsic cues in gradient-based models to trigger GCP differentiation with increasing distance to the region of the outer EGL [16]. Finally, the EGL disappears at about three weeks after birth in mice.
PTCH1 was identified as a frequent target of inactivating mutations or genomic loss in sporadic MBs [17]–[19] that belong to the molecular subtype hallmarked by an aberrant activity of hedgehog signaling. The monoallelic inactivation of the Ptch1 gene in mice and thus downstream activation of the Hh pathway leads to MB development at a frequency of about 10–15% [20]. This mouse model has provided substantial insights into the pathogenesis of Hh-dependent MBs and has been used in different cross-breeding experiments to investigate tumor suppressor gene functions in this particular context [21], [22].
Nitric oxide (NO) is a highly reactive gaseous molecule involved in various physiological processes ranging from vasculature modulation to neurotransmission [23], [24]. NO is produced by three distinct enzyme isoforms: i) neuronal nitric oxide synthase (nNos/Nos1), ii) inducible nitric oxide synthase (iNos/Nos2), and iii) endothelial nitric oxide synthase (eNos/Nos3). Though being constitutively expressed in their respective tissue, nNos and eNos activity strongly depends on calcium [25], whereas calcium-independent iNos is primarily regulated by transcriptional induction, e.g. by inflammatory cytokines and endotoxins [26], which permits higher quantities of NO generation. The role of NO in cancer initiation and progression is heterogeneous with opposing effects in different malignancies [27]. Considering effects of tumor stroma, increased angiogenesis was reported to be associated with elevated Nos activity [28] and some immune-related processes were found to be mediated by NO [29], including cytotoxicity of activated microglia [30]. Finally, NO released by vascular endothelial cells was reported to build a niche-like microenvironment for maintenance of glioma stem cells [31]. In the context of cerebellar development, Nos2 (inducible Nos) is initially expressed in early GCPs, whereas Nos1 (neuronal Nos) is hardly present before postnatal day 7 (Cerebellar Development Transcriptome Database [32]). Successively, Nos1 expression increases along with granule cell differentiation [33] and predominantly contributes to the common NO signaling that becomes apparent in the IGL as development proceeds [34]. Evidence has been provided that NO negatively acts on proliferation of neuronal precursors during adult neurogenesis [35]. Similarly, Ciani and colleagues demonstrated enhanced proliferation of cerebellar precursor cells upon inhibition of NO synthases [36].
Here, we report on the generation of Ptch1+/− Nos2−/− mice to investigate the impact of Nos2 on tumor development in Ptch1 hemizygous mutant mice. Interestingly, we observed an approximately two-fold increase in the incidence of spontaneous MB in Ptch1+/− Nos2−/− mice in comparison to Ptch1+/− Nos2+/+ mice. To characterize the molecular pathomechanism underlying the tumor-promoting effect of Nos2 deficiency in Ptch1+/− mice, we performed comprehensive expression and DNA copy number profiling of MB tumors (Ptch1+/− Nos2+/+ versus Ptch1+/− Nos2−/−) as well as expression profiling of normal cerebellar tissue samples from different developmental stages and various genotypes (Ptch1+/− Nos2+/+, Ptch1+/− Nos2−/−, Ptch1+/+ Nos2−/− and wild-type mice). Downregulation of the growth-associated protein 43 (Gap43) was the most striking feature in the cerebellum of Nos2-deficient mice when compared to Ptch1+/− Nos2+/+ and wild-type mice. Subsequent functional analyses and results from in situ studies of GCPs in postnatal cerebellum allowed us to formulate a model for the tumor promoting role of Nos2 deficiency in Ptch1 mutant mice via deregulation of Gap43-dependent migration of GCPs.
Survival analyses of 315 wild-type mice, 412 Ptch1+/+ Nos2−/− mice, 215 Ptch1+/− Nos2+/+ mice and 221 Ptch1+/− Nos2−/− mice demonstrated a significantly higher MB incidence in the group of Ptch1+/− Nos2−/− mice relative to the group of Ptch1+/− Nos2+/+ mice (p = 0.0007, Logrank test, Figure 1A). In total, 11% of the Ptch1+/− Nos2+/+ mice (24/215) and 21% of the Ptch1+/− Nos2−/− mice (47/221) were sacrificed due to the development of cerebellar MB. None of the 315 wild-type and the 412 Ptch1+/+ Nos2−/− mice developed MBs. These observations indicate a MB-promoting role of Nos2 deficiency in Ptch1+/− mice.
In humans, Hh-dependent MBs typically correspond to the desmoplastic subtype. MBs in Ptch1+/− mice, however, microscopically resemble the classic MB subtype [20]. Histological analysis of MBs in Ptch1+/− Nos2+/+ and Ptch1+/− Nos2−/− mice demonstrated similar morphological features (Figure 1B–1E). The tumors were composed of densely packed sheets of cells with hyperchromatic carrot-shaped nuclei and scant cytoplasm. There were no obvious histopathological differences between MBs of the two genotypes.
For an initial assessment of the molecular tumor characteristics, gene expression of hedgehog signaling pathway components were measured in 21 MBs and 24 normal (adult) cerebellar tissue samples from both Ptch1+/− Nos2+/+ and Ptch1+/− Nos2−/− mice. Using quantitative real-time PCR (qRT-PCR), significant downregulation of the wild-type Ptch1 transcript and upregulation of the Shh target genes Gli1 and N-myc were generally observed in the tumor samples (Figure S1), indicating all examined MBs to be of the same Hh-dependent molecular subtype. However, there were no significant differences for these genes between MBs of the two genotypes. Furthermore, targeted genetic analyses showed a loss of the wild-type Ptch1 allele in 10 of the 21 MBs investigated, while none of the tumors demonstrated a Tp53 mutation or N-myc amplification. The Cdkn2a/p16INK4a locus was retained in all tumors while a single MB demonstrated a homozygous p19ARF deletion (see Table S1 and Text S1 for details).
In order to identify the molecular pathomechanism contributing to the increased MB rate in Nos2-deficient Ptch1 mutant mice, we performed array-based gene expression profiling of three Ptch1+/− Nos2+/+ versus six Ptch1+/− Nos2−/− and comparative genomic hybridization (array-CGH) analyses of five Ptch1+/− Nos2+/+ versus seven Ptch1+/− Nos2−/− MB tissue samples. All specimens investigated had tumor cell contents between 70% and 90% as determined on corresponding formalin-fixed and paraffin-embedded (FFPE) reference sections. Differential expression of selected candidate genes was validated by qRT-PCR on an expanded, partially overlapping tumor set of seven Ptch1+/− Nos2+/+ versus seven Ptch1+/− Nos2−/− MB samples.
The expression profiling analysis revealed a total of 87 differentially regulated genes between tumors of the two genotypes (Table S2) with the vast majority (87%) showing lower transcript levels in Ptch1+/− Nos2−/− when compared to Ptch1+/− Nos2+/+ mice. As expected from the initial targeted qRT-PCR measurements, there was no difference detectable concerning the activation of Hh pathway genes. Due to the important role of Nos2 during angiogenesis and cancer-associated immune response, including microglia, stromal effects need to be particularly considered in a systemic Nos2 knockout model. However, neither the set of significantly deregulated genes nor selective determination of marker expression for pericytes, vascular endothelial cells or microglia suggested any differences in the tumor stroma between the two genotypes (see Table S3 and Text S1 for details). According to the findings of Ciani and co-workers [37], reduction of NO enhances GCP proliferation through an increased expression of the proto-oncogene N-myc. Therefore, protein levels were particularly examined for differences between tumor samples from Ptch1+/− Nos2+/+ and Ptch1+/− Nos2−/− mice. The results shown in Figure S2, however, revealed similar expression of N-myc in all MBs.
Analyses of genomic copy number alterations revealed a trisomy of chromosome 6 in the majority of MBs from both groups (11/12, Figure 2A and 2B). Moreover, a small region on chromosome 13, approximately 1.5 Mb upstream of the Ptch1 gene, showed a hemizygous deletion in healthy cerebella of Ptch1-mutant mice (data not shown) but a homozygous deletion in most tumors (10/12). Similarly, a second small region 3.8 Mb downstream of the last Ptch1 exon exhibited a loss in 9 of 12 MBs. This suggests structural changes flanking the Ptch1 locus that likely contribute to inactivation of the wild-type allele. The frequencies of genomic aberrations showed a more heterogeneous karyotype with gross structural changes in Ptch1+/− Nos2+/+ MBs when compared to Ptch1+/− Nos2−/− MBs (see Figure 2A, 2B and Text S1 for details). However, a general difference in chromosomal instability was not obvious between both genotypes. Only a small region (205.6 kb) on chromosome 14 containing the Entpd4 (ectonucleoside triphosphate diphosphohydrolase 4) gene was more frequently gained in Ptch1+/− Nos2−/− MBs (7/7) than in Ptch1+/− Nos2+/+ MBs (1/5, Figure 2A). Accordingly, Entpd4 expression appeared to be specifically upregulated in expression profiles of Ptch1+/− Nos2−/− tumors. QRT-PCR validation confirmed an elevated mean expression in Ptch1+/− Nos2−/− compared to Ptch1+/− Nos2+/+ MBs in those samples that overlapped with the array-CGH analysis but revealed no significant difference across the expanded tumor set (Figure 2C).
As GCPs are considered the cells of origin for the Hh-dependent MB subtype, we examined the effect of Nos2 ablation in the context of cerebellar development. Therefore, gene expression profiles of normal cerebellar tissue samples from postnatal day 9 (P9), 6 weeks after birth (6W), and 1 year of age (1Y) were generated from wild-type, Ptch1+/− Nos2+/+, Ptch1+/− Nos2−/−, and Ptch1+/+ Nos2−/− animals. While specimens of mature cerebellum (6W and 1Y) were investigated separately in 3 biological replicates per genotype and developmental stage, samples of postnatal cerebellum consisted of pooled RNA from 4–5 individuals processed in technical replicates due to limited tissue amounts.
An unsupervised hierarchical cluster analysis of transcriptome data clearly separated developing cerebellum (P9) of wild-type mice and the two Ptch1-mutated genotypes from mature cerebellum. Interestingly, P9 cerebellum of Ptch1+/+ Nos2−/− mice displayed different properties highly similar to mature cerebellum (Figure 3A). Expression profiles of MBs formed a distinct cluster clearly separated from all healthy tissue samples.
A direct comparison between gene expression profiles from Ptch1+/+ Nos2−/− and wild-type P9 cerebellar tissue samples resulted in a total of 984 deregulated genes with 755 genes (76.7%) showing a decreased expression in Ptch1+/+ Nos2−/− mice (Table S4). P9 cerebellum from Ptch1+/− Nos2+/+ and Ptch1+/− Nos2−/− mice revealed only 5 and 32 deregulated genes relative to wild-type, respectively (Table S5 and Table S6). This large deviation of postnatal gene expression in the Ptch1+/+ Nos2−/− genotype included a set of downregulated genes that are essential for proliferation of GCPs (e.g. cyclin D1, cyclin D2 and N-myc, Figure 3B). As hedgehog signaling constitutes the main regulatory pathway for neonatal cell proliferation in GCPs of the EGL, the 984 deregulated genes were analyzed for enrichment of Gli transcription factor targets. Matching this list to a set of recently identified Gli-targets in GCPs [10] yielded a significant overrepresentation of Gli1-regulated genes (p = 0.005, chi-square test). Hence, the reduced transcript levels of these target genes suggests an attenuated hedgehog signaling in postnatal Ptch1+/+ Nos2−/− cerebellum compared to wild-type (or any other genotype).
Notably, the decreased expression of Gli1-targets and proliferation-associated genes observed in Nos2-deficient cerebellar tissue was abolished upon additional inactivation of the hedgehog receptor Ptch1 (in Ptch1+/− Nos2−/− mice). Therefore, we examined the transcript levels of patched receptors themselves in more detail. While neither Ptch1 nor Ptch2 expression was changed between wild-type and Ptch1+/− Nos2+/+ P9 cerebellum, a significant increase of Ptch1 and a minor increase of Ptch2 expression were observed in Ptch1+/+ Nos2−/− mice relative to wild-type mice (Figure 3C). Notably, in Ptch1+/− Nos2−/− cerebellar tissue samples, Ptch2 expression was more elevated than Ptch1. However, since Ptch2 is not capable of inhibiting smoothened (Smoh), it probably failed to take over the attenuating effect on Gli activity [38]. MB specimens from Ptch1+/− Nos2+/+ versus Ptch1+/− Nos2−/− mice showed no significant difference in expression levels of either patched receptor, with Ptch2 being substantially increased over Ptch1 in both groups (Figure 3C). These findings indicate that Nos2 deficiency leads to an upregulation of Ptch1 in GCPs, which results in a downregulation of mitotic genes and Gli-targets only in a Ptch1-wild-type background.
So far, Nos2 inactivation was shown to counteract proliferation and antagonize hedgehog signaling in developing cerebella. To identify those Nos2-dependent effects promoting MB induction, we determined the features that were common to Ptch1+/+ Nos2−/− and Ptch1+/− Nos2−/− genotypes and persisted in the tumor tissues. Accordingly, the overlap of differential gene expression from three comparisons was built: i) Ptch1+/+ Nos2−/− versus wild-type P9 cerebellum, ii) Ptch1+/− Nos2−/− versus Ptch1+/− Nos2+/+ P9 cerebellum; and iii) Ptch1+/− Nos2−/− versus Ptch1+/− Nos2+/+ MB. As a result, only 2 genes were observed to be deregulated in a Nos2-dependent manner during cerebellar development and in MBs (Figure 4A). Although Stmn1 (stathmin 1) appeared to be upregulated in Ptch1+/− Nos2−/− MBs relative to Ptch1+/− Nos2+/+ MBs, this could not be confirmed by qRT-PCR (Figure S3). Gene expression of Gap43 was consistently reduced in Nos2-deficient cerebellar tissue samples and downregulation in Ptch1+/− Nos2−/− tumors relative to Ptch1+/− Nos2+/+ tumors was also significant in the expanded validation set (Figure 4C and 4D). To further assess the immediacy of Nos2 inactivation and Gap43 deregulation, Gap43 transcript levels were determined in expression profiles of healthy cerebella from all developmental stages (P9, 6W and 1Y). Groups for comparison were built according to presence or absence of Nos2, irrespective of the Ptch1 status. The results clearly demonstrated a close association of altered Gap43 transcript levels and Nos2 status (Figure 4B), and indicated downregulation of Gap43 to be the most common effect of Nos2 deficiency in the cerebellum.
To investigate differences in Gap43 expression on protein level in situ we performed immunofluorescent double stainings of Gap43 and the proliferation marker Ki-67 on FFPE sections of P9 cerebella from wild-type, Ptch1+/− Nos2+/+, Ptch+/+ Nos2−/− and Ptch1+/− Nos2−/− mice. As illustrated in Figure 4E, Gap43 immunofluorescence was particularly prominent in the outer region of the molecular layer (ML) that is connected to and partially comprised of radial GCP process extensions. Image quantification further indicate a quantitative difference of Gap43 expression in this region with sections from wild-type and Ptch1+/− Nos2+/+ mice showing a more intense staining than sections from Ptch1+/− Nos2−/− and Ptch1+/+ Nos2−/− mice (Figure 4F).
The association of Nos2 inactivation and decreased Gap43 expression suggests a gene-regulatory function of NO signaling. In order to investigate this possible link in vitro, we used the murine cerebellar precursor cell line c17.2 and the human MB cell line D458 (see Text S2 for details). Both cell lines were treated either with the Nos inhibitor L-NAME (Nω-nitro-L-arginine methyl ester) to reduce NO levels or solvent control (Figure S4). Relative expression of Gap43 was assessed every 24 hours by qRT-PCR. In c17.2 cells, Gap43 transcript abundance was generally low and increased with culture duration. We observed a slightly decreased expression of Gap43 upon L-NAME treatment that reached significance (p = 0.023) after 120 hours (Figure 5A). NOS inhibition in D458 human MB cells resulted in a significant reduction of Gap43 transcript levels starting already after 72 hours with further decrease after 96 hours and 120 hours (Figure 5B). FACS analyses of apoptosis and cell cycle excluded these observations to be attributed to secondary effects of changing cell conditions (Figure S5). These results suggest Gap43 downregulation as a direct consequence of reduced NO levels in murine neuronal precursors and human MB cells.
The dependency of Gap43 expression on NO signaling suggests this gene as key mediator of the effects observed in Nos2-deficient P9 cerebellum and Ptch1+/− Nos2−/− MB, in particular, the upregulation of functional Ptch1 in Ptch1+/+ Nos2−/− mice. Mishra et al. recently reported a central role of Gap43 in the polarization of developing GCPs by regulating centrosome positioning and thus defining correct orientation towards the IGL [39]. Since this is a prerequisite for directed migration, reduced levels of Gap43 in P9 cerebellar tissue may lead to retention of GCPs in the EGL. To test these hypotheses, shRNA-mediated knockdown of Gap43 was performed in c17.2 cells (see Text S1 for details). Upon knockdown of Gap43 we observed a strong inverse behavior of Ptch1 and Gap43 transcript levels (Figure 5C). Changes in migration characteristics were assayed in a Boyden chamber using recombinant SDF-1α (CXCL12) as chemoattractant, which was reported to participate in guiding migration of embryonal GCPs in vivo [40]. Downregulation of Gap43 yielded a significant decrease in cell migration between 14% (p = 0.013) and 20% (p = 0.007) (Figure 5D; Figure S6C). A pseudo-effect of the knockdown due to altered proliferation of the v-myc-immortalized c17.2 cells was excluded by FACS-based cell cycle analysis (Figure S7).
Transcriptome and functional analyses suggest that a decreased Gap43 expression accounts for Ptch1 upregulation and impairment of directed neuronal precursor migration in vitro. Accordingly, Ptch1+/+ Nos2−/− P9 cerebella are supposed to increasingly retain GCPs with reduced mitotic activity in the EGL compared to wild-type and Ptch1+/− Nos2+/+ mice. Moreover, the Ptch1+/− Nos2−/− genotype is also expected to exhibit retention of GCPs, but not to show any cell cycle arrest. To further verify this hypothesis in situ we performed immunofluorescent double staining of proliferating (Ki-67+) and post-mitotic GCPs on FFPE sections of postnatal cerebellum (Figure 6A). Here, post-mitotic cells were delineated by the neuronal marker NeuN (neuronal nuclear antigen A60) [41]. At least three different regions of each mouse cerebellum were analyzed in three to four animals per genotype using confocal laser scanning microscopy. In accordance with the microarray data, averaged cell counts of wild-type and Ptch1+/− Nos2+/+ mice did not show significant difference. In contrast, an increase of post-mitotic GCPs (NeuN+, Ki-67−) was detectable in the EGL of Ptch1+/− Nos2−/− and Ptch1+/+ Nos2−/− mice (Figure 6C). Concurrently, the ratio of dividing to non-dividing GCPs was similar in Ptch1+/− Nos2−/−, wild-type and Ptch1+/− Nos2+/+ P9 cerebella but markedly decreased in Ptch1+/+ Nos2−/− mice. This recapitulated the downregulation of mitotic genes observed in the expression profiles. However, the total amount of proliferating GCPs per EGL section was significantly higher in Ptch1+/− Nos2−/− mice compared to any other genotype (Figure 6C). These results demonstrate a tissue phenotype that corresponds to the effects of reduced Gap43 in developing cerebellar neuronal precursors (in vitro). The increased accumulation of proliferating GCPs in the EGL observed in the Ptch1+/− Nos2−/− genotype supposedly leads to a larger pool of cells susceptible to neoplastic transformation and is therefore likely to promote medulloblastoma development.
The Ptch1+/− MB mouse model has been intensively studied and has greatly contributed to our understanding of Hh-dependent MB tumorigenesis in the context of cerebellar development. The data presented here indicate a role of Nos2 and hence NO signaling in Hh-dependent MB by demonstrating a significantly increased MB rate in Ptch1+/− Nos2−/− mice compared to Ptch1+/− Nos2+/+ mice. The global genome-wide screens performed in the present study did not reveal obvious molecular differences between MBs in Ptch1+/− Nos2+/+ versus Ptch1+/− Nos2−/− animals. Assessment of genomic alterations using array-CGH identified trisomy of chromosome 6 as a recurrent feature in tumors of both genotypes. This corresponds to a recent report on MBs of the same molecular subtype with inactivated double-strand break repair proteins targeted to neuronal progenitors of p53−/− mice [42]. The most common loss identified in our analyses affected two small regions on chromosome 13 encompassing the Ptch1 gene and possibly indicate acquired homozygosity for the mutant allele or somatic rearrangements rather than a broad deletion of the locus. Targeted duplex PCR further confirmed loss of the functional wild-type allele to be a frequent event in these MBs. Notably, tumors of the Ptch1+/− Nos2−/− genotype showed a higher frequency of a small gain on chromosome 14. The affected Entpd4 gene encodes for an apyrase located at the internal membrane of lysosomal vacuoles and the Golgi apparatus. It preferentially catalyzes the hydrolysis of UDP to UMP [43] and thereby facilitates the inverse directed import of UDP-GlcNAc [44]. This in turn was reported to increase glycosylation of surface receptors (e.g. EGFR and PDGFR) and foster cell growth [45]. According to the microarray and qRT-PCR expression data, Entpd4 transcript levels were indeed increased in tumors with this chromosomal gain. However, this effect did not turn out to be Nos2-dependent in an expanded sample set. Consequently, Entpd4 likely plays a role in MB pathogenesis but is not directly linked to loss of Nos2.
The examination of tumor-relevant changes in developing cerebellum as a consequence of impaired Nos2 activity and hence NO signaling surprisingly revealed a decreased proliferation of GCPs in the cerebellum of Ptch1+/+ Nos2−/− mice. The concurrent upregulation of Ptch1 and the significant enrichment of downregulated Gli1-target genes strongly suggest that this effect is a consequence of reduced hedgehog signaling. Moreover, this phenotype was completely abrogated by a concomitant Ptch1 mutation. The slight increase of Ptch2 in Ptch1+/− Nos2−/− cells points to a compensatory effect and further supports the notion of an inhibitory function of Nos2 loss on the hedgehog pathway in postnatal cerebellum. Since neither a Smoh-regulating domain [38] nor a function for cell cycle arrest through seizing cyclin B1 [46] were reported for Ptch2, its upregulation may be insufficient for preventing MB induction. In contrast to these observations, Ciani et al. demonstrated that proliferation of cultured GCPs increased upon withdrawal of NO and that this effect was mediated by augmented N-myc levels [37]. However, N-myc was not differentially expressed between Ptch1+/− Nos2−/− and Ptch1+/− Nos2+/+ MBs of our series. A possible explanation for this discrepancy might be an unrecognized heterogeneity in the isolated cerebellar cell population used in the Ciani study. Since eNos and nNos are known to attenuate the mitotic activity of subventricular neuronal stem cells [47], [48] Nos inhibitor treatment possibly resulted in a selective growth advantage over GCPs. Downregulation of Gap43 was the only feature observed in Nos2-deficient versus Nos2-proficient postnatal cerebella irrespective of the Ptch1 status. This difference was also conserved between Ptch1+/− Nos2−/− and Ptch1+/− Nos2+/+ MBs, and particularly visible in outer regions of the molecular layer, where maturating GCPs of the EGL develop contact forming projections prior to radial migration. Other studies already suggested a link between Gap43 mRNA levels and NO signaling due to co-induction of nNos and Gap43 during axon regeneration and reactive synaptogenesis following injury of spinal motoneurons [49] and sensory neurons [50]. Furthermore, a downregulation of Gap43 was reported after silencing of soluble guanylate cyclase subunits, the central elements of cGMP-mediated NO signaling [51]. Finally, the present study demonstrates Gap43 downregulation to be a consequence of NO withdrawal in neuronal progenitors and MB cells. A possible mechanism for this regulation refers to decreased protein levels of the poly(U)-binding and degradation factor AUF1 upon NO-dependent cGMP production [52]. AUF-proteins generally bind to AU-rich elements of the 3′UTR (untranslated region) of coding transcripts and associate with proteins of the ELAV-like family to control gene expression via mRNA decay [53]. Tsai et al. demonstrated that Gap43 mRNA levels are post-transcriptionally regulated during neuronal differentiation and that elements of the 3′UTR confer transcript instability, which is abolished upon TPA treatment (inter alia inducing NOS2) [54]. At the same time, Chung et al. demonstrated that indeed ELAV-like family member HuD was binding to 3′UTR regions of GAP43 [55]. Taken together, NO accumulation possibly decreases cellular levels of mRNA-destabilizing AUF1 protein and thus might contribute to a high transcript abundance of Gap43.
Gap43 is a membrane-anchored protein at the cytoplasmic side of neuronal cell projections and found to be highly expressed during development of the CNS [56]. It is particularly localized in axonal growth cones and participates in the coordination of extrinsic stimuli and intrinsic cell remodeling [57] by regulating cytoskeleton dynamics [58]. Granule cell (GC) migration follows a sequence of tangential and radial movements controlled by successive formation of leading projections [59]. As maturating GCPs exit cell cycle, positioning of the centrosome determines the site of axon growth cone emergence and thus neuronal polarity including localization of such projections [60]. This defines the structural orientation of GCPs in terms of directing its dendrite to descend across the molecular and Purkinje cell layers to populate the IGL. However, centrosome positioning and therefore accurate polarization of GCPs require phosphorylated Gap43 to bind to the centrosome-associated microtubule-organizing center [61]. Hence, inaccurate GCP migration was observed in Gap43−/− animals [39], a finding that is in full agreement with our data from the functional Gap43 knockdown assays. Downregulation of Gap43 in Nos2-deficient P9 cerebellum therefore likely mediates the retention of GCPs observed in FFPE sections. Accordingly, NO/cGMP signaling was demonstrated to be crucial for accurate migration of the neuronal precursor cell line NT2 [62]. Furthermore, slice culture experiments of neonatal cerebella (P9) exhibited a substantial reduction of proliferation and migration of maturating granule cells to the IGL upon application of NO synthases inhibitors [63]. The elevation of Ptch1 levels upon Gap43 reduction in vitro fits to the data by Shen et al. who reported an upregulation of Ptch1 gene expression in inner EGL regions of Gap43−/− mice compared to wild-type animals. Moreover, cultured Gap43-deficient GCPs show decreased proliferation in response to administered recombinant Shh protein [64]. A possible regulatory link was recently provided as the activation of the hedgehog signaling component Smoh was found to depend on PI4P (phosphatidylinositol 4-phosphate) levels that immediately increase when Shh binds to Ptch1 or when functional Ptch1 is absent [65]. The authors further showed that imbalanced conversion of the precursor molecule PI into PI4P influences hedgehog pathway activity. Alternatively, the production of PI4P can also result from a specific dephosphorylation of PI(4,5)P2 [66]. In this context, Gap43 protein was recently demonstrated to build oligomeric structures in the plasma membrane which sequester specifically PI(4,5)P2 [67]. A similar finding has been reported earlier showing that GAP43 participates in the accumulation of plasmalemma rafts, which promoted retention of PI(4,5)P2 [68]. The amount of Gap43 associated with the plasma membrane therefore possibly modulates the utilization of PI(4,5)P2, including its conversion into PI4P, which in turn directly affects hedgehog signaling through Smoh activation. However, the effective impact on downstream Gli-targets would still be difficult to conclude regarding the multitude of responses to Shh, including negative feedback regulation [13]. Further studies applying depletion and enrichment of specific phosphatidyl derivatives and selective silencing of hedgehog pathway elements will be necessary to elucidate the molecular nature of this proposed signaling axis.
The increased accumulation of mitotic granule cells at the EGL seen in the combined Ptch1+/− Nos2−/− genotype supposedly gives a special clue to MB induction. In contrast to the classical view of neonatal EGL organization, which describes radial migration of granule cells to follow a proliferation stop, more and more evidence arises showing that cell cycle arrest is not a prerequisite for migration but rather occurs during a temporally coordinated interplay of gene expression patterns [69]. This corroborates our data shown in Figure 6 (white arrows), where proliferation is still maintained in migrating granule cells and even in cells of the IGL of wild-type cerebellum. Regulation of such expression patterns is largely dependent on Shh stimuli being most intensive in the EGL [70], as well as gradients of other soluble factors such as Bmps, which were reported to account for a regulatory environment along the transition through the cerebellar layers [71]. Further evidence for a niche-like-concept was provided by Choi et al. in Bdnf−/− mice that displayed a severe retardation of GCP migration [16]. The authors could demonstrate that mitotic activity of maturating GCPs was significantly enhanced when cells were retained in the EGL and declined with increasing distance from outer EGL regions. Therefore, the accumulation of GCPs in the EGL in combination with the insensitivity to Ptch1-mediated cell cycle arrest in Ptch1+/− Nos2−/− mice provide a growth advantage and increase the number of putative transformation targets over Ptch1+/− Nos2+/+ mice (Figure 7C).
In conclusion, the following picture emerged from our data: Homozygous deletion of Nos2 leads to a reduction of basic NO levels in immature GCPs of the EGL during postnatal development of the cerebellum. This reduction causes a downregulation of Gap43 expression, which results in an increased expression of Ptch1 and impaired directed migration of maturating GCPs. As a consequence, undifferentiated granule cell precursors exit cell cycle and are retained at the EGL (Figure 7B). In case of an additional heterozygous Ptch1 mutation, upregulation of this receptor does not suffice to exert the anti-proliferative stimulus following Gap43 decrease, which results in an increased fraction of continuously dividing cells in the EGL (Figure 7C). As reduced migration towards the IGL further leads to a withdrawal of growth-limiting signals, expansion of the GCP population is additionally supported. Finally, this advances medulloblastoma development in Ptch1+/− Nos2−/− mice compared to Ptch1+/− Nos2+/+ mice. The mechanism described here illustrates a new tumor-promoting concept in MB showing that the localization of pre-neoplastic cells within the developing cerebellum is important for pathogenesis.
Ptch1+/− mice (B6; 129P2-Ptch1tm1Mps/Ptch+; [20]) and Nos2−/− mice (B6;129P2-Nos2tm1Lau; [72]) were obtained from the Jackson Laboratory (Bar Harbor, Maine, USA) and crossbred to generate double heterozygous mice (Ptch1+/− Nos2+/−). The F1 hybrids were backcrossed with Ptch1+/+ Nos2−/− mice to generate Ptch1+/− Nos2−/− mice. Later on, Ptch1+/− Nos2−/− mice were directly mated. For details on housing and genotyping see Text S2 and Table S9. All animal experiments were approved by the responsible federal authorities (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Recklinghausen, Germany, Az. 50.05-230-17/06).
Total RNA of tumor specimens and normal cerebellar tissue samples was isolated via CsCl density gradient centrifugation [73] and assessed for integrity using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, USA). For gene expression microarrays, linear amplification of mRNA and labeling of cDNA were conducted on samples and Mouse Universal Reference RNA (Stratagene, La Jolla, USA) according to the TAcKLE protocol [74]. Both were combined for two-color hybridizations with each sample being performed as two replicates of inverse dye orientation. Global gene expression profiling was performed on self-printed oligonucleotide microarrays. Further details of microarray production and hybridization are described in Text S2. Genomic DNA of tumor specimens was isolated from the interphase of the CsCl gradient by ethanol precipitation, proteinase K digest, and phenol/chloroform extraction. DNA samples were monitored for purity and adequate fragment size using spectrophotometric measurements and gel electrophoresis. Array-based comparative genomic hybridization (array-CGH, matrix-CGH, [75]) was performed on Mouse Genome CGH 244 k Microarrays (Agilent). Cy5-labeled tumor DNA was combined with corresponding Cy3-labeled reference (wild-type genomic DNA) to receive either sex-matched sample pairs or pairs of different gender for internal negative or positive control. Sample preparation, microarray hybridization, and washing procedures were carried out as described in the manufacturer's protocol. Microarray data are available in GEO (http://www-ncbi.nlm.nih.gov/geo), under accession number GSE29201.
Total RNA isolated from cell culture samples using the RNeasy Mini Kit (Qiagen, Hilden, Germany) or RNA from tissue specimens was subjected to oligo(dT)-primed reverse transcription. QRT-PCR measurements were conducted in an ABI PRISM 7900HT thermal cycler (Applied Biosystems, Foster City, USA) using the SYBR green reaction and detection system (ABgene, Epsom, UK). For relative quantification mean ratios were calculated between genes of interest and a set of five housekeeping genes (Table S7) according to the Pfaffl method [76].
The expression levels of Ptch1, Gli1, N-myc, and Nos2 were determined by real-time reverse transcription PCR analysis using the ABI PRISM 5700 system (Applied Biosystems) as reported before [73]. For these experiments, the mRNA expression level of mitochondrial ribosomal protein L32 (Mrpl32) served as housekeeping reference. All primer sequences are depicted in Table S8.
Inhibition of NO synthases was performed in c17.2 and D458 cells which were seeded at densities of 2×105 and 4×105 cells per well in 12-well plates, respectively. Cells were daily treated with either 1 mM of the inhibitor L-NAME or 1× PBS as solvent control. For knockdown experiments of Gap43, c17.2 cells were grown in a 12-well plate to 80% confluency and transfected with 2 µg of pLKO.1-puro vector that contained either shRNA constructs targeting Gap43, shRNA against GFP, or non-target shRNA as a control (Sigma-Aldrich, St. Louis, USA) using 9 µl FuGene HD reagent (Roche, Basel, Switzerland). Transfection was repeated 2 times each after 8 hours and subjected to selection conditions (1 µg/ml puromycin) for 24 hours. Subsequently, cells were trypsinized, adjusted to 4×105 cells/ml and seeded into the inserts of a Costar Polycarbonate Membrane Transwell plate (8 µm pores, Corning, USA). After 24 hours cells were either harvested for gene expression and protein analyses or 0.1 µg/µl recombinant SDF-1α was applied to the lower compartment for migration assays. Following 12 hours of incubation, cells at the bottom of the insert membrane were methanol-fixed, hematoxylin-stained, and counted.
FFPE sections of postnatal cerebella were pre-processed as described in Text S2. For immunofluorescence co-staining, Gap43 (Sigma-Aldrich, clone GAP-7b10) or NeuN (Millipore, clone A60) first primary antibodies were diluted 1∶1000 or 1∶200, respectively and applied using the Dako REAL Detection System (Dako, Glostrup, Denmark). Following over night incubation at 4°C, washing in TBS, and blocking of residual biotin/streptavidin, sections were subsequently incubated with biotinylated anti-mouse secondary antibody (Dako) and stained with 20 ng/µl FITC-conjugated streptavidin (Invitrogen, Carlsbad, USA). The second primary antibody against Ki-67 (Novocastra, Wetzlar, Germany) was diluted 1∶1000 and accordingly applied using biotinylated anti-rabbit secondary antibody (Dako) and 20 ng/µl Cy5-conjugated streptavidin (Invitrogen). Co-stained sections were then covered with DAPI-containing VECTASHIELD Mounting Medium (Vector, Burlingame, USA) and subjected to confocal laser scanning microscopy.
Quantification of Gap43 staining was performed for areas of interest using Image J software (NIH). Numbers of dividing and non-dividing cells in the EGL of postnatal cerebellar tissue sections were counted manually and normalized to the corresponding length of the EGL edge. Cell counts for each region were averaged across three sections, each with 10–20 µm distance in z-axis, and per individual.
Kaplan-Meier survival plots were calculated for a total of 1167 mice, including 315 wild-type mice, 412 Ptch1+/+ Nos2−/− mice, 215 Ptch1+/− Nos2+/+ mice and 221 Ptch1+/− Nos2−/− mice. MB-free survival was plotted using the GraphPad Prism 5 software (GraphPad, La Jolla, USA). The logrank test was applied to compare survival (tumor occurrence) of the different genotypes. For comparisons of differences of means between two groups of replicates, p-value calculations were performed using an unpaired, two-tailed t-test, unless indicated otherwise. Calculated error bars represent the standard error of the mean (SEM). For details on microarray statistics please refer to Text S2.
|
10.1371/journal.ppat.1002696 | Analysis of Functional Differences between Hepatitis C Virus NS5A of Genotypes 1–7 in Infectious Cell Culture Systems | Hepatitis C virus (HCV) is an important cause of chronic liver disease. Several highly diverse HCV genotypes exist with potential key functional differences. The HCV NS5A protein was associated with response to interferon (IFN)-α based therapy, and is a primary target of currently developed directly-acting antiviral compounds. NS5A is important for replication and virus production, but has not been studied for most HCV genotypes. We studied the function of NS5A using infectious NS5A genotype 1–7 cell culture systems, and through reverse genetics demonstrated a universal importance of the amphipathic alpha-helix, domain I and II and the low-complexity sequence (LCS) I for HCV replication; the replicon-enhancing LCSI mutation S225P attenuated all genotypes. Mutation of conserved prolines in LCSII led to minor reductions in virus production for the JFH1(genotype 2a) NS5A recombinant, but had greater effects on other isolates; replication was highly attenuated for ED43(4a) and QC69(7a) recombinants. Deletion of the conserved residues 414-428 in domain III reduced virus production for most recombinants but not JFH1(2a). Reduced virus production was linked to attenuated replication in all cases, but ED43(4a) and SA13(5a) also displayed impaired particle assembly. Compared to the original H77C(1a) NS5A recombinant, the changes in LCSII and domain III reduced the amounts of NS5A present. For H77C(1a) and TN(1a) NS5A recombinants, we observed a genetic linkage between NS5A and p7, since introduced changes in NS5A led to changes in p7 and vice versa. Finally, NS5A function depended on genotype-specific residues in domain I, as changing genotype 2a-specific residues to genotype 1a sequence and vice versa led to highly attenuated mutants. In conclusion, this study identified NS5A genetic elements essential for all major HCV genotypes in infectious cell culture systems. Genotype- or isolate- specific NS5A functional differences were identified, which will be important for understanding of HCV NS5A function and therapeutic targeting.
| Hepatitis C virus (HCV) is a major public health burden and leads to chronic liver disease, including liver cirrhosis and liver cancer. Understanding the biological functions of the virus is crucial to the development of a vaccine and to improve current therapy through development of directly-acting antiviral compounds. The NS5A protein is a promising antiviral target, but much remains to be understood about its role in the viral life cycle. Great diversity among the seven major HCV genotypes poses challenges for broadly active inhibitors. Here we used infectious cell culture systems for NS5A of the seven major HCV genotypes, and demonstrated that all genotypes depended on the NS5A amphipathic alpha-helix, domain I, low-complexity sequence (LCS) I and domain II for viral replication. Interestingly, effects on replication and virus production by changes in LCSII and domain III varied greatly among NS5A isolates. Furthermore, we found that genotype 2 had evolved genotype-specific residues in domain I of importance for viral function. Thus, the highly diverse sequence of the NS5A protein reflected functional differences between HCV genotypes and isolates. Such differences will be important to consider in understanding HCV biology and for future development of antiviral compounds.
| Hepatitis C virus (HCV) chronically infects 130–170 million people and leads to increased risk of severe liver disease. HCV belongs to the Flaviviridae family and has a positive-strand RNA genome containing one long open reading frame (ORF). The ORF encodes a polyprotein that is co- and post-translationally cleaved into structural proteins (Core, E1, E2), p7 and nonstructural (NS) proteins NS2, NS3, NS4A, NS4B, NS5A and NS5B. Significant diversity is found among HCV isolates, which in phylogenetic analysis cluster into seven major genotypes and numerous subtypes [1], [2]. Genotypes, subtypes and isolates/strains differ at around 30%, 20% and 2–10%, respectively, at the nucleotide and amino acid level. A higher variability is found in certain genome regions. Among different genotypes, the NS5A protein sequence varies up to 50% in composition and by more than 20 residues in length. Important differences between HCV genotypes were identified in biology [3] and in sensitivity to neutralizing antibodies [4]–[7]. The HCV genotype is an important determinant for response to the current interferon (IFN)-α based treatment regimens; sustained virological response is achieved for 80–90% of genotype 2 and 3 and for around 50% of genotype 1 and 4 infected patients [8]. Several HCV genes, including E2, NS3 and NS5A, were suggested to influence the response to IFN [9]. Directly-acting antiviral compounds are currently being developed with the NS3 protease, the NS5B polymerase and NS5A as primary targets [10]. Genotype- and isolate-specific responses to treatment with directly-acting antivirals have been reported in vitro [11], [12] and in clinical trials [13], [14].
The NS5A phosphoprotein is a component of the viral replication complex [15] and consists of three domains separated by low-complexity sequences (LCS) [16]. It is anchored to intracellular membranes through the N-terminal amphipathic alpha-helix [17], [18]. A crystal structure was solved for domain I [19], [20], which contains a zinc-binding motif [16] and a highly basic channel with RNA binding capacity [21]. The amphipathic alpha-helix, domain I and II are essential for the genotype 1b replicon system [16], [22], [23]. In LCSI and domain II, covering parts of the binding site for the IFN-induced dsRNA-dependent protein kinase R (PKR) [24], an IFN-sensitivity determining region (ISDR) was described for genotype 1b [25], although, its existence is controversial [26]. Several studies on genotype 2a suggest a primary role for domain III in production of infectious particles [27]–[29]. For most genotypes, however, the role of NS5A in the viral life cycle has not been studied.
Until development of the genotype 2a JFH1 cell culture system [30], and the more efficient J6/JFH1 system with the Core-NS2 region from another 2a isolate [31], studies on HCV NS5A relied on the genotype 1b and 2a replicon systems, recapitulating only parts of the viral life cycle [9]. J6/JFH1-based infectious cell culture systems for NS5A genotypes 1–7 [12] allowed us to study the function of the highly variable NS5A protein for all major HCV genotypes in context of the complete viral life cycle. We analyzed individual NS5A domains for their influence on steps of the viral life cycle and found genotype- and isolate-specific effects of introduced mutations and modifications.
To determine the importance of the individual NS5A domains for the major HCV genotypes in context of the complete viral life-cycle, we introduced selected mutations (Figure 1A) into J6/JFH1-based NS5A genotype 1–7 recombinants, which we previously demonstrated were infectious and efficient in vitro (Materials & Methods) [12]. In previous studies, it was found that the Con-1(genotype 1b) replicon system was inhibited by the I12E mutation shown to disrupt the hydrophobic face of the amphipathic alpha-helix [22], by C57G and C59G mutations interfering with zinc-ion binding of domain I [16], and by mutating the conserved W329 in domain II [23] (Numbering throughout is according to individual H77 reference proteins, GenBank accession number AF009606). The S225P mutation in LCSI was shown not to be permissible in vivo when introduced into the Con-1(1b) full-length infectious clone [32], even though it had an enhancing effect on the Con-1(1b) replicon system [33], [34]. These residues were all highly conserved among HCV patient isolates (Figure 1B). To analyze the effect of these mutations in cell culture, RNA transcripts of more than 30 mutants of the H77C(1a), J4(1b), JFH1(2a), S52(3a), ED43(4a), SA13(5a), HK6a(6a) and QC69(7a) NS5A recombinants were transfected into Huh7.5 cells in parallel with the positive control J6/JFH1 (for consistency denoted JFH1(2a) according to the NS5A isolate) [31]. While around 30% of cells were HCV positive one day after JFH1(2a) transfection, the I/V12E, C57G/C59G, S225P and W329A genotype 1–7 mutants were all highly attenuated, as no or very few HCV positive cells were observed in immunostainings one and three days post-transfection.
Since reversion was observed in vivo for the Con-1(1b) S225P mutant coinciding with detection of high viral titers [32], we followed the J4(1b) S225P mutant until it, after two weeks, infected most cells. In virus recovered after passage to naïve cells, S225P had reverted. Thus, for this mutant the findings in the infectious cell culture system were in accordance with findings in vivo but not with findings in the replicon system [33], [34]. To determine whether another mutant with a replicon-enhancing alteration was also attenuated we changed the J4(1b) NS5A recombinant to encode the Δ47 deletion, which replaces residues 235–282 in NS5A LCSI/domain II by a single tyrosine (Figure 1A) [33]; this mutant was followed for three weeks without detection of HCV positive cells. Thus, the positive effect of NS5A replicon-enhancing mutations apparently led to the opposite effect in the infectious cell culture system.
We further investigated whether the phenotype observed for the highly attenuated NS5A mutants corresponded to that of mutants of other HCV nonstructural proteins previously shown to abrogate HCV RNA replication. No HCV positive cells were observed one day after transfection of JFH1(2a) mutants of the NS3 protease active site (NS3pro−, S139A [35]), the NS3 helicase active site (NS3hel, D290A [35]), the NS4A transmembrane segment (G21V [36]), the NS4B C-terminal end (W252S [37]), the NS5B polymerase active site (GND-mutant, D318N [38]), for a JFH1(2a) mutant with a stop codon in the NS3 N-terminus (Y6[stop]), or for the JFH1(2a) NS5A domain I mutants C57G, C59G or C57G/C59G included for comparison. However, infection emerged subsequently for the NS3hel, NS4BW252S, NS5AC59G, NS5BGND and NS3Stop mutants when following cultures for more than two weeks. Reversion of the knockout mutations and the presence of silent marker mutations engineered to exclude contamination were confirmed by sequencing. We found that in a total of 13 experiments done with NS5BGND, 10 led to emerging infection. Viruses recovered from five of these cultures were subsequently sequenced, in all of which the attenuating mutation had reverted; the marker mutations were maintained. Thus, the phenotype of attenuated NS5A mutants corresponded to that of single mutations of other non-structural genes expected to be detrimental for viral replication. For these mutants, reversion was detected in most cases. We speculate that this was due to transfected RNA pools containing genomes that were reverted to wild-type as a result of the high error-rate of T7 polymerase driven in vitro RNA transcription.
The W329 in the C-terminal end of NS5A domain II was found to be of critical importance for replication of NS5A genotype 1–7 recombinants, while it was previously reported that a JFH1 NS5A mutant with a deletion of residues 246–308 covering a large N-terminal region of domain II (Δ2222–2280 mutant) apparently was fully viable [27]. We had similar findings for a J6/JFH1 virus without NS5A residues 250–293 [39], and therefore wanted to investigate whether deletion of the corresponding residues was generally permitted for NS5A isolates. Thus we constructed H77C(1a) and TN(1a) Δ250–293 mutants (Figure 1A and S1) and transfected RNA transcripts into Huh7.5 cells. Infectivity titers after transfection were reduced for these mutants, most significantly for H77C(1a); as previously observed, titers for the JFH1(2a) mutant were not decreased (Figure 2A). This corresponded to immunostainings with 20–30% HCV positive cells one day post-transfection, except for the H77C(1a) Δ250–293 mutant, for which only 5% were positive. All recombinants spread to the majority of cells within 8 days. Analysis of NS5A of recovered passaged viruses of JFH1(2a), H77C(1a) and TN(1a) Δ250–293 mutants confirmed the deletion. Amino acid changes D444G (D2416G) and C447R (C2419R) were identified in H77C(1a) and TN(1a) Δ250–293, respectively. No NS5A changes were observed for the JFH1(2a) mutant.
To investigate whether viral replication, assembly or release was affected for H77C(1a) and TN(1a) Δ250–293 mutants, we transfected CD81-deficient Huh7-derived S29 cells that are not susceptible to HCV infection [40]. HCV Core levels were used as a measure for RNA replication [41]; validation experiments confirmed the correlation between intracellular HCV RNA and Core levels (Figure S2). From separate cultures, assembly and release was evaluated by titration of intra- and extra-cellular infectivity. While replication and virus production was not affected for the JFH1(2a) mutant, minor decreases in replication levels and intracellular virus production was observed for TN(1a) (Figure 2B and C). The H77C(1a) mutant exhibited greater reductions, in particular for intra- and extra-cellular virus production that was reduced more than 10-fold. Thus, while the C-terminal region of domain II was critical for replication of all HCV genotypes, deletion of the N-terminal residues 250–293 led to a highly attenuated phenotype for H77C(1a), an intermediate phenotype for TN(1a) and no attenuation of JFH1(2a).
The NS5A LCSII region was shown to influence HCV replication and virus production [23], [42]. The Con-1(1b) replicon system was reported to be highly attenuated by the P343A LCSII mutation [23]. However, changes at this genotype-specific position (P343A for genotype 1 and 5, A343P for other genotypes, Figure S1) did not influence infectivity titers after transfection of Huh7.5 cells for any of the NS5A genotype 1–7 recombinants (Figure 3A and data not shown). No changes were identified in the complete ORF of the recovered JFH1(2a) A343P mutant, and residue 343 had not reverted for mutants of other NS5A isolates. This was in agreement with findings in a more recent Con-1(1b) replicon study [42].
In addition, we generated the JFH1(2a) P346A/P351A/P354A and P346A/P351A/P353A/P354A/P355A LCSII mutants, in which highly conserved prolines (Figure 1B and S1) in two putative SH3 interaction domains [42]–[44] were mutated. Infectivity titers after transfection of Huh7.5 cells were decreased <10-fold (Figure 3A), and the mutations did not revert after passage to naïve cells. Next, the P346A/P351A/P354A mutations were introduced into the NS5A genotype 1–7 recombinants, and tested for viability in Huh7.5 cells. While supernatant HCV infectivity titers were only slightly decreased for the J4(1b) mutant, reductions by 10-fold or more were observed for H77C(1a), S52(3a), ED43(4a), SA13(5a), HK6a(6a) and QC69(7a) mutants (data not shown). To analyze whether attenuation was caused by reduced replication capacities, we measured intracellular HCV Core after transfection of S29 cells. RNA replication of ED43(4a) and QC69(7a) P346A/P351A/P354A mutants was highly attenuated, with intracellular HCV Core levels similar to those observed for the NS5BGND mutant (Figure 3B). These findings were in agreement with immunostainings, which on day 1 after transfection of Huh7.5 cells were HCV negative for the ED43(4a) and QC69(7a) mutants. For all other LCSII mutants, replication was decreased with up to 10-fold reductions in intracellular HCV Core levels 48 and 72 hours post-transfection. While intra- and extra-cellular infectivity titers from transfected S29 cells were only slightly decreased for the J4(1b) and JFH1(2a) mutants (Figure 3C), greater than 10-fold reductions were observed for the H77C(1a), S52(3a), SA13(5a) and HK6a(6a) mutants. Thus the P346A/P351A/P354A mutations had a profound impact on intracellular virus production for some NS5A isolates but not for others. No infectivity was observed for the ED43(4a) and QC69(7a) mutants, as expected from RNA replication assays. Thus, various NS5A isolates tolerated changes in LCSII to different extents. The ED43(4a) and QC69(7a) mutants were highly attenuated in viral replication compared to other isolates, while assembly of intracellular infectious viral particles was affected by LCSII mutations at an isolate-specific level.
In genotype 2a studies, NS5A domain III was reported to have a primary role in virus production, e.g. deletion of JFH1 domain III significantly attenuated virus production while replication was less affected [23], [27], [28]. The high degree of sequence conservation of residues 414-428 (Figure S1) suggested an important function of this region; thus we generated deletion mutants for the NS5A genotype 1–7 recombinants (Figure 1A). While the JFH1(2a) Δ414-428 mutant was not attenuated, infectivity titers after transfection of Huh7.5 cells were decreased up to 10-fold for deletion mutants of other NS5A isolates (Figure 4A). This correlated with previous reports on a 382–428 JFH1 deletion mutant [27]. After spread of infection in culture and passage to naïve cells, sequencing of the ORF from recovered viruses confirmed the deletion for all recombinants and identified changes for the H77C(1a), TN(1a), S52(3a), SA13(5a) and QC69(7a) deletion mutants, primarily in the NS5A region downstream of the deletion and in p7, NS2 and NS3 (Table 1).
To analyze whether attenuation of Δ414-428 mutants was caused by reduced replication capacities, we measured intracellular HCV Core after transfection of S29 cells. For most mutants, replication was decreased less than 2-fold, but greater than 3-fold reductions in intracellular Core were observed at multiple time points after transfection with H77C(1a), TN(1a) and S52(3a) mutants (Figure 4B). The effect of deletion mutants on intra- and extra-cellular infectivity after transfection of S29 cells corresponded to observations from Huh7.5 culture supernatant, with greater than 3-fold reductions in titers for the H77C(1a), J6(2a), S52(3a), ED43(4a) and SA13(5a) recombinants; in this assay JFH1(2a) and QC69(7a) were not affected by the deletion (Figure 4C). For most recombinants, reductions in replication capacities could explain the decreases in intra- and extracellular titers observed. However, the less than 2-fold decrease in replication capacity for the ED43(4a) and SA13(5a) mutants is not likely to explain the greater than 3-fold reductions in intra- and extracellular titers observed. This indicated an effect on virus assembly in addition to the effect on replication for certain NS5A isolates.
To determine whether the genotype 2-specific 20 residue insertion immediately downstream of residue 428 (Figure 1A and S1) could be responsible for the efficient replication and virus production of the JFH1(2a) Δ414-428 mutant (Figure 4), we replaced the 414-428 region of the H77C(1a) recombinant by the JFH1-specific insertion. For this mutant, “Δ414-428+20aa”, replication capacity was slightly decreased, but unlike for the H77C(1a) Δ414-428 mutant, no attenuation of intra- or extracellular infectivity was observed and no additional mutations were identified after passage to naïve cells. Thus, the genotype 2-specific insertion could potentially compensate for the deletion of residues 414-428.
Serines in domain III play an important role for genotype 2a NS5A function [28], [29], [45]; in particular the S437A mutation (Figure 1) was reported to attenuate J6/JFH1 virus production [28]. After transfection of Huh7.5 cells with RNA transcripts of S437A mutants of NS5A genotype 1–7 we did, however, not identify significant attenuation of virus production (Figure 5 and data not shown). No changes were identified in the ORF of the recovered JFH1(2a) S437A mutant and residue 437 had not reverted for mutants of other NS5A isolates.
We further analyzed the conserved arginine/lysine motif at residue 356–359 and the partially conserved 363–380 region that covers the completely conserved residues 363, 372 and 376. Huh7.5 cells were transfected with JFH1(2a) mutants containing point mutations or a deletion of the 363–380 region (Figure 1 and S1). A minor decrease in infectivity titers was observed for the R356A/R357A/R358A/R359A mutant, but not for L363A/L372A, L363H/L372H or Δ363–380 mutants (Figure 5). We confirmed the presence of point mutations and the 363–380 deletion in recovered virus after passage to naïve Huh7.5 cells. Thus, no major effect of the introduced mutations was observed, even after deletion of a larger conserved region of NS5A domain III.
NS5A domain III was previously reported to be important for efficient virus production of genotype 2a [27], [29]. The present findings indicate that domain III is also of importance for replication, with isolate-specific dependence on the highly conserved 414-428 region for replication and virus assembly.
To further investigate the functional impact of mutations in LCSII and domain III, we analyzed the potential effect on NS5A stability. Due to limitations in available NS5A antibodies recognizing domain III deletion mutants of the different isolates, we focused this analysis on H77C(1a) and the mutants P346A/P351A/P354A in LCSII as well as Δ414-428 and “Δ414-428+20aa” in domain III. Huh7.5 cells were transfected with the various recombinants, and cultures with >80% infected cells where lysed for analysis in western blots. To control for impact on replication, NS5A amounts were normalized to amounts of Core. All analyzed mutants had relative NS5A amounts decreased to 30–50% of that of the original H77C(1a) (Figure 6). Thus, assuming that changes in NS5A did not influence stability of Core, we concluded that the P346A/P351A/P354A mutations in LCSII and the Δ414-428 deletion in domain III decreased stability of H77C(1a) NS5A.
As seen in Table 1, two recombinants with modifications in NS5A domain III acquired the F26S (F772S) mutation in p7. We previously reported that this mutation compensated for the exchange of NS5A in the J6(2a) and ED43(4a) recombinants, and in the DH6(1a) NS5A recombinant this residue changed to leucine [12]. To further investigate a putative genetic linkage between NS5A and p7 sequences, we introduced p7 F26S into H77C(1a), TN(1a) and JFH1(2a) NS5A recombinants, which did not rely on adaptation [12]. After transfection and passage to naïve Huh7.5 cells we sequenced p7 and NS5A of viral genomes recovered from supernatants. The H77C(1a) and TN(1a) mutants acquired C447R (C2419R) or V446L (V2418L) in NS5A, respectively, while the JFH1(2a) mutant did not acquire additional mutations (Table 2). This was seen in four separate experiments for each mutant, while four replicate experiments with the original H77C(1a), TN(1a) and JFH1(2a) recombinants did not lead to accumulation of any mutations (Table 2). No variation is observed among genotype 1a and 2a isolates in the HCV database at these two residues. To cross-check whether the observed changes in NS5A of the genotype 1a recombinants compensated for the p7 mutation, we transfected H77C(1a) NS5AC447R and TN(1a) NS5AV446L mutants in triplicates. After passage to naïve cells, all H77C(1a) NS5AC447R cultures and two of three TN(1a) NS5AV446L cultures acquired changes in p7, including F26S (Table 2). Most cultures in addition acquired changes in NS5A. We hypothesized that if the F26S (p7) and C447R or V446L (NS5A) mutations were compensatory, combining these mutations should abolish the need for further mutations. Indeed, after transfection and subsequent passage of H77C(1a) p7F26SNS5AC447R and TN(1a) p7F26SNS5AV446L mutants in triplicates, no additional mutations were observed except for one additional coding change in NS5A for one TN(1a) p7F26SNS5AV446L culture (Table 2). Thus, modifications of NS5A from several isolates induced changes in p7, and p7 mutations induced changes in NS5A for genotype 1a. JFH1(2a) apparently better tolerated the change in p7 and did not rely on compensatory NS5A mutations.
In phylogenetic analysis of NS5A, genotype 2 clusters separately from other genotypes. To investigate the functional significance of genotype-specific residues in the highly variable NS5A protein, we changed positions in JFH1(2a), where genotype 2a residues were different from all or almost all isolates of genotypes 1, 3, 4, 5 and 6 (Figure 7). Most of these positions were in NS5A domain I, and in most cases genotype 2b isolates had residues identical to genotype 2a. Mutations were introduced into JFH1(2a) singly or combined: E95T/Q97P, I110L, H124V, S126D, I140C, S151T/W152E, Q157R, P165C/F168L/F169L and C436V. After transfection, <10% of Huh7.5 cells were HCV positive for the I140C and S151T/W152E mutants, while around 30% were positive for JFH1(2a) and other mutants. Virus production was attenuated for these two mutants and for the E95T/Q97P, Q157R and P165C/F168L/F169L mutants (Figure 8A). After immediate or delayed spread of infection in the transfection culture, supernatants were passaged to naïve cells and the NS5A gene of recovered mutants was sequenced; the I140C and S151T/W152E mutants both acquired V130I. In addition the I140C mutant acquired L188F, and the S151T/W152E mutant acquired T122M. In two independent experiments, the H124V mutant acquired I289T or S300P in domain II. No mutations in NS5A were observed for the other mutants. In functional analyses we investigated the five mutants attenuated after transfection of Huh7.5 cells. As demonstrated by measurement of intracellular HCV Core after transfection of S29 cells, replication was highly attenuated for the I140C and S151T/W152E mutants while Core accumulation was delayed for the E95T/Q97P, Q157 and P165/F168/F169 mutants (Figure 8B). Intra- and extracellular infectivity titers after transfection of S29 cells were reduced up to 10-fold for the E95/Q97, Q157R, S151T/W152E and P165C/F168L/F169L mutants, while reductions of 100-fold or more were observed for the I140C mutant (Figure 8C). Thus the genotype 2 or 2a-specific residues E95T/Q97P, Q157R and P165C/F168L/F169L seemed to be important both for viral replication and intracellular virus assembly.
To address whether the observed attenuation of JFH1(2a) mutants was indeed due to genotype-specific requirements at the analyzed positions, we introduced the reverse mutations into the H77C(1a) NS5A recombinant. After transfection of Huh7.5 cultures, around 10% HCV positive cells were observed for the T95E/P97Q, C140I, T151S/E152W and R157Q mutants, while 30% were positive for the original H77C(1a) NS5A recombinant; no positive cells were observed for the C165P/L168F/L169F mutant. This was reflected by decreased infectivity titers for all mutants (Figure 8D). Thus, mutation of the H77(1a) NS5A recombinant to genotype 2- or 2a-specific residues also led to attenuation.
Since NS5A function depended on several genotype-specific residues in domain I, we wanted to determine whether this domain could function as a genotype-specific entity. We thus replaced NS5A domain I for JFH(2a) by H77C sequence and for H77C(1a) by JFH1 sequence, thereby generating two J6/JFH1 recombinants with either domain I or domain II–III from H77C. After transfection into Huh7.5 cells, the H77C domain II–III recombinant had slightly delayed viral spread and infectivity titers decreased by more than 10-fold compared to JFH1(2a), while the domain I recombinant was highly attenuated (Figure 8E). Data from our previous study showed attenuation even for a J6/JFH1-based recombinant with domain I from the J6 isolate also of genotype 2a [12]. Thus, NS5A function depended on genotype-specific residues in domain I and genotype-specific interactions existed between NS5A domain I and domain II–III.
Due to the clinical and biological importance, there has been great interest in the study of HCV genotype-specific functional differences [9]. However, most functional studies of HCV in infectious culture systems have depended on a single HCV isolate (JFH1). The NS5A protein was so far primarily studied in genotype 1 and 2 replicon systems or in the JFH1 genotype 2a cell culture system. Interestingly, replicon-enhancing mutations were not permissible in vivo [32], emphasizing that conclusions from the replicon systems should be drawn with caution. In this study, we used infectious NS5A genotype 1–7 recombinants [12], and demonstrated a universal dependence of viral replication on the NS5A amphipathic alpha-helix, domain I, LCSI and domain II. Thus, it was demonstrated that all HCV genotypes require these domains for replication. Interestingly, isolate-specific effects of mutations in LCSII and of deletions in domain II and III revealed novel functional differences between NS5A isolates. Furthermore, NS5A function was shown to depend on genotype-specific residues in domain I, a finding that could influence the effect of directly acting antiviral compounds directed against this NS5A domain. Thus, functional isolate-specific differences are emerging for HCV [3], [42], which will be of critical importance for our understanding of HCV biology and for development of antiviral strategies that target NS5A and other regions with isolate variability.
We addressed the importance of the various regions of NS5A for replication and virus production, by analyzing more than 80 mutants in culture. We showed that viral replication in context of the complete life-cycle was critically dependent on the NS5A amphipathic-alpha helix, domain I, LCSI and domain II for isolates of genotypes 1a, 1b, 2a, 3a, 4a, 5a, 6a and 7a. The absence of HCV positive cells by immunostaining early after transfection corresponded to observations for knockout mutants of non-structural proteins NS3, NS4A, NS4B, and NS5B, and a mutant with an introduced stop-codon, all expected to abolish replication. We demonstrated that reversion could occur even for such single-residue mutants, potentially due to the high error-rate of the T7 polymerase used for RNA in vitro transcription. Thus, abrogation of replication due to mutations shown to disrupt the hydrophobic face of the amphipathic alpha-helix, mutation of two zinc-binding cysteines in domain I, or due to exchange of a conserved tryptophan in domain II confirmed and extended previous findings in the Con-1(1b) replicon system [16], [22], [23]. Contrarily, dependence on LCSI in the infectious cell culture system was demonstrated by the highly attenuated phenotype caused by the replicon-enhancing mutation S225P [33], which was also not permissible in vivo [32]. Thus, at least for this mutant the infectious NS5A cell culture systems reflected findings in vivo better than replicon systems. Recently, S225P was shown to enhance replication but inhibit HCV Core release of the full-length Con-1(1b) isolate in vitro [46], which was not in agreement with our data from infectious culture systems. In theory, inhibition of virus production but not replication could be specific for the Con-1(1b) isolate, however, the Con-1(1b) and J4(1b) NS5A protein sequences deviate by less than 5%. Replication seemed to depend also on highly conserved residues downstream in LCSI (Figure S1), as deletion of J4(1b) NS5A residues 235–282 (Δ47) in this study abolished replication, while residues 246–308 previously were deleted in J6/JFH1 with no apparent effect on viability [27]. Interestingly, we observed decreased infectivity titers and mutations in the C-terminus of NS5A for domain II Δ250–293 mutants of genotype 1a but not 2a (Figure 2 and [39]). Thus, HCV apparently shows an isolate-specific dependence on the highly variable N-terminal region of NS5A domain II. The genotype 2 protein sequence deviates in this region from most other genotypes (Figure S1), potentially reflected by differences in structure or function and thereby also in the effect of the deletion. With the critical dependence on selected residues or regions of the NS5A amphipathic alpha-helix, domain I, LCSI and domain II for replication, these regions would be obvious targets for antiviral therapy [47].
Great differences in the effect on viral infectivity titers were observed among the NS5A genotype 1–7 mutants when residues 346, 351 and 354 in LCSII were mutated, with J4(1b) and JFH(2a) being least affected. Measurements of intracellular HCV Core showed that most NS5A LCSII mutants had up to 10-fold reduced replication capacities, while ED43(4a) and QC69(7a) mutants were severely attenuated. Greater than 10-fold reductions in intra- and extracellular infectivity for the H77C(1a), S52(3a), SA13(5a) and HK6a(6a) mutants indicated an additional effect on virus production for these isolates. Findings on the JFH1(2a) LCSII mutant were in agreement with previous findings that a Con-1(1b) but not a JFH1(2a) replicon depended on P346 for replication, while mutation of the infectious full-length JFH1 recombinant led to lower levels of infectivity [42]. In line with this previous study, we found that mutation of P346/P351/P354 inhibited replication at an isolate-specific level. Interestingly, replication of the J4(1b) mutant was only slightly decreased, while the Con-1(1b) P346A replicon mutant was severely attenuated [42]. This might be due to differences between the isolates or between the two in vitro systems studied. Our QC69(7a) LCSII mutant lacked 14 nucleotides in the poly-pyrimidine tract (Materials & Methods), however, it is unlikely that this caused the observed attenuation of replication since previous studies demonstrated efficient replication of J6/JFH1 with much shorter poly-pyrimidine tracts [48]. In agreement with previous findings for residue 343 mutants [42], we could not confirm dependence on this position [23] for replication of any of the NS5A recombinants.
Deletion of the highly conserved residues 414-428 led to isolate-specific decrease in infectivity titers and replication (Figure 4). The deletion had very limited effect on the JFH1(2a) mutant, while most other NS5A isolate mutants had reduced replication capacities; most pronounced decreases were observed for H77C(1a), TN(1a) and S52(3a) mutants. Reductions in replication capacities might explain the decrease in intra- and extracellular titers observed for most mutants, however, the less than 2-fold reduction in replication capacity for ED43(4a) and SA13(5a) mutants compared to the greater than 3-fold reduction in produced infectious particles indicated an effect of the 414-428 deletion on assembly of intracellular infectious particles for these recombinants. The finding that a genotype 1a mutant with residues 414-428 replaced by the downstream genotype 2-specific insertion of 20 residues produced infectivity titers comparable to the original NS5A recombinant, indicated that this insertion could compensate for the 414-428 deletion. Others previously studied JFH1 mutants without the 382–428 region [27] or the genotype 2 insertion sequence [49] and found only minor effects on replication, production of infectious particles and co-localization of NS5A and Core on lipid droplets [27]. This correlated with our results with the JFH1(2a) 414-428 deletion mutant. However, deletion of the 408–437 region [29], which covered the 414-428 region and the genotype 2 -specific sequence, or the almost entire domain III (residues 356–439) [27] of genotype 2a, was highly attenuating for viral infectivity but not for replication. Thus, domain III residues outside the 414-428 region appear to be the most important determinants of efficient virus production.
Experiments with H77C(1a) demonstrated that selected LCSII and domain III mutants led to decreased amounts of NS5A normalized to Core protein levels. Assuming that Core stability was not affected by introduced mutations, this suggested a decreased stability of the NS5A protein. Similar observations were recently reported when two or four of the serines S432, S434, S437 and S438 were changed to alanine in a JFH1 recombinant carrying NS5A domain III from H77 [50]. Such effects could possibly be associated with disruption of NS5A co-localization with Core on lipid droplets [27] or disruption of interactions with annexin A2 [51].
The J6/JFH1 S437A mutation was previously shown to significantly decrease infectivity titers 48 hours after transfection [28], however, another study indicated that mutation of at least two of the S432, S434 and S437 residues was required for a significant reduction of particle production and impaired NS5A-Core interaction, and that viral kinetics were attenuated only early after transfection [29]. Similar findings were reported in the JFH1 background with H77 NS5A domain III [50]. This might explain why we observed only a less than 10-fold reduction in infectivity titers on day 3 but not thereafter (Figure 5). Surprisingly, deletion or point mutation of the relatively conserved 363–380 region in domain III (Figure S1) did not significantly affect JFH1(2a) virus production. Similarly, mutation of arginines 356–359 in a putative nuclear localization signal [52] led to only slight reduction of infectivity titers for JFH1(2a). Most interestingly, changing highly conserved residues in NS5A LCSII and domain III of various HCV isolates led to very different effects on replication and virus production. This emphasizes the evolutionary development of functional differences among the HCV genotypes in particular in variable regions of the genome [3]. Investigation of the role of these regions in vivo would be of interest for future studies.
When replacing the entire NS5A gene of J6/JFH1 with that of other isolates [12], or conducting NS5A reverse genetic experiments as done here, putative compensatory mutations were observed outside of NS5A; particularly in p7, NS2 and NS3 (Table 1, [12]). The F26S mutation in p7 provided adaptation to several NS5A recombinants [12], and in this study F26S was acquired by NS5A Δ414-428 deletion mutants. In addition, the p7 mutation also adapted J6/JFH1 recombinants with genotype-specific NS3/4A protease [53]. Interestingly, introduction of F26S into the genetically stable H77C(1a) and TN(1a) NS5A recombinants led to compensatory mutations in NS5A domain III. Moreover, introduction of these NS5A mutations led to mutations in p7, while combining the p7 and NS5A mutations led to stable recombinants. These findings indicated a genetic linkage between NS5A and p7. NS5A interactions might also involve NS3/NS4A, where several Δ414-428 mutants acquired changes, possibly in concert with p7 or other proteins in the Core-NS2 region, since I399T in NS3 (I1425T) acquired by the TN(1a) mutant also adapted a Core-NS2 genotype 1a recombinant [54]. Additionally, a number of potentially compensatory mutations for NS5A deletion mutants in the third transmembrane domain of NS2 (Table 1), for the J4(1b) NS5A recombinant [12], and for Core-NS2 recombinants [7], indicated an interaction between NS5A and NS2. Pull-down, co-localization and reverse genetic experiments demonstrated NS2 interactions with E1, E2, p7, NS3 and NS5A, and studies of compensatory mutations identified the importance of such interactions during production of virus particles [55]–[58]. Furthermore, the p7 F26L mutation was shown to compensate for changes in Core [59], possibly reflecting involvement of these proteins in viral assembly. Unfortunately, useful antibodies targeting p7 are scarce and the protein can not be directly detected in immunostainings. Tagging of p7 was not compatible with efficient virus production (data not shown and [56]), and was therefore not a useful approach for studying protein interactions likely to take place during viral assembly and release. Thus, despite elaborate efforts we failed to establish co-immunoprecipitation or co-localization based evidence for an interaction between p7 and NS5A in the infectious culture system; better reagents are needed for conclusive studies on p7 interactions.
Mutation from genotype 2a to 1a or vice versa of the NS5A genotype 2 or 2a-specific residues 95/97, 140, 151/152, 157 or 165/168/169 led to attenuation in cell culture. This was of particular interest, since the introduced amino acids led to attenuation irrespective of the presence of these amino acid residues at the given position in numerous infectious HCV isolates. This indicated that genotype 2 evolved specific sequence requirements for function of NS5A domain I. Putative genotype-specific compensatory mutations were identified for the JFH1(2a) H124V mutant that changed serine at position 300 (domain II), present in all genotype 2 and only few other isolates, and the JFH1(2a) S151T/W152E mutant that changed threonine at position 122, present only for genotype 2 and 1b isolates. Residues 130 and 188 that changed in the I140C culture are located in close proximity to residue 140 in the NS5A domain I structure [19], suggesting a functional interaction between these residues. Compensatory mutations in domain II and the reduced viability of domain I exchange recombinants suggested important genotype-specific interactions between domain I and other regions of NS5A. Such interactions might in particular be between domain I and II, since a recent study demonstrated that domain III alone could be exchanged between H77 and JFH1 recombinants [50]. To our knowledge this is the first time the function of an HCV protein was found to depend on genotype-specific residues. Since NS5A inhibitors under current development target domain I [14], [47], this may pose challenges for future antiviral therapy.
Although this study significantly increases the number of studied NS5A isolates, only a single isolate was analyzed for most genotypes. Since differences were observed between isolates of the same genotype, e.g. between domain II deletion mutants of H77C(1a) and TN(1a), studies of more isolates will be important to discriminate between isolate- and genotype-specific findings. Since the original NS5A recombinants used in this study all were efficient without requirement for further adaptation, any effects observed for NS5A mutants are likely to be attributed to the particular NS5A isolate. However, it is possible that particular mutations would render NS5A non-functional in the J6/JFH1 genetic background but not in a full-length background of that particular isolate. Thus, it will eventually be of importance to develop full-length cell culture systems for all HCV genotypes.
In conclusion, we demonstrated that all major genotypes depended on the NS5A amphipathic alpha-helix, domain I, LCSI and domain II for viral replication. Interestingly, dependence on LCSII and domain III for HCV RNA replication and virus production varied with the NS5A isolate. Additionally, functional genotype-specific differences of NS5A domain I residues were identified. Our study highlights the emerging evidence of significant functional differences between diverse HCV isolates. Observed differences in NS5A will be important to consider in functional understanding and therapeutic targeting of this protein. Further studies in vitro and in vivo will be important for understanding and targeting this pleiotropic viral protein.
Reverse genetic studies were done with the J6/JFH1 recombinant [31], and the J6/JFH1-based NS5A genotype 1–7 recombinants H77C(1a), TN(1a), J4(1b)R867H,C1185S, J6(2a)F772S, S52(3a)D1975G, ED43(4a)F772S,Y1644H,E2267G, SA13(5a)R1978G,S2416G, HK6a(6a)I2268N and QC69(7a), expressing the entire NS5A protein (Numbering of mutations according to the H77 reference polyprotein) [12]. Culture adaptive mutations in NS5A are indicated in Figure S1. All mutations analyzed in this study were introduced using site-directed mutagenesis. Marker mutations (according to the H77 reference ORF sequence) introduced to exclude contamination in studies of reversion were T3431C (NS3stop), G3830A (NS3pro), C4280T (NS3hel), G5369A (NS4AG21V), C6224T (NS4BW252S), C6419T (NS5AC57G), C6440T (NS5AC59G), T2719C or C8558T (NS5BGND). The complete HCV sequence of final plasmid preparations was confirmed, except for NS5A domain mutants that did not acquire mutations in the ORF after passage in cell culture and did not produce decreased infectivity titers after transfection. Sequencing identified the following exceptions; J6/JFH1P165C/F168L/F169L carried the additional non-coding C5485T mutation. The J4(1b) Δ414-428 and QC69(7a) P346A/P351A/P354A mutants lacked 2 and 14 nucleotides in the poly-pyrimidine tract, respectively.
Culturing of Huh7.5 hepatoma cells [31] was done as described [60]. One day before transfection or infection, 4×105 cells were plated per well in six-well plates. In vitro transcription of RNA was described previously [5]. For transfection, 2.5 µg RNA were incubated with 5 µL Lipofectamine2000 (Invitrogen) in 500 µL Opti-MEM (Invitrogen) for 20 min at room temperature. Cells were incubated with transfection complexes for 16–24 hours in growth medium. The individual transfection efficiencies of 20 independent experiments, as measured by HCV Core ELISA (see below) after 4 hours, varied less than 2-fold from the positive control. Intra- and extracellular infectivity titers after transfection of S29 cells [40] were determined as described [61]. For infection experiments, cells were inoculated with virus-containing supernatant for 16–24 hours. Supernatants collected during experiments were sterile filtered and stored at −80°C.
Infected cultures were monitored by immunostaining using mouse anti-HCV Core protein monoclonal antibody (B2, Anogen) as described [5], [60]. Infectivity titers were determined by adding 100 µL of triplicate sample dilutions (diluted 1∶2 or more) to 6×103 Huh7.5 cells/well plated out the day before on poly-D-lysine-coated 96-well plates (Nunc). Cells were fixed and immunostained for HCV 48 hours after infection using a previously established protocol [60]. Primary antibody was Hepatitis C Virus NS3 antibody (H23, Abcam). The previously used anti-NS5A 9E10 antibody [31] gave no signal for the J6(2a) NS5A recombinant and for Δ414-428 mutants and suboptimal signals for several other NS5A recombinants. The number of focus-forming units (FFU) was determined by manual counting or on an ImmunoSpot Series 5 UV Analyzer (CTL Europe GmbH) with customized software as previously described [61]. HCV RNA quantification was done using an in-house assay as described [60].
For measurement of intracellular HCV Core, 105 S29 cells [40] per well plated the day before in 24-well plates were transfected with HCV RNA transcripts for the indicated time period. After 4, 24, 48 and 72 hours, cells were trypsinized, centrifuged at 1000× g for 5 minutes at 4°C, washed in cold PBS and lysed in cold RIPA-buffer supplemented with protease inhibitor cocktail set III (Calbiochem). Cell lysates were clarified at 20,000× g for 15 minutes at 4°C before measuring HCV Core levels using ORTHO HCV antigen ELISA test kit (Ortho Clinical Diagnostics).
Huh7.5 cells were trypsinized, washed in cold PBS and lyzed in 200 µl RIPA buffer (Thermo Scientific) with protease inhibitor cocktail set III (Calbiochem) on ice for 10 min. Lysates were treated with RQ1 DNase (Promega) to reduce viscosity, and clarified by centrifugation at 20,000×g for 15 min at 4°C. Protein lysates were loaded on 10% Bis-Tris gels (Invitrogen) and subsequently transferred to 0.45 µm Hybond-P PVDF membranes (GE Healthcare Amersham). Following overnight incubation with specific antibodies (anti-core C7-50, Enzo Life Science) or anti-NS5A (H26, Abcam) at 4°C, unsaturated chemiluminescense images were acquired and protein amounts were quantified based on band intensity using ImageJ.
RNA extraction, RT-PCR and direct sequence analysis (Macrogen Inc) [60] as well as primers specific for the NS5A region [12] were previously described. Sequence analysis was performed with Sequencher (Gene Codes Corporation). HCV sequences were retrieved from the European HCV database and the Los Alamos HCV sequence database. Sequence logos were done using WebLogo [62].
|
10.1371/journal.pcbi.1004659 | Structural and Energetic Characterization of the Ankyrin Repeat Protein Family | Ankyrin repeat containing proteins are one of the most abundant solenoid folds. Usually implicated in specific protein-protein interactions, these proteins are readily amenable for design, with promising biotechnological and biomedical applications. Studying repeat protein families presents technical challenges due to the high sequence divergence among the repeating units. We developed and applied a systematic method to consistently identify and annotate the structural repetitions over the members of the complete Ankyrin Repeat Protein Family, with increased sensitivity over previous studies. We statistically characterized the number of repeats, the folding of the repeat-arrays, their structural variations, insertions and deletions. An energetic analysis of the local frustration patterns reveal the basic features underlying fold stability and its relation to the functional binding regions. We found a strong linear correlation between the conservation of the energetic features in the repeat arrays and their sequence variations, and discuss new insights into the organization and function of these ubiquitous proteins.
| Some natural proteins are formed with repetitions of similar amino acid stretches. Ankyrin-repeat proteins constitute one of the most abundant families of this class of proteins that serve as model systems to analyze how variations in sequences exert effects in structures and biological functions. We present an in-depth analysis of the ankyrin repeat protein family, characterizing the variations in the repeating arrays both at the structural and energetic level. We introduce a consistent annotation for the repeat characteristics and describe how the structural differences are related to the sequences by their underlying energetic signatures.
| Ankyrin Repeat Proteins (ANKs) are usually described as composed of linear arrays of tandem copies of a ∼33 residues length motif with a canonical helix-loop-helix-β-hairpin/loop fold. Being one of the most common protein motifs in nature [1], these molecules are known to function as specific protein-protein interactors. The diversity of unrelated molecules with which they interact is reflected at the many cellular processes in which they are involved [1, 2]. The Notch receptor is a key molecule for metazoan development, involved in cell-cell signaling. Notch has been associated in many types of cancer and several Notch inhibitors are being evaluated for cancer treatments [3]. P16 is a tumor suppressor protein that regulates cell cycle, and mutations on P16 protein are related to several malignancies [4]. The IκB family constitutes a group of related molecules that act as inhibitors of the NF-κB transcription factors. The IκB/NF-κB system is involved in many cellular processes such as cell adhesion, immune and proinflammatory responses and apoptosis [5] as well as in Alzheimer’s disease, diabetes, AIDS, and many types of cancer [6].
Having an elongated architecture, interactions between residues are confined within repeats or between adjacent repeats. These molecules constitute useful models to study sequence-structure-function relationships. However, this non-globular fold represents new challenges [7]. Divergence in the primary structure can be high in comparison to the tertiary structure, with sequence identities between repeats lying on the twilight zone of sequence alignments [8], thus hindering the analysis of repeats at the family level.
We developed and applied a systematic method to consistently identify and annotate the structural repetitions over the members of the complete Ankyrin Repeat Protein Family. We consistently defined the repeats along all the members of the family, and performed comparative studies. We statistically characterized the number of repeats, the folding of the repeat-arrays, their structural variations, insertions and deletions. An energetic analysis of the local frustration patterns reveal the basic features underlying fold stability and its relation to the functional binding regions.
All proteins with at least one detectable ANK repeat were retrieved from the Protein Data Bank (PDB, http://www.rcsb.org/ as it was at May 2014). Repeats were detected by using the hmmsearch module from HMMER (http://hmmer.wustl.edu) and the Hidden Markov Models (HMMs) from Pfam that correspond to the ANK Clan (Pfam ID = CL0465). This Clan is composed of 7 individual domains which are supposed to be evolutionary related (Ank, ID: PF00023; Ank2, ID:PF12796; Ank3, ID: PF13606; Ank4, ID: PF13637; Ank5, ID: PF13857; DUF3420, ID: PF11900; DUF3447, ID: PF11929). We merged the results for which at least one hit was found to any of the HMMs in the ANK Clan. To avoid missing ANKs which could have extremely divergent sequence to be detected by any HMM in the ANK Clan, we looked for structures using the TopSearch application [9]. TopSearch performs structural alignments between a given structure or fragment of it (query) and the entire PDB. We used three different queries: the protein 3ANK (PdbID: 1n0q,A), the protein 4ANK (PdbID: 1n0r,A), and the fragment belonging to the two internal repeats of 4ANK. No further structures were found with this strategy. At last, we obtained 164 PDB structures that contain 54 unique Uniprot entries (http://www.uniprot.org/) and 44 structures of designed ANKs proteins that do not have Uniprot entries associated (S1 Table). Among the 54 proteins, only 25 are related to a unique structure while the others are associated to 2–8 different PDB structures. We created a non redundant set of structures by selecting the one that maximized the coverage of its corresponding uniprot entry. 19 designed ANKs were selected according to different structural and sequence parameters, such as the number of repeats, construction method, number of variations with respect to the consensus.
Mapping between PDB structure chains and Uniprot entries was assigned from the PDBSWS resource [10] and their sequences were retrieved from the UniprotKB database.
The energy landscapes of natural proteins are funnelled towards the native ensemble, in accordance with the Principle of minimal frustration [14]. However, energetic conflicts can remain in the native state, and these may have functional consequences for the dynamics and activity of the protein [40]. The frustration index, Fi, [15] allows to localize and quantify the energetics frustration present in protein structures. Given a pair of contacting residues in the native state, their interaction energy is compared to the energies that would be found by placing different residues in the same native location (mutational frustration index) or by creating a different environment for the interacting pair (configurational frustration index). When comparing the native energy to the energy distribution resulting from these decoys, the native contacts are classified as highly, neutrally or minimally frustrated according to how distant the native energy is with respect to the mean value of the decoys, taking into account the standard deviation from the distribution. An analogous approach used to calculate Fi for contacts can be used for residues. In this case, the set of decoys is constructed by shuffling the identity of only one residue, keeping all other parameters and neighboring residues in the native location, and evaluating the total energy change upon mutation, i.e integrating the interactions that the residue establishes with all its neighbors (single-residue-level frustration index). The algorithm is available at www.frustratometer.tk [16].
Information content obtained from multiple sequence alignments (MSA) measures the conservation of amino acids at specific positions. Also, the information content can be calculated according to the local frustration states (from the frustratometer algorithm on the structural level [16]). The Schneider’s approach [17] was used to compute the information content for both sequences and local frustration states.
The information content per position or per contact (R) can be thought as the reduction of uncertainty regards to the maximum possible (Hmax). The uncertainty observed in a system as defined by Shannon [18] is H o b s e r v e d = - ∑ i = 1 M P i l o g 2 P i, where Pi is the probability that the system is in state i. In this work the states are the amino acids at characteristic positions (for sequences) and the frustration class of contacts (for local frustration). The probabilities are normalized, i.e ∑ i = 1 M P i = 1, where M is the size of the alphabet (20 for amino acids, 3 for the frustration index). Therefore, the information content can be calculated as R = Hmax − Hobserved.
Generally, it is considered that Hmax is reached for a uniform distribution of states: then P i m a x = 1 / M and Hmax = log2(M). Nevertheless, if states are not equally likely to occur the probability distribution of states should be used to estimate Hmax. We set Hmax = log2(20) for sequence IC calculations, and used the distribution of states reported by Ferreiro et al [15] to calculate the Hmax for frustration IC calculations.
The MSAs derived from structural alignments produced by the tiling procedure were used to generate HMMs (using the hmmbuild module from the HMMER suit) specific for each type of repeat, i.e N-terminal, internal or C-terminal. These HMMs are available in https://github.com/proteinphysiologylab/StructuralAnks_Parra2015. The hmmsearch module from HMMER can be used to look for hits to any of these models over specific sequences of interest.
We collected and curated all ANKs structures from the PDB Data Bank, with a grand total number of 169 entries. Since many entries correspond to the same protein, we defined a non redundant dataset of 74 proteins, where 55 correspond to natural ANKs and 19 correspond to human-designed ANKs (see Methods). The structures are composed of 3 to 12 ANK repeats (S1 Table) with an overall α-solenoid fold architecture according to the classification defined by Kajava [19]. Although the majority of the structures are mainly composed of ANK arrays, 5 entries contain non ANK domains in the same chain (Murine Arf-GAP, PdbID: 1dcq,A; Human Arf-GAP, PdbID: 3jue,B; AnkX protein from Legionella pneumophila, PdbID: 4bet,A; Human Osteoclast-stimulating factor 1, PdbID: 3ehq,A and the Serine/threonine-protein phosphatase PP1-beta catalytic subunit from Gallus gallus, PdbID: 1s70,B).
ANKs appear to have very similar structures when observed globally, but subtle differences can be detected when looking carefully. We used the TopMatch tool [11] in order to analyze how (dis)similar the structures are, and pairwise aligned all the proteins from the non redundant dataset. TopMatch implements a metric for structural similarity [12], S, roughly equivalent to the number of residues that can be structurally aligned. A sequence identity value, I, is also calculated as the number of structurally aligned residues that are identical. Normalized values are used to correct S and I due to the differences in length of the aligned structures, with la and lb being the lengths of the two molecules being aligned, relS = 2 ⋅ Sab/(la + lb), as defined in [12], and relI = 2 ⋅ Sab ⋅ (I/100)/(la + lb). As a general trend, we observe that relI values are much lower (mean of 0.22 and sd = 0.16) than the corresponding relS values (mean of 0.7 and sd = 0.17), for two aligned proteins (Fig 1). Alignments between designed proteins are the ones that have the highest relI and relS values with mean values of 0.6 (sd = 0.17) and 0.85 (sd = 0.1) respectively. Alignments between natural proteins have a wider range of structural variability, with relS distribution centered at 0.55 (sd = 0.16) while relI is lower, with a distribution centered at about 0.17 (sd = 0.14). We found that there is a roughly linear correlation between sequence identity (relI) and structural variability (relS) distributions when aligning natural ANKs. The pairwise alignments show that despite a considerable sequence variability, structures with a relI as low as 0.2 can have a relS value of 0.8. In order to analyze which features of the family are captured by relI and relS a hierarchical analysis based on each of these measures was applied (S1 Fig).
A clustering analysis based on relI distinguishes the proteins mainly by their orthology and paralogy. Designed proteins segregate in a clearly separated cluster evidencing that natural proteins are mainly composed of non-consensus repeats. When the clustering analysis is performed using the relS values the clusters are mainly composed of proteins with similar number of repeats. We observe an outer group nucleating non-eukaryotic proteins, displaying long non-repetitive regions or belonging to a specific class of ion channels (TRPV family) which have strong structural modifications on their repeats. In order to find the repeat motif that better describes the family, we analyzed the dendrogram obtained from the clustering based on the relS parameter. Starting from the base, the leaves were collapsed in pairs from the bottom to the root. Each time a branch consisting of two members was collapsed, the common structure between the two is annotated. Once the whole dendogram is collapsed, the substructure that is common to all the dataset is obtained. The resulting regions match with the two internal repeats of the 4ANK protein [20].
Repeat proteins are the result of a collection of analogous motifs, organized in tandem arrays and related by different types of transformations between neighboring units (i.e. rotational, translational or screw [21, 22]). We have shown that a tileability score [13] captures how collectively periodic a structure is, taking into account how repetitive it is at different length-scales. Tileability scores for the ANKs in the non redundant dataset are shown in Fig 2A. Despite the fact that all these proteins share a common architecture with highly similar motifs disposed in tandem, small perturbations within or between repeats propagate to the global structure decreasing the overall symmetry. Proteins where the geometrical transformations between adjacent repeats are homogeneous along the array and do not present big insertions or deletions have high Tileabilities (e.g. Notch1, PdbID: 2he0,A; Ankyrin-1, PdbID: 1n11,A; ANKRA2, PdbID: 3so8,A). On the other side of the Tileability spectra we find proteins that hold a considerable proportion of non-repetitive regions (e.g. Serine/threonine-protein phosphatase PP1-beta catalytic subunit, PdbID: 1s70,B; AnkX, PdbID: 4bet,A) or display several structural perturbations (Cell cycle regulatory protein SWI6, PdbID: 1sw6,A).
Repeat proteins usually need modified versions of their terminal repeats in order to be soluble enough and to maintain their overall stability [23]. We separated the structurally detected repeats into three groups, i.e N-terminal, Internal and C-terminal repeats, and analyzed their structural and sequence properties. The first helix is usually extended in the case of the N-terminal repeats while in the case of C-terminal repeats, the second helix is extended (S2 Fig). In order to analyze how repetitive is a structure at different regions, we defined a per-residue function that measures how frequently a residue is covered by copies of fragments defined at all possible lengths and phases. Thus, we were able to locate which regions break the overall symmetry of the molecule. While some ANK structures, as the one corresponding to Notch1, are highly periodic for all tile sizes and at many phases (Fig 2B), some others are barely periodic with many regions that are non-repetitive as for SWI6 (Fig 2C). In this last example, we detected a region at the N-termini that has a combination of secondary structure elements that can be aligned with ANK repeats. This type of secondary structure arrangements, present also in other cases, may serve as non repetitive caps.
Ankyrin domains are composed of a variable number of repeats that are arranged in tandem. In order to consistently define the structural repetitions we applied a structural search method that analyses the periodicities and repetitions in the protein structures [13]. Given their repetitive nature, ANK structures can be decomposed into minimal units whose copies can be translated and rotated to reconstruct the overall protein structure. We observe that most ANKs are markedly periodic, identifying fragments of 33 residues length as the peak of highest amplitude, and multiples of it (Fig 3A). This is in agreement with the repeat length reported by other studies [1, 24]. While a periodic pattern is evident, definition of limits between adjacent repeats (or periods) is not straight-forward.
To consistently annotate the repeats we considered the phase definition that better explains the structural observations in all the known ANKs. For each protein we selected the fragment of 33 residues length that has the highest coverage score and defined its phase by comparing it to the prototypical ANK Hidden Markov Model (HMM) deposited in the Pfam Database. We observed that in 70% of the structures, a multiplicity of fragments defined at different phases share the highest score. For the remaining 30% of the dataset a unique phase has the highest score, and there is a distribution of phases that are selected (Fig 3B). Thus there is not a clearly conserved geometric phase in natural ANK proteins.
Protein and repeat array stabilities may be related to the selection of a preferred phase of the repeat-unit. Each repeating unit has similar structural elements that interact with the nearest neighbors, and their folding process could be depicted by many folding funnels that merge upon interaction [25]. Following this idea, we took all the possible fragments of 33 residues length and calculated their relative foldability score, Θ = Δ E / ( δ E N ) [26]. We calculated the energy that corresponds to the sum of the internal interactions as measured by the AM/W energy function [27], and compared it to the mean energy of a set of N structures (ΔE) and its variance (δE), for all the other fragments of 33 residues length in the ANK array. We took the fragment of 33 residues length with the highest Θ value in each protein, and assigned it a phase according to the ANK HMM. We observe that there is a highly conserved phase, starting at position 28 of the ANK HMM, for most of natural ANKs. Cases that deviate from this prevalent phase correspond mainly to designed proteins and members of the TRPV channels that constitute a specific subgroup of proteins where several structural modifications are present at individual repeats. Interestingly, proteins designed by Peng’s group in which covariations among the positions were respected [20], show a phase that agrees with most natural proteins (Fig 3C).
Once both the period and the phase were unequivocally defined, we proceeded to detect and annotate the repeats on the ANKs comprised in the non redundant dataset (S1 Table). We selected the first internal repeat from the 4ANK protein, the one that maximizes the structural similarity, defined at the phase identified previously. We used this fragment to cover all other target proteins with repetitions of it. As a result, a set of structure-based sequence alignments between the 4ANK fragment and the matched subfragments within each target protein were obtained. For each alignment we redistributed the gaps satisfying that i) any gap modification for the query sequence was also applied to its target sequence and ii) globally align all the query sequences, which only differ in the number of gaps and location of them. Thus, all the target sequences became transitively aligned, producing a MSA using only structural information. Additionally, the pairwise analysis retrieved from the procedure allows to assign canonical positions to the repeating units as well as insertions and deletions, defined by those regions with gaps in either the query or target sequences, respectively. The length of the detected repeats range from 24 up to 48 residues (Fig 4A). Repeats that are located at the termini display largest deviations towards shorter lengths (Fig 4B). The most affected ones are the N-terminal, that lack the first β-hairpin region. In the case of the C-terminal repeats, deletions correspond in many cases to a shorter second helix. Insertions and deletions are not homogeneously distributed along the repeats but localized at specific regions. Insertions are mainly located at the β-hairpins and between the two α-helices (Fig 4C) and deletions are particularly located at the end of the first helix (Fig 4D). Both insertions and deletions show variable lengths, and they are not correlated with their locations. (Fig 4C and 4D). A total of 390 repeats were obtained with our structural method which is an increment of ∼18% respect with the 321 repeats that are detected using the ANK HMM in Pfam. If we consider only natural proteins, 311 repeats are detected in structure and 253 over sequences with the ANK HMM in Pfam. HMMs specific for the N-terminal, internal and C-terminal repeats were created from the specific MSAs for each type, using the hmmbuild module from the HMMER suit. 310 repeats are detected over the natural ANK sequences analyzed before, by just using the HMM derived from the internal repeats MSA.
ANKs are described as being specific protein-protein interactors [1]. Protein evolution is the result of different forces that select for their ability to fold into a stable structure and perform a biological function. In this sense, energetic conflicts present in the ANK repeat fold could be related with the ability of these proteins to efficiently bind their protein targets. We used the Frustratometer tool [16] in order to locate in the structures these energetic conflicts and quantify them by analyzing the different frustration indices [15].
There are 83 structures from the dataset where the ANKs are in complex, either forming homoligomers or heteroligomers. An entry belongs to the non-redundant set if there is no other complex containing the same ANK Uniprot ID or if it is in complex with a different partner than the entries already included. A total of 33 complexes were analyzed (S3 Table). ANK repeats involve an average of 20% of their residues in binding (sd = 11%). 80% of the binding residues correspond to canonical positions in the repeats (sd = 0.24%), 15% correspond to insertions either between or within repeats (sd = 0.19%) and only 5% can be mapped to non repetitive regions.
In general, the whole repeat array is at some point involved in protein-protein interactions (S2B Fig). We observe that interactions are enriched in the β-hairpins, as previously described [1]. However the inter helices region is also enriched in interacting residues. Thus there is no conserved mechanism for ANKs binding other proteins (Fig 8A–8D). In addition to a non localized epitope at the canonical structure of repeats, interactions usually involve non repetitive regions, as is the case of the complex between the ANK repeat-containing protein, myosin phosphatase targeting subunit 1 (MYPT1, PdbID: 1s70) and the ser/thr phosphatase-1 (delta) (Fig 8A) where a helix is connected to the repetitive array by a loop which directly interacts with the partner. Another example of this is the IκBβ/NF-κB p65 homodimer complex (PdbID: 1k3z) (Fig 8B) where a loop with no defined secondary structure interacts with NF-κB. A related protein, IκBα, also displays a loop region at the C-terminal that binds the NF-κB P50/P65 heterodimer. This region is disordered in the free protein and has profound implications for the binding energy between the molecules [41]. While in the case of IκBβ all the repeats are involved in the interaction with the other molecule, in other ANKs like MYPT1 or in the case of the ANK repeat-containing protein YAR1 when binding the 40S ribosomal protein S3 (PdbID: 4bsz) (Fig 8C), the interface is concentrated at specific locations along the array. ANKs are also able to form homodimeric complexes involving different types of interfaces with one of the most interesting cases being the Tankyrase-1 (PdbID: 3utm) (Fig 8D), where the monomers are intertwined in a crossed fashion involving their central repeats, that have extended helices due to insertions. It was previously described that binding sites are enriched in high local frustration [15]. We compared the configurational frustration index calculated in both the unbound ANK monomers and in the bound state (Fig 8E). The frustration distribution for both states are similar, indicating that both correspond to largely minimally frustrated networks of interactions. We compared the frustration change at the contacts established by those residues that are involved at the interface between the ANKs and their partners, and observed that the majority of them, ∼60%, change towards lower frustration values. This can be attributed to the change in the solvent accessibility upon binding that is captured by the burial term in the AMW energy function. The remaining ∼40% of highly varying contacts, change in the opposite direction. We observe that while much of the frustration present at the unbound state is released in the complexed state, some new highly frustrated interactions arise as a product of the interface formation. It is tempting to speculate that the frustration that appears upon oligomerization has functional consequences for further conformational transitions once the complexes are assembled.
In order to characterize the ANK repeat protein family, detection and consistent annotation of repeats within the repetitive arrays is required. The high sequence diversity among repeats and their short lengths pose a problem when trying to accomplish the annotation using sequence-based methods. We applied a method that analyses periodicities and repetitions at protein structures by tiling the structural space [13]. By combining this approach with the relative foldability function [26] we were able to consistently define both the length and the phase of the ANK repeats. The structural detection of ANK repeats allowed us to detect instances that are completely missed when using other structural methods as AnkPred [42] that fails to detect, for example, the sixth repeat on IκB-α (PdbID: 1ikn,D) or the repeats within the K1 protein from the Vaccinia virus (PdbID: 3kea,B). While the Console method [43] is able to detect the six repeats on IκB and almost all repeats in the K1 protein, it is not possible to select the phase at which the repeats appear, constituting an obstacle for comparative studies. With our procedure, all repeats from the protein structures in a non-redundant dataset (S1 Table) were consistently annotated, together with the insertions and deletions. The new sequence profiles derived from these annotations, based on a strictly structure-based multiple sequence alignment, give a new perspective about the sequence divergence that ANK repeats can tolerate. Moreover, these profiles could be used in combination with the existent ones in order to improve the annotation and sequence coverage of ANKs at sequence databases such as Pfam, Uniprot and definitively would be useful in order to perform consistent high-quality annotation on structural databases such as RepeatsDB [44] in which all repeat-containing proteins are being deposited and characterized.
We have shown that the N-terminal, internal and C-terminal repeats are different in their sequence and energetic signatures, having distinct secondary structure compositions. Internal repeats display the highest conservation on both sequence and energetic signatures while the N-terminal repeats were the least conserved ones. If sequence conservation is compared to the conservation of local frustration at the structure level, we observe that there exists a positive linear and significant correlation between them for the C-terminal and internal types while no significant correlation was observed for the N-terminal case. This correlation suggests that the more similar to the consensus sequence a repeat is the more foldable it will be. Consensus mutations have proven to be effective at stabilizing proteins [45] with some exceptions [46]. Destabilizing consensus mutations were observed at highly co-varying positions or invariant ones where hidden correlations can occur [4, 47, 48]. We calculated which interactions within the canonical structure of ANK repeat pairs are the most conserved leading to a set of highly conserved residues, at the sequence level, that are connected by a network of minimally frustrated interactions. The majority of these interactions are established between hydrophobic residues. Non hydrophobic interactions involve the GXTPLHLA motif interacting with the β-hairpin at the single repeat level or interactions with the loops at adjacent repeats, as well as interactions with their respective GXTPLHLA motifs (S2 Table).
Stabilizing interactions within and in between ANK repeats are encoded in the consensus sequence of a single, self-compatible repeat as evidenced by the success of consensus design by stacking identical repeats [20, 29]. Folding cooperativity of repeat proteins is highly influenced by the intrinsic stabilities of the different repeats and their interfaces. It has been computationally shown that these proteins are ‘poised’ at particular ratios of inter-repeat and intra-repeat interaction energies that allow them to undergo partially unfolding under physiological conditions which would be a requirement to perform their biological functions [49, 50]. We have mapped the interactions that are the most energetically favored between residues composing the canonical structure of ANK repeats. Identifying which of these interactions, within and in between natural repeats, are not satisfied at natural repeat arrays could help trace the determinants of differential folding behaviors between different ANKs which, despite having the same number of repeats, display substantial different dynamical properties.
ANK repeats can undergo several modifications, i.e insertions and deletions at their canonical 33 residues length framework. Analysis of the energetics of the interactions that surround these structural modifications, showed us that there exists an enrichment of highly frustrated interactions around them. This suggests that insertions and deletions occurring at ANK repeats may have functional consequences for the entire molecule. Moreover, binding residues, identified from co-crystallized complexes, showed that these are also enriched in highly frustrated interactions, that are mostly relieved upon binding.
ANKs are adapted to perform their principal function of specifically binding other proteins. Their sequences and structures can vary in order to maximize their recognition properties, introducing considerable structural deviations and even displaying order/disorder transitions in many cases. Their modularity allows for an exquisite tuning of individual repeats giving differential dynamic properties to different regions within the repetitive array. The presence of frustration at binding sites, insertions and deletions shows that, in many cases, evolution seems to have kept energetic conflicts that destabilize the individual ANK structures but at the same time should be critical to drive the recognition process and further release of frustration when complexed. ANKs seem to strategically combine the introduction of structural perturbations by mutation of key residues at the canonical sequence of the different repeats while maintaining some of them unchanged. This tinkering of the sequence not only modifies the global structure and modulate the affinity and specificity for partners recognition, but it may also encode complex dynamic behaviors, as the presence of a multiplicity of folding intermediates or an increased conformational plasticity that arises due to these variations [51].
|
10.1371/journal.pcbi.1000769 | Energy-Information Trade-Offs between Movement and Sensing | While there is accumulating evidence for the importance of the metabolic cost of information in sensory systems, how these costs are traded-off with movement when sensing is closely linked to movement is poorly understood. For example, if an animal needs to search a given amount of space beyond the range of its vision system, is it better to evolve a higher acuity visual system, or evolve a body movement system that can more rapidly move the body over that space? How is this trade-off dependent upon the three-dimensional shape of the field of sensory sensitivity (hereafter, sensorium)? How is it dependent upon sensorium mobility, either through rotation of the sensorium via muscles at the base of the sense organ (e.g., eye or pinna muscles) or neck rotation, or by whole body movement through space? Here we show that in an aquatic model system, the electric fish, a choice to swim in a more inefficient manner during prey search results in a higher prey encounter rate due to better sensory performance. The increase in prey encounter rate more than counterbalances the additional energy expended in swimming inefficiently. The reduction of swimming efficiency for improved sensing arises because positioning the sensory receptor surface to scan more space per unit time results in an increase in the area of the body pushing through the fluid, increasing wasteful body drag forces. We show that the improvement in sensory performance that occurs with the costly repositioning of the body depends upon having an elongated sensorium shape. Finally, we show that if the fish was able to reorient their sensorium independent of body movement, as fish with movable eyes can, there would be significant energy savings. This provides insight into the ubiquity of sensory organ mobility in animal design. This study exposes important links between the morphology of the sensorium, sensorium mobility, and behavioral strategy for maximally extracting energy from the environment. An “infomechanical” approach to complex behavior helps to elucidate how animals distribute functions across sensory systems and movement systems with their diverse energy loads.
| Animals thrive by sensing their environment and using the information they've gathered to guide their movement. But collecting better information can result in less efficient movement: Bicycling while standing up on the pedals may help you see over obstacles ahead of you, but it causes more air drag, forcing your legs to work harder. Nocturnal weakly electric fish search for prey with their body tilted. This tilting more than doubles the resistance to movement from the water, but because the fish's ability to sense prey improves when tilted, it is better to swim this way. Beyond a certain amount of tilt, the costs of movement become too great. Interestingly, the benefit of tilting is dependent on the shape of the volume around the fish where it detects prey. We also found that if the fish was able to swivel its region of prey sensitivity, like a vision-based animal can shift its gaze, it would save energy. This conclusion helps us understand why animals like us can move our eyes. A Polish folk saying succinctly captures the gist: “He who doesn't have it in the head has it in the legs” (Ten kto nie ma w głowie ma w nogach).
| Animals must constantly negotiate trade-offs in sensory and motor performance. The most well known of these trade-offs occur within either movement or sensory systems, rather than between them. As an example within motor systems, fish body shapes and styles of movement that maximize cruising efficiency may suffer from poor maneuverability [1]–[3]. In sensory systems, converging signals from large numbers of photoreceptors for increased sensitivity results in reduced spatial resolution.
What about trade-offs between movement and sensory systems? For example, for a fixed amount of available energy from food sources, is it better to expend that energy on a larger visual sensing range (via a larger eye and the brain tissue to process signals), or to move the body more so that the effective area that is scanned is similar? One challenge in assessing such trade-offs is that it is difficult to compare measures of movement performance, such as energy efficiency, to sensory performance, such as acuity. Ultimately, however, these different subsystem performance measures translate into net energy gains and losses for the animal [4]. Consequently, examining energy provides a lens through which to look at how an animal can best trade off movement and sensing. Given that neuronal tissue requires about 20 times more energy than skeletal muscle per unit mass in mammals, where it has been measured ([5], after [6]), we already know that brains and sensory systems are metabolically expensive compared to movement systems. Recent studies have shown the important influence of the energetic costs of sensory systems, such as the role of these costs in the evolution of sensory systems (review: [7]). Although looking at energetics enables comparison of the costs of movement and sensing in behaviors where these are closely interrelated, such an analysis has rarely been performed [7].
One simple source of trade-offs between movement and sensing can be easily understood. A key role of a sensory system is to support scanning the environment for food, threats, mates, competitors, or anything else which may affect the animal's continued existence. But the space where these items of interest exist will typically exceed the range of the sensory system. To scan a larger volume of space, an animal can move its body, or evolve increased sensory range. Either approach has its associated costs. In the case of body movement, it is the cost of locomotion. The amount of locomotion needed will depend on the range of the sensory system being used, with less movement needed by long-range systems, such as vision, and more movement needed for short-range systems, such as touch. If, for example, you need to detect the location of a split on a wood table, you can use your visual system and glance at the entire surface at once (little dependence on movement), or you can move your hands across the surface and use your sense of touch to detect the split (maximal dependence on movement). In the case of evolving increased sensory range, the associated costs include more neuronal tissue, development costs, maintenance, and the cost to carry the weight of the sensory system (not insignificant for flying animals: the fly uses 3% of its energy simply to keep its visual system aloft [8]).
The above type of trade-off between movement and sensing is indirect because the problem is how best to expend a fixed amount of energy (more on movement, and less on sensing, or vice versa)— but not a case where improvement in one domain comes at the expense of performance in the other domain. An example of a more direct trade-off like this is how moving the eye faster to increase the speed of visually inspecting an area of space can directly conflict with visual performance. The conflict arises when an image passes over more than one photoreceptor acceptance angle per response time, since this results in the visual percept being degraded by motion blur [9].
A thought experiment can help expose another direct way in which a trade-off between movement and sensing can occur, one similar to the kind at issue in this study. As the effective range of a given sensory epithelium approaches zero (contact sensing), to increase the amount of space that is scanned while moving through space (for example, in a straight line) can require reorienting the sensory epithelium in a way that results in less efficient movement. For example, imagine your finger was an autonomous organism. Suppose this finger is feeling its way along a novel surface in a water current (or a stiff wind), with the long axis of the finger parallel to the direction of movement so as to minimize drag effects. Now, the back portion of the finger is scanning the same surface as was already scanned by the front. To increase the amount of space being scanned per unit time, the sensory epithelium needs to be reoriented. Ideally, the finger would be oriented perpendicular to the line of travel. This way the rate of surface scanning is maximized; but now there is also maximal projected area in the direction of travel, and thus maximal drag.
Contrast this situation with that involving a sensory epithelium whose range is far from zero, such as the retina of an eagle flying high and looking for prey on the ground. Now, suppose that the eagle is looking straight downward. The eagle's visual sensorium can be idealized as a cone whose apex is the eagle's head. The area scanned per unit time will be the width of the cone times the velocity of the eagle. If instead of looking straight down, the eagle sweeps its conical sensorium from side to side by moving its eyes, it will greatly increase the area scanned per unit time. In this case, however, to reorient the sensory epithelium through eye rotation comes at no change in the projected area of the body in the direction of travel, and thus no added costs due to increased drag. If the eyes were not movable, the eagle would have to turn its head, which could result in more drag; if the eye and head were not movable, the whole body would need to be reoriented, incurring even more costs. Note, however, that having the ability to reorient the sensory epithelium without changing body orientation can incur significant neuronal processing costs, since it may require coordinate transformations from a sensory organ-fixed coordinate frame (e.g., retinotopic coordinates) to body-fixed coordinates.
With sensors distributed over a sensory epithelium consisting of the entire body surface, as occurs in somatosensory and electrosensory systems, it becomes progressively less possible to reorient the sensory epithelium independently of full body reorientation. For example, it conflicts with the strategy of concentrating the sensors on one portion of the body which is moved with muscles, as with some eyes and pinnae. Full body reorientation, however, can be quite costly if the relative velocity between the body and the surrounding environment is sufficient to produce drag forces on the body — for example, if the animal is moving rapidly through the air.
An example of this type of trade-off between sensing and movement can be found in chemosensory behavior of the blue crab [10]. Blue crabs move sideways up-current, with their body slightly rotated into the flow. The slight rotation into the flow is believed to result in improved sensing of the local gradient of odorant molecules, as this rotation causes their primary chemosensory appendages for this behavior—their legs [11] —to be sensing slightly across the flow. Without this slight rotation, the downstream legs receive fluid in which the odorant has been mixed and diluted from hitting the upstream legs, compromising the ability to detect and localize the odorant. With the body rotated into the flow, the crab avoids this dilution and can use bilateral comparisons between chemosensory input along the legs to help guide the body to the source. However, turning the body into the flow also increases drag. As Weissburg and coworkers increased flow speed in their experimental apparatus, they found a speed at which the crab chose not to rotate the body into the flow. The cost of movement at the increased drag appears to outweigh the gain in sensory performance at this critical flow speed.
Here we present an analysis of a conflict between efficient movement and sensory performance using the model system of weakly electric fish (Figure 1A), a leading system for the analysis of sensory function in vertebrates. These fish hunt for small insect prey at night in rivers of the Amazon Basin, through the use of an active electrosensory system. The fish generate an oscillating electric field ( near the body), that surrounds the whole animal. When prey enter the fish's electric field, a small change in voltage occurs across the skin () [12], [13]. This change in voltage is detected by electroreceptors covering the entire body surface. These voltage modulations are then transformed into changes in the firing rate of primary electrosensory afferents that terminate in the hind brain of the animal for further processing (reviews: [14], [15]).
While searching for prey, these fish were previously shown to hold their body with the head down at a pitch while searching for prey [16], as illustrated in Figure 1B. We show that this posture significantly increases the cost of movement. However, this increased cost is more than offset by the increase in sensory performance resulting from the posture. We observed that this increase in sensory performance is dependent upon the fish having an elongated sensorium. When we examined the effect of the fish having a non-elongated sensorium, such as a blunt-shaped sensorium or a forwardly-directed visual sensorium similar in aspect ratio to a visually-guided aquatic predator, we found that there was no benefit to increasing the pitch of the body. We show that if the black ghost could swivel its sensorium independently of body movement, as visually-guided animals can swivel their sensoria, the fish would obtain a significant benefit through reduced energy expenditure for prey search.
Body movement through any medium results in lost energy due to friction between the medium and the body. In air these effects are slight except for flying animals. In water, with 1,000 times the density of air, these effects are significant even at relatively low speeds. As mentioned above, black ghost knifefish tilt their body while searching for prey. To estimate the energetic consequences of tilting their body from neutral (horizontal) body pitch to the measured , the force needed to overcome the resistance to movement (drag) through water needs to be estimated at different body pitches and movement speeds. The energy needed to overcome this resistance is then simply this force times the distance moved.
We estimated the drag in two ways. First, we performed high resolution computational fluid dynamic simulations of the black ghost as it was being virtually towed through water. The forces on the body are easily recovered from the simulations, as are the flow patterns, which give insight into the basis of the drag forces corresponding to each body pitch angle. The computed flow patterns are shown in Figure 2. Second, we towed an accurate urethane cast of the knifefish through a large water tank at constant, behaviorally relevant velocities, measuring the steady-state resistance to movement with a force sensor that the cast was attached to.
We highlight results for 15 cm/s, because our prior prey capture study with the black ghost knifefish found search velocities of 9.34.3 cm/s (mean and std) [16]. In that study, the tank in which we made our observations had to be small due to imaging constraints, making 15 cm/s a reasonable choice to focus on here. The drag force results are shown in Figure 3. At 15 cm/s, the measured drag force was 2.00.4 mN (), 5.20.4 mN (), and 8.10.5 mN (). The corresponding computed drag forces were 1.0, 6.1, and 12.2 mN.
The measured drag was typically lower than the computationally estimated drag. As shown in the snapshots of the computed flow patterns around the fish being virtually towed at 15 cm/s in Figure 2, at the flow separation is higher than in the other cases. Because of this degree of separation, computational fluid dynamic simulations that incorporate the effect of turbulence may be required to fully resolve the flows around the body. If turbulence is present in the empirical experiments with the fish cast, this could potentially reduce the drag. Given the disparities between measured and computed drag forces, we use the measured drag forces for the remainder of the study. Our key result, that observed pitch angles during search behavior are consistent with minimizing costs, are not affected by this choice.
The fish has an omnidirectional field of prey sensitivity [12] (Figure 4A) because of the broad distribution of sensors and electric field described above. This volume is relatively uniform, although there are significant non-uniformities in electric field strength and sensory receptor density [12]. As shown in Figure 4A, as the fish increases its body pitch, the amount of space that it scans while moving increases. The volume the fish can sense prey within while moving is the product of the frontal area of the sensorium (the area that results from projecting the volume to a plane at right angles to the direction of motion), and the distance traveled. For a cuboidal idealization of the complex natural shape of the sensory volume (see Materials and Methods), we found that the projected frontal area increased with body pitch up to a maximum at a body pitch of (Figure 5A). At neutral body pitch, the frontal area was , going up by 190% to at and up by 235% at .
Our energetics model estimates the amount of energy needed to overcome drag forces for the fish to swim to a single prey of the kind used in quantifying the size and shape of the sensorium, Daphnia magna. These prey are typically found in stomach content analyses of Apteronotus albifrons [17]–[19] and have known energy content (Table 1). We assume that prey are uniformly distributed at the density shown in Table 1.
As derived below in Materials and Methods, the equation for estimating the energy in joules needed to overcome drag to reach a single prey is(1)
is the power needed to overcome drag at the reference velocity (during steady state swimming, thrust power must be equal to the power needed to overcome drag). We fixed to the power needed to overcome the experimentally measured tow drag at pitch and 15 cm/s, which was 0.3 mW (15 cm/s times the drag force at this velocity, 2 mN, Figure 3). is the density of prey (see Table 1). is a function of body pitch angle which returns the area of the sensorium projected to a plane perpendicular to the path of motion. is a function of body pitch such that the drag force is equal to , where is the velocity of the fish.
As shown in Figure 5B for the curve labeled “2.2 (natural)” the energy needed to encounter one prey at neutral pitch was slightly over 25 J, going down by around 40% to near 15 J at the optimal pitch of just over , with a similar value at a pitch of .
Changing the pitch of the body not only affects the drag on the body, and the search rate, it also affects propulsion. The black ghost knifefish generates force by undulating the extended ribbon fin along its underside (Figure 1A) while keeping its body semirigid except for bends to turn left or right [20], [21]. The fin undulations are approximately sinusoidal and travel from one end of the fin to the other—from head to tail for forward movement. The fin generates two different forces: one along the length of the fin (called surge), and one smaller force perpendicular to the fin, pushing the body up (called heave) [21]. As the fin tilts, the forward propulsive force reaches a maximum when the fin base is at an angle of approximately to the horizontal. This is its angle when the body axis is horizontal (e.g., when , then the fin base is at angle in Figure 1B, approximately ). As the fin base tilts past ( body pitch), the sum of the surge and heave forces projected to the forward direction decreases. This effect is shown by Figure 6, which depicts a family of curves relating forward propulsive force to body pitch (Figure 6). For the purposes of this illustration, we assume that the fish varies its frequency of undulation to vary propulsive force. This appears to be true [20].
We examined the influence of sensorium shape on the energy needed to encounter prey. We define the “elongation factor” as the ratio of the length to height of the sensorium. The effect of elongation factor on projected sensorium area and energy to encounter one prey is shown in Figure 5. The naturally observed elongation factor is 2.2. When the elongation factor was 1.0 (sensorium length equal to height), the energy needed per prey decreased negligibly at low angles before increasing with body pitch angle; essentially, there was no improvement in performance with pitching the body. When the elongation factor was 4.0 (sensorium length four times its height), the energy needed decreased with body pitch angle up to pitch angles of . With this elongation factor, the energy needed per prey encounter was typically less than half the energy per prey encounter for the 2.2 elongation factor at relevant body pitch angles. Sensorium elongation makes body pitching progressively more advantageous.
The effect of blunt versus elongated sensoria was further explored through a scenario in which the black ghost has a frontally-directed visual sensorium (see Figure 7A) rather than its normal omnidirectional sensorium (Figure 4A). A fish called the stone moroko (Pseudorasbora parva) is a visual predator whose vision-based sensorium for Daphnia has been measured ([22]) and is shown in Figure 4B. A cuboidal approximation of the stone moroko visual sensorium is 11.9 cm high (vertical)×12.0 cm long (distance of leading edge from the eyes)×18.7 cm wide (left-right extent). The elongation factor, length over height, is therefore close to 1.0. Given this aspect ratio, there is only a very slight increase in the swept volume of the sensorium with swiveling of the volume in pitch (see the “1.0 (blunt)” curve in Figure 5A). As shown in Figure 4B, this cuboidal approximation overestimates the effect of pitching the conical visual sensorium of the stone moroko.
We will simplify the analysis slightly by 1) making the idealization that projected area does not change with pitch angle because of the aspect ratio of the visual sensorium, and by 2) allowing the projected area of the electrosensory sensorium at , (Figure 5), stand for the projected area of the visual sensorium at , which is 11.9 cm18.7 cm or . This facilitates comparison to the electrosensory case.
The energetic consequence of this visual sensorium is then obtained by clamping the projected area () term of the equation to its value when the body is pitched at , as shown in Figure 7C by the black ‘+’ curve. The energy to overcome drag monotonically increases; the benefit of holding the body at a pitch is lost.
Long range sensing organs, such as eyes and pinna, are often clustered and invested with muscles that enable them to rotate, which in turn rotates their associated sensorium. What effect does sensorium mobility have on the amount of energy needed to encounter prey? In another hypothetical scenario, we examined the consequences of the fish being able to pitch its sensorium around its head without moving its body, illustrated in Figure 7B. We do this by clamping the drag force () term of the equation for above to its value at , with the result shown in Figure 7C by the red ‘x’ curve. There is a substantial decrease in energy needed per prey. Whereas this sensorium mobility is not biologically possible due to the near-field and broadly distributed nature of electrosense, this example serves to illustrate how sensorium mobility for a far-field sensory system can have beneficial consequences.
Given the limited availability of energy, all animals must balance the energy load of sensory and neuronal systems with motor and other body systems. However, active sensing animals such as bats, dolphins, and electric fish, have a particularly stringent constraint: they must generate the energy required to perceive their world. Both emitted energy and energy reflected from objects falls as [23], so that the total power attenuation is inversely proportional to . By this, a doubling of sensory range takes sixteen times more energy.
As an example of how constraining the physics of active sensing energy attenuation is, we consider the power the electric fish has to emit to detect prey. The electric fish's self-generated electric field allows them to detect prey at less than a body length away from the body [16]. The energetic cost of electric signal generation was recently measured at 3–22% (depending on time of day and gender) of the total metabolic rate [24]. For a 350 J/day total energy budget for the black ghost knifefish [25], this amounts to a peak of up to about 80 J/day. This power level enables them to detect prey at up to 3 cm [16]. To detect prey at twice this distance, or 6 cm, would require , or 16 times more energy, or 1,280 J—four times the total energy budget of the fish. Although the signal generation power measurements used here are for a different species of South American weakly electric fish, the argument is hardly affected even at an order of magnitude lower power.
Given these simple estimates, while all animals have to contend with trade-offs between more energy devoted to sensory systems versus other systems, we can expect these trade-offs to be especially clear in active sensing animals such as electric fish.
Figure 8 shows one of our key results in summary form. We have found that as the body pitch increases from zero to , the drag force increases by a factor of between 2–4 times at a search swimming velocity of 15 cm/s. However, this increase in pitch angle also results in a near doubling in the search rate as quantified by projected area of the sensorium. In the simplified model, the balance of these two factors, which is quantified by the energy required to reach one prey (Figure 5B), results in a best pitch angle of around . This results in a 40% energy saving over swimming at . Put another way, the number of prey encountered over a given distance of movement will be nearly doubled due to the near doubling of projected sensorium area.
The measured fish pitch angle during search was [16], significantly different from the optimum found here. There is an additional factor which will have the effect of reducing this disparity. This is a reduction in propulsive force from the ribbon fin with increased pitch angle, as shown in Figure 6. In this figure, each solid curve shows how the thrust from the fin, with a traveling wave at the indicated frequency, decreases as the body pitches. Across the different undulation frequencies, the propulsive effectiveness of the fin drops around 25% at . If this effect were to be fully incorporated, the optimal swimming angle would clearly be less than .
To illustrate this relationship, consider the dashed line of Figure 6, which shows the drag force on the body when total power expended for swimming () is clamped at a specific value, calculated given the drag force () at a pitch angle of and a velocity () of 15 cm/s (0.2 mN15 cm/s, which is 0.3 mW). Given that , an increase in drag requires a decrease in velocity to keep power fixed, resulting in a lower velocity as drag increases with pitch angle. Therefore, the dashed line indicates the thrust needed to overcome this drag when the power available is the same as when the fish swims horizontally. The intersection between the dashed line and the thrust curve indicates the approximate pitch angle required to move with a constant velocity, as propulsive thrust balances body drag resulting in zero acceleration. In particular, at the highest undulation frequency shown, 6 Hz, the fish would need to swim at , significantly below the pitch that would be best if loss of thrust with increased pitch were not a consideration.
While we have found that the mechanical energy needed to find each prey is on the order of tens of microjoules, a small fraction of the energy gained per prey (on the order of a joule; see Table 1), the mechanical energy expended to overcome drag is only a fraction of the total energy the animal will use in finding each prey. This is because 1) not all the energy in food is converted to available energy [26]; 2) not all the available energy is used for swimming muscles (e.g., we estimate the mechanical power used for swimming at pitch and 15 cm/s is 0.3 mW (the velocity times the drag at this velocity, 15 cm/s2 mN), while metabolic rate is on the order of 0.4 mW [25]); and 3) not all the energy used for swimming muscles is converted into thrust. These factors combined are around a factor of ten. There is also significant uncertainty in the prey density numbers. The energy needed per prey will double if the prey density is half that used for these estimates (5,000 prey per cubic meter). The density appears to vary between 1,000–5,000 individuals per cubic meter for rivers typically inhabited by Amazonian electric fish [27], [28]. However, this includes many different insect species and it is unclear what fraction of these are prey the fish would eat. Despite these uncertainties, the fish has few ways at its disposal for increasing search rate at a given velocity beyond changing pitch angle. For example, it cannot increase its sensorium size because it does not vary its electric field strength, although another species of weakly electric fish has recently been shown to vary its electric field strength [29]. Thus the increased mechanical load on the fish with increased body pitch is an appropriate variable to examine.
A key factor in the beneficial effect of pitching the body is the shape of the sensorium. More specifically, how the projected area of the sensorium changes as a function of the sensorium position control variable, in this case body pitch (), is crucial. As the sensorium becomes less elongated, the increase in projected area with increased pitch angle becomes negligible, and thus the benefit of body pitching disappears. This is shown in Figure 5A and B. As the sensorium becomes more elongated (elongation factor 4.0), the projected area increases more rapidly with pitch angle, and the net energy needed per prey decreases more rapidly. The opposite holds for the cube-like sensorium (elongation factor 1.0): there is nearly no increase in projected area with pitch angle, and thus the energy needed per prey only increases with pitch angle due to increased drag forces.
As another way to examine this effect, we computed the energetic consequences of the black ghost using a visual sensorium, illustrated in Figure 7A. The visual sensorium for the detection of the same type of prey used in this study, Daphnia, in a visual predatory fish (the stone moroko) has been measured to be 11.9 cm high (vertical)×12.0 cm long (distance from eye to leading edge)×18.7 cm wide (left-right) (Figure 4B). This sensorium has an elongation factor (length to height) of unity, so the projected area changes little with rotation in pitch. As a visually-guided animal with movable eyes, the stone moroko can choose to rotate its eyes with its oblique muscles to control the pitch angle of its sensorium [30]. For the purpose of this example, let's facilitate the comparison to the elongated body-fixed sensorium of the black ghost by supposing that this artificial visual sensorium is also body-fixed, as depicted in Figure 7A. Thus, the fish changes the pitch of its body to change the pitch of the sensorium. The effect of this faux visual sensorium on energy is shown in Figure 7C. There is no benefit to pitching when the effect of the elongated sensorium is removed, and only the cost of overcoming drag remains for the artificial case of a body-fixed visual sensorium.
These results indicate that an elongated sensorium is beneficial. In this particular group of species, an elongated sensorium goes along with an elongated body that is characteristic of the knifefish body plan, common across some 180 different species (Gymnotidae) [31], and the distributed nature of the electrosensory system of these fishes.
For the stone moroko, a fish which swims by “tail-wagging” (the carangiform mode), the instability in yaw induced by tail beating results in high yaw maneuverability [1], and would facilitate prey capture lateral to the fish body. In addition, left-right eye movements will sweep the sensorium in azimuth. Therefore, this fish's vertically flattened sensorium, over one-and-a-half times wider than it is tall, seems likely to be beneficial. Further amplifying this point, Figure 4B shows that pitching the sensorium would in fact decrease the swept volume slightly. The relevant elongation factor for this fish will be length to width, since height will only have a constant factor effect on how projected area changes with azimuthal angle.
Weakly electric fish have a body-fixed sensorium. If it were at all possible to change the position of the sensorium without changing body position, as animals that rotate their eyes or turn their heads can [9], one possible scenario would allow the animal to have all of the sensory advantages of pitching the body, with none of the drag costs. In this scenario, imagine the fish could tilt the back of its sensorium up as illustrated in Figure 7B, but without tilting the body—analogous to how some animals can rotate their visual sensorium without moving their bodies. We can assess the energy implications of this scenario through the use of Eq. 2, by fixing the drag force () to its value when the body is at pitch, while allowing to vary. The result is shown by the ‘x’ curve of Figure 5B. Being able to dynamically reposition the sensorium without moving the body results in more than a factor of two decrease in energy per prey at , and even more at larger angles.
Decoupling sensorium movement from whole body movement has been an ancient theme of vision, our most powerful teleceptive sensory modality. Independent eye movement and stabilization goes back to the very first vertebrates [32]. There are many benefits to eye movements, such as minimization of motion blur due to self movement and movement of the object of fixation [9], but clearly not having to reposition the body to see something initially out of view can economize on energy [30]. Given that body mass is considerably larger than sensory organ mass, it also saves on time. One cost, however, is the need to translate the coordinates of perceptual information arriving in sensory-organ-fixed coordinates to the coordinates of the body, demanding significant neuronal processing. The tectum, or superior colliculus, is one structure where this occurs (review: [33]).
Whereas eye movement is quite ancient, the ability to turn the head is relatively recent in vertebrates. Our earliest evidence of this ability is from a 375 million year old fossil of an animal that appears to be a transitional form between fish and tetrapods, Tiktaalik roseae [34]. Some active sensing animals exploit head movements for sweeping their sensoria horizontally and vertically while keeping their body on a fixed course. For example, bats nearly double the angular range of their sonar-based sensorium by combining pinna and head movements [35]–[37], and dolphins have also been shown to use head movements to manipulate their sonar-based sensorium to a similar end [38]. Rats also exploit this freedom, combining head movements with whisker movements to palpate objects [39].
Having relatively light and independently movable sensory appendages is a ubiquitous feature of animal body plans. It is particularly powerful for teleceptive systems such as vision and audition. The analysis here highlights how advantageous it can be to decouple sensorium movement from whole body movement from an energetics standpoint. It may also suggest that when developing assistive technologies for people with sensory challenges, a sensorium whose movement is independently controllable from body movement can be particularly helpful.
Although there is a significant literature of how mechanical considerations enter into sensory performance in a large number of systems, and a growing literature on the metabolic cost of information, there has been little examination of how these two domains overlap and trade-off with one another. While measures of performance in these two areas typically are not commensurable, the impact of a change in sensing or movement on the net energy balance of an animal provides a basis of comparison. We have been able to quantify how this animal trades-off movement efficiency for sensory performance in prey search behavior. A simplified model illuminates why the animal searches with its body in a drag-inducing position, and suggests a possible basis for why this group of animals has evolved an unusual degree of elongation in their body plan. This model also illustrates the benefits of sensorium mobility that is decoupled from whole-body movement.
In the traditional view, the nervous system performs the computational “heavy lifting” in an organism. This view neglects, however, the critical role of morphology, biomaterials, passive mechanical physics, and other pre-neuronal or non-neuronal systems. Given that neurons consume forty times more energy per unit mass than structural materials such as bone [40], and twenty times as much as muscle ([5], after [6]), there are clearly advantages to distributing tasks between these tissues in a way that improves energetic efficiency. In this “bone-brain continuum” view [41], animal intelligence and behavioral control systems can only be understood using integrative modeling approaches that expose the computational roles of both neural and non-neural substrates and their close coupling in behavioral output. The infomechanical approach taken here, in which information and mechanics are jointly examined with regard to energy consequences, is one such approach that can facilitate a more integrative understanding of animal system design.
An accurate urethane cast of a 190 mm long Apteronotus albifrons made for a prior study [42] was bolted to a rigid rod. This was suspended from a custom force balance that used three miniature beam load cells (MB-5-89, Interface Inc., Scottsdale AZ USA). For force balance and calibration details, see [43]. The fish cast was towed through a large tank that was in length, width, and depth (GALCIT towtank, Caltech) using a gantry system driven by a speed-controlled DC servomotor above the tank [43]. Trials were conducted at three speeds: 10, 12, and 15 cm/s, and three angles to the flow: , , and . Only the data collected after the startup force transient had settled was analyzed, until just before the end of the towing distance (300 cm). The data was filtered with a digital Butterworth low pass filter (cutoff at 5 Hz) to remove transducer transients prior to further statistical analysis.
We used a custom computational fluid dynamics solver to obtain the drag force on a fish model at different towing velocities. The fish model was derived from the same urethane cast as was used for the tow-tank measurements [42]. It is assumed to be rigid. In the numerical simulations, it is towed at 10, 15, and 20 cm/s, and three angles to the flow: , , and . All of the simulations were performed using the San Diego Supercomputer Center's IA-64 Linux Cluster, which has 262 compute nodes each consisting of two 1.5GHz Intel Itanium 2 processors running SuSE Linux. The computational fluid dynamics code was written in Fortran 90 and C (for details, see [21]).
In a prior study we used a combination of empirical measurements and computational models to determine the 3D volume around the fish body where a typical prey item, Daphnia magna, could be detected (Figure 4A) [12]. We idealized the resulting electrosensory sensorium as a cuboid (Figure 9) whose width, height, and length is matched to the maximal dimensions of this volume, after scaling for body size (the body length for the [12] study was 14.4 cm, while it is 19.0 cm in this study). The resulting dimensions are shown in Table 1.
As shown by Figure 9, the projected area of this cuboidal sensorium in the direction of travel (its silhouette if you were to look at it directly along the path of its approach) is simply . We varied the ratio of the length to the height of the cuboidal sensorium. To assess the impact of elongation of this volume on projected area with pitch angle , we varied the ratio of the length to the height of the cuboidal sensorium (the elongation factor). These two dimensions were chosen because by the above equation for , varying the width only results in a constant factor change in the projected area with respect to . The naturally observed elongation factor was 2.2.
We assume that prey are uniformly distributed at the density shown in Table 1. As shown in Figure 9, the projected search area is . Thus, the total water volume scanned for prey when the fish moves distance will be . The number of prey detected in that volume will be the volume times the prey density, or . The distance travelled to get one prey will then be .
We fit our measured drag data to a function of the form , where is in degrees. The result is , where and , with an of . Thus thrust power . We can rearrange this to solve for .
We rearrange to solve for and use the solution for from above to solve for , the time required to find one prey. Then we multiply this by to solve for the energy expended to overcome drag in obtaining one prey:(2)
For , we used the power needed to overcome the experimentally measured tow drag at pitch and 15 cm/s, which was 0.3 mW (15 cm/s2 mN) (Figure 3).
For a previous study we used a computational model of a non-translating, non-rotating fin deforming in a sinusoidal pattern with time [21]. The instantaneous velocity of each point on the fin is specified as a function of time. The no-slip and no-penetration boundary conditions are imposed on the surface of the fin using an immersed boundary formulation, and the fluid flow around it is fully resolved using finite difference methods of order in space and order in time. The complete details of the computational algorithm and method are given in [21], [44].
Mean forces on the fin were calculated as the time average of the hydrodynamic forces on the fin over at least one period of oscillation, after a quasi-steady state is reached. As shown in [21], the force in newtons from the fin followed the correlation(3)where is a constant equal to 86.03, is the density of water (), is the frequency of the traveling wave on the fin (Hz), is the maximal angular excursion of the traveling wave (radians), is the fin length (m), is the height of the fin (m), is the wavelength of the traveling wave (m), and is a function of the specific wavelength which can be approximated by:(4)
This equation estimates the propulsive force parallel to the fin, or surge force. However, in addition to this force, the fin also generates a small force that is perpendicular to the fin base, pushing the body upward. This force, termed heave, has a magnitude of about 25% of the surge force for typical motion patterns [21]. Because of the relative magnitudes of the surge and heave forces, the angle of the fin that would maximize forward thrust is . This angle is nearly identical to the observed fin insertion angle on the body ( in Figure 1B) when the fish is swimming straight. By knowing the surge force, and this angle, we can therefore compute the heave force as the tangent of the fin base angle times the surge force. As the body pitches, the contribution of the parallel surge force to thrust will vary with the cosine of the sum of the body pitch angle and fin base angle , whereas the contribution of the normal heave force will vary with the sine of the sum of these two angles. Thus the net force will be:(5)where is the body pitch angle, and is the angle of the fin base with respect to the body axis at body pitch (; shown in Figure 1B). For these force estimates, we used the length of the fin of the fish used for drag estimates (12.7 cm), a fin height of 1 cm, and typically observed kinematic values of an , and two waves along the fin ( cm) [20], [21].
To compare thrust to drag when the power expended on swimming is fixed, we derive the relationship between the drag function and swimming power. Based on and , . Thus .
To assess the effect of sensorium shape, we examined elongation factors of 1.0 and 4.0 by changing the sensorium length to be equal to its normal height, and four times its normal height, respectively. We then examined the energetic consequences of these sensorium morphologies. This was done through the equation describing the energy needed per prey encounter described below (Equation 2) through changing the function () that returns projected sensorium area given the pitch of the body.
For the artificial elongation factors of 1.0 and 4.0, we make the following simplification. A change in elongation factor normally would be accompanied by a change in body elongation. This is because the electric organ and sensors, which together form the sensorium [12], are along the full length of the fish; therefore a change in relative length of the sensorium would necessitate a change in body length. Any change in body length would in turn affect the drag force on the body and thus the energy needed per prey through the term of Equation 2. Although this was not considered here due to the extensive computational demands of the drag study, the results of simple sensitivity analyses suggest that this simplification has negligible effect on the qualitative trends.
We examined the energetic consequences of two “what if” scenarios: 1) There is no increase in projected sensorium area as the body pitches. To do this, we clamp to its value at . 2) There is no increase in body drag as the body pitches. This would be the case if the fish were able to independently control the pitch angle of its sensorium, analogous to how animals with movable eyes or heads can change the position of their visual sensory volume without changing body position. To do this, we clamp the drag term to its value at .
|
10.1371/journal.pntd.0003342 | Rapid and Sensitive Detection of Bartonella bacilliformis in Experimentally Infected Sand Flies by Loop-Mediated Isothermal Amplification (LAMP) of the Pap31 Gene | Carrion' disease, caused by Bartonella bacilliformis, remains truly neglected due to its focal geographical nature. A wide spectrum of clinical manifestations, including asymptomatic bacteremia, and lack of a sensitive diagnostic test can potentially lead to a spread of the disease into non-endemic regions where competent sand fly vectors may be present. A reliable test capable of detecting B. bacilliformis is urgently needed. Our objective is to develop a loop-mediated isothermal amplification (LAMP) assay targeting the pap31 gene to detect B. bacilliformis.
The sensitivity of the LAMP was evaluated in comparison to qPCR using plasmid DNA containing the target gene and genomic DNA in the absence and presence of human or sand fly DNA. The detection limit of LAMP was 1 to 10 copies/µL, depending on the sample metrics. No cross-reaction was observed when testing against a panel of various closely related bacteria. The utility of the LAMP was further compared to qPCR by the examination of 74 Lutzomyia longipalpis sand flies artificially fed on blood spiked with B. bacilliformis and harvested at days (D) 1, 3, 5, 7 and 9 post feeding. Only 86% of sand flies at D1 and 63% of flies at D3 were positive by qPCR. LAMP was able to detect B. bacilliformis in all those flies confirmed positive by qPCR. However, none of the flies after D3 were positive by either LAMP or qPCR. In addition to demonstrating the sensitivity of the LAMP assay, these results suggest that B. bacilliformis cannot propagate in artificially fed L. longipalpis.
The LAMP assay is as sensitive as qPCR for the detection of B. bacilliformis and could be useful to support diagnosis of patients in low-resource settings and also to identify B. bacilliformis in the sand fly vector.
| Carrion's disease, caused by Bartonella bacilliformis remains truly neglected due to its focal geographical nature. A wide spectrum of clinical manifestations, including asymptomatic bacteremia can potentially lead to a spread of the disease into non-endemic regions. The PCR-based approach is sensitive for detection of B. bacilliformis but requires a thermocycler, thus limiting its use in remote endemic areas. LAMP is a simple method capable of detecting B. bacilliformis DNA within an hour under isothermal conditions, requiring less specialized equipment for amplification, thus enabling diagnosis in rural areas. This study demonstrated that the detection limit of LAMP, targeting the pap31 gene of B. bacilliformis, was comparable to that of qPCR. With a high, targeted selectivity, LAMP showed a high specificity as no cross-reaction was observed when testing a panel of closely related bacteria. The utility of the LAMP assay was further demonstrated by the examination of sand flies artificially fed on blood spiked with B. bacilliformis. The results showed that LAMP was able to detect B. bacilliformis in all flies confirmed positive by qPCR. This study showed that LAMP can be useful to support diagnosis of patients in low-resource settings and also to identify B. bacilliformis in the sand fly vector.
| Carrion's disease, caused by Bartonella bacilliformis remains truly neglected due to its focal geographical nature, occurring in small rural communities of inter-Andean valleys between altitudes of 500–200 meters above sea level in Peru, Columbia and Ecuador [1]. The disease typically presents with acute fever and devastating hemolysis known as Oroya fever, followed by asymptomatic bacteremia and chronic eruptive disease, called Verruga peruana [2]. As the period of asymptomatic bacteremia may vary and last for up to 15 months, these individuals can presumably serve as a reservoir for the bacteria, leading to potential transmission by the putative sand fly vector, Lutzomyia verrucarum, to either populations living in endemic areas or travelers visiting such regions [3]. The increase of human migration, habitat destruction, as well as adaptation of sand flies association with human, may also aid the spread of the potential vector [4]. With changes in weather patterns or host dynamics, sporadic and occasional epidemics of bartonellosis have occurred over a much larger area than previously described [5]–[8].
L. verrucarum have been found throughout the western half of Peru between 1500 and 3200 meters above sea level in Occidental and Interandean valleys of the Andes mountain [9]. This species was presumed to be the primary vector of B. bacilliformis, based on the evidence derived from the presence of the flies in the areas in which the disease occurred [8]. However, Carrion's disease is found in areas where L. verrucarum is not present which implied that other Lutzomyia sand flies may serve as a vector [4]. Additionally, the mechanisms by which transmission occurs, mechanically through surface contamination of the sand fly or biologically through B. bacilliformis infected sand flies, remain unanswered.
With a recent increase of migration in human populations and the possible existence of other unrecognized vector species as well as a great abundance of other sand flies species such as Lutzomyia longipalpis at the lower elevations in South America, a growing concern is whether or not other sand flies species are susceptible to B. bacilliformis and have the capability of transmitting infection. Generally, oviposition of sand flies starts on the fifth day after blood meal. If sand flies were able to transmit B. bacilliformis, the bacteria should be able to establish infection in sand flies through this period and persist long enough to be transmitted [10].
Diagnosis of bartonellosis remains challenging because each test has its own limitations. Gimsa-stained blood smear is the cheapest and quickest diagnostic method being used to diagnose acute disease but the test suffers from low sensitivity (36%) [11]. Serological tests such as IFA or immunoblot show promising results but require paired samples of acute and convalescent phases and may not be useful to diagnose acute disease [12]. Isolation of the bacteria may require up to 4 weeks before they are considered negative [13], thus it may not be useful for case management. PCR amplification of DNA shows promising sensitivity and specificity but requires a high-precision thermal cycler, thus it is impractical for diagnosing the disease in remote areas [11]. A rapid and sensitive test capable of detecting B. bacilliformis DNA is of great clinical importance, not only for early diagnosis and treatment but for better understanding of true disease burden, natural history of the disease as well as the role of L. verrucarum or other Lutzomyia species in transmission of the disease.
Loop-mediated isothermal amplification (LAMP) is a nucleic acid amplification method, generating up to 109 fold amplification within an hour under isothermal conditions; hence, it is simpler and requires less specialized equipment than conventional or real time PCR [14]. In this study, we aimed to develop the LAMP assay targeting heme-binding protein pap 31 gene for detection of B. bacilliformis compared to qPCR using known positive samples including plasmid containing the targeted gene and B. bacilliformis genomic DNA. Additionally, the utility of the LAMP assay was evaluated by testing L. longipalpis fed on blood infected with B. bacilliformis and comparing the detection limit to that of qPCR. A recent study has shown that an in vitro feeding method using a natural skin membrane and an artificial feeder is a viable alternative to the use of the in vivo method using anesthetized hamsters for blood-feeding sand flies [15]. In addition, in the situation where there is no known non-human host like for human bartonellosis, an artificial feeding system is best fitted for studying transmission. According to the results obtained, we should be able to determine whether B. bacilliformis would be able to propagate in L. longipalpis sand flies.
This study was reviewed and approved by the ethics committee of the Uniformed Services University of the Health Sciences. Human blood used in this study was obtained from stored blood that expired from Armed Services Blood Program (ASBP), Maryland. There was no IACUC review the animal care in this study since there was no animal used other than skin from dead quails which were residual from toxicology testing at U.S. Army Public Health Command (USAPHC).
To determine the analytical sensitivity of the LAMP assay, a plasmid containing the pap31 gene sequence of ATCC strain 35685 (KC 583) (accession number DQ207957) of B. bacilliformis was constructed as previously described [16]. In brief, the sequence between the forward primer (5′- gcagcatatgttatgatcccgcaagaaata-3′) and the reverse primer (5′-ctaaaggcacaaccacaacgcattcttaag-3′) was amplified. The amplified product was then cloned into pET24a (Novagen, San Diego, CA) and sequence confirmed to be used as the standard for nucleic acid amplification.
The fluorescent strain of B. bacilliformis KC583 (ATCC 35685) (gloBart) was grown in heart infusion agar blood plates (HIAB; heart infusion agar [Difco, Detroit, Mich.] containing 4% sheep red blood cell [Quad 5, Ryegate, Montana] and 2% sheep serum) using the standard methods described by Minnick for cultivation of B. quintana [17], [18].
Two-to-four day-old colony-reared female sand flies were placed in a 1-pint paper feeding carton fitted with a screen top. All flies were fed through a quail-skin membrane (US Army Public Health Command, Aberdeen Proving Ground, MD) on human blood inoculated with a fluorescent strain of B. bacilliformis (gloBart). A membrane feeder fitted with the quail skin membrane was placed on top of the feeding cup with the membrane pressed against the screen. Heparinized human blood inoculated for an hour with fluorescent strain of B. bacilliformis (gloBart) was added, using a sterile glass pipette inserted through the top of the feeder into the inner chamber until it was approximately two-thirds full with blood. The flies were allowed to feed under conditions of 22–24°C and 75–80% relative humidity. The outer chamber of the membrane feeder was connected to a circulating water bath to maintain the blood temperature at 37°C throughout the feeding period. Fully engorged sand flies were retained in the cages for another 24 h before being transferred to fresh cages. They were kept under the same conditions and were provided a solution of 15% sucrose before being harvested at days (D) 1, 3, 5, 7, and 9 post blood feeding. At least 10 fully engorged females were harvested at each day.
DNA extraction of the whole genome of B. bacilliformis was performed using the DNeasy Blood & Tissue Kit (QIAGEN, Germany) following the manufacture's protocol. To extract DNA from non-infected sand flies spiked with B. bacilliformis, 20 µL of a culture suspension of B. bacilliformis was added to 180 µL of Buffer ATL (QIAGEN, Germany) and a single sand fly was added to the solution. Similarly, 180 µL of Buffer ATL was added to single or pooled, artificially-fed sand flies before being homogenized with a disposable plastic homogenizer. The mixture was added together with 20 µL of proteinase K and incubated at 56°C for 2 h. Next, 200 µL of Buffer AL (QIAGEN, Germany) was added to the sample and incubated at 70°C for 10 min. Then, the DNA extraction was performed using the manufacture's protocol. Two elutions were done with 50 µL of low EDTA buffer (final volume was 100 µL) and each time the buffer was incubated on the column for 10 min.
The oligonucleotide primers used for the LAMP assay were designed based on the pap31 gene sequence from the ATCC KC 583 strain of B. bacilliformis. Using PREMIER Biosoft (http://premierbiosoft.com), a highly specific set of primers including two outer (F3 and B3), two inner (FIP and BIP) and loop primers (LF and LB) were used. All primers were synthesized by Eurofins MWG Operon (Huntsville, AL) and are described in Table 1.
The LAMP reactions were carried out in a 25 µL reaction volume containing 1.6 µM of each of the FIP and BIP primers, 0.4 µM of the F3 and B3 primers, 0.8 µM of the LF and LB primers, 20 mM Tris-HCl (pH 8.8), 10 mM KCl, 8 mM MgSO4, 10 mM (NH)4SO4, 0.1% Triton X-100, 0.8 M betaine (Sigma-Aldrich, St Louis, MO), 1.4 mM dNTP mixture (New England Biolabs, Beverly, MA), 8 U Bst DNA polymerase (New England Biolabs, Beverly, MA), and 5 µL of DNA template. All reactions were prepared in duplicate. The reaction mixture was incubated in a Biometra thermocycler (Applied Biosystems, Foster City, CA) at 59°C for 60 minute. Each reaction was terminated by adding 5 µL of 10X BlueJuice (Invitrogen, Carlsbad, CA). The LAMP products were examined by electrophoresis on a 2% agarose gel stained with a 1∶20,000 dilution of ethidium bromide (Sigma-Aldrich Products, St Louis, MO) and the positive samples were identified by the appearance of a typical ladder banding pattern. LAMP was considered positive when both duplicates were positive.
This assay used primers designed against the pap31 gene sequence of the ATCC KC 583 strain of B. bacilliformis (https://www.genscript.com/ssl-bin/app/primer). The primer sequences are described in Table 1. In brief, each reaction mixture contained 750 nM of the forward primer, 750 nM of the reverse primer, 1×RT2 SYBR Green qPCR Mastermix (SA-Biosciences, Frederick, MD), 5 µL of plasmid or genomic DNA of B. bacilliformis or 1 µL of DNA consisting of human or sand fly DNA spiked with B. bacilliformis or from that of artificially-fed sand flies and water added to a final volume of 20 µL. The qPCR reactions were performed and analyzed using the 7500 Fast Real-time PCR system (Applied Biosystems, Foster City, CA), with an initial 5 minute activation step at 95°C, followed by 40 cycles of 95°C for 10 seconds, 60°C for 30 seconds and a melting curve determination cycle. The samples were tested in duplicate. The results were reported as mean of Ct values and a detection range expressed in copies/µL.
To determine the analytical sensitivity, 10-fold serial dilutions of plasmid DNA across a range of 106 to 1 copy (per reaction) and sterile water were used to define the limit of detection (LOD) as compared to qPCR. When using genomic DNA of B. bacilliformis as the template, the copy number was determined by qPCR and the LOD of LAMP was determined as compared to that of qPCR. In addition, to ensure that various host components would not interfere with the detection limit of the LAMP assay, the genomic DNA of B. bacilliformis was spiked with normal human plasma and a single clean sand fly before extraction.
The analytical specificity of LAMP was determined using genomic DNA of various Rickettsia species including R. conorii, R. rickettsii, R. typhi, Orientia tsutsugamushi, Leptospira interrogans, B. henselae, the species of bartonella that are abundant in Peru including B. quintana and B. rochalimae (ATCC BAA-1498) as well as Leishmania spp. that are present in Peru including L. mexicana and L. braziliensis. Furthermore, interference studies were done by spiking the various bacteria or protozoa (100 copies per reaction) with the genomic or plasmid DNA of B. bacilliformis (10 copies per reaction). Based on BLAST search using pap31 against the entire B. bacilliformis genomic DNA, the number of organisms (i.e. number of copies) is equivalent to the number of pap31 gene.
The utility of the LAMP was further compared to qPCR by examining 74 laboratory reared colonies of L. longipalpis female sand flies that fed on B. bacilliformis infected blood. Sensitivity was calculated as (number of true positive)/(number of true positives + number of false negatives), and specificity was calculated as (number of true negatives)/(number of true negatives+ number of false positives).
The detection limit of the LAMP assay compared to qPCR against different types of known positive samples for B. bacilliformis is shown in Table 2. The LAMP assay was as sensitive as qPCR for the detection of B. bacilliformis, with a detection limit of 1 to 10 copies/µL, depending on the sample metrics.
When testing the LAMP assay using DNA from closely related bacteria including various Rickettsia spp. (n = 3), Orientia tsutsugamushi (n = 1), L. interrogans (n = 1) as well as B. henselae (n = 1) as the template, no amplification was observed for any of these bacteria (Fig. 1A). The LAMP assay developed here did not detect B. henselae even though both possess pap31 gene. Furthermore, no interference of amplification was observed when testing the LAMP assay with the samples containing genomic DNA of B. bacilliformis and that of other bacteria (Fig. 1B). When testing LAMP assay using only genomic DNA of B. quintana (n = 1) or B. rochalimae (n = 1) and two species of Leishmania including L. mexicana (n = 1) and L. braziliensis (n = 1) at 5,000 copies per reaction (i.e. 500 fold more than the detection limit of B. bacillformis [10 copies per reaction]), no amplification was observed for any of these organisms. Additionally, the presence of 10 fold excess of B. quintana, B. rochalimae or two species of Leishmania DNA to B. bacilliformis plasmid DNA showed that there were no interferences on amplification of B. bacilliformis DNA by pap31 LAMP assay. Furthermore, there was no cross reactivity between B. bacilliformis primers and DNA extracted from human blood.
Laboratory reared colonies of L. longipalpis female sand flies (n = 74) were fed through a quail skin membrane on blood containing B. bacilliformis. The number of blood-fed flies that were harvested at the day of interest was as follows: D1 (n = 15), D3 (n = 11), D5 (n = 12), D7 (n = 16) and D9 (n = 20). The bacterial DNA was extracted from individual sand flies from D1 and D3. Four individual sand flies from D5, 7 and 9 along with 2 pools (4 flies/pool) from D5, 3 pools (4 flies/pool) from D7 and 4 pools (4 flies/pool) from D9 were extracted. Flow diagram of a diagnostic accuracy in detection of pap31 in sand flies by LAMP compared to qPCR was shown in S1 Figure.
As shown in Table 3, 86% of L. longipalpis (13/15) at D1 post feeding on infected blood were positive by qPCR, with the mean copy number of 64.6 copies/µL (range 7–382 copies/µL) and 63% of flies (7/11) at D3 post feeding were positive by qPCR, with the mean copy number of 71 copies/µL (range 10-302 copies/µL). All individual and pooled flies after D3 were negative by qPCR assay.
As shown in Table 4, the sensitivity of the LAMP assay at D1 and D3 was 100%. The specificity of the LAMP assay at D3, D5, D7 and D9 was 75%, 66.6%, 100% and 100%, respectively. Our results suggest that LAMP targeting the pap31 gene is highly sensitive and moderately specific for detection of B. bacilliformis in experimentally infected sand flies compared to qPCR.
We reported here the development of the LAMP assay targeting the pap31 gene to detect B. bacilliformis. The limit of detection (LOD) for the LAMP assay ranged from 1 to 10 copies/µL, depending on the sample metrics, which was comparable to that of real-time PCR which ranged from 2 to 18 copies/µL. In addition, no cross-reaction was observed when testing a panel of other closely related bacteria. We also demonstrated the capability of the pap31 LAMP assay to identify B. bacilliformis in the sand fly vector. LAMP was able to detect B. bacilliformis in all those flies confirmed positive by qPCR. However, none of the flies after D3 were positive by either LAMP or qPCR.
Carrion's disease caused by B. bacilliformis, has been described as an exotic disease, confined to remote areas of certain mountain regions in South America [1]. Once infected, the patients can remain persistently bacteremic due to the existence of the intra-erythrocytic phase, thus providing a protective niche for the bacteria that is competent for vector transmission [19]. The PCR-based approach appears to be a sensitive method of detection of B. bacilliformis. Previous studies used PCR methods targeting several genes including the gltA gene, the rpoB gene, the 16S-23S rRNA ITS, or the ftsZ gene in attempts to classify the Bartonella species [20]. There have been reports of using PCR for diagnosis of infection caused by Bartonella spp. and Carrion's disease [21]–[23]. Although these molecular techniques offer a high sensitivity and specificity, it requires a dedicated thermocycler, thus limiting its use in remote rural endemic area. Alternatively, the LAMP method is rapid and simple to perform, requiring only a water bath or heating block for amplification. A sufficient amount of amplified product can be produced within an hour under isothermal conditions, enabling a rapid, molecular diagnosis in rural areas [14].
In the recent years, several studies have indicated that LAMP has a high sensitivity and specificity compared to either conventional, nested or real time PCR for the detection of several intracellular bacteria, such as Coxiella burnetii or Orientia tsutsugamushi [24], [25] or in distinguishing species of several intraerythrocytic protozoan parasites, such as Plasmodium spp. or Babesia spp. [26], [27]. The LAMP assay is also specific due to recognition of six distinct sequences on the targeted gene by six primers. This was demonstrated in our results as no cross-reaction was observed when testing other closely related bacteria. In addition, our targeted gene sequence (pap31) is unique to B. bacilliformis which has only 8% identity with other Bartonella spp. Since most of the homologous sequences are located in the 3′ end of the DNA and all our primers are located outside of the homologous region, thus the target is unique in detecting B. bacilliformis. In addition, neither genomic DNA of human, sand fly nor that of closely related bacteria inhibited DNA amplification in the pap31 LAMP assay.
Previous studies have shown the capability of LAMP for detecting pathogens in mosquitos including Plasmodium spp., Dirofilaria immitis and Wuchereria bancrofti [28]–[30]. This study is the first report whereby LAMP assay was used for detecting B. bacilliformis in sand fly vectors. The utility of the pap31 LAMP assay was evaluated by examining artificially-fed L. longipalpis sand flies on blood infected with B. bacilliformis. When using qPCR as a reference standard, 86% of flies at D1 and 63% of flies at D3 post feeding were positive by qPCR. Overall, LAMP shows a sensitivity of 100% at D1 and D3 and specificity of 100% at D5 and D9 post feeding. However, DNA from one sand fly at D3 and 2 sand flies at D5 was unable to be amplified by qPCR but showed positive for B. bacilliformis via the LAMP assay, resulting in lower specificity of the LAMP assay. Regarding the selectivity of the LAMP assay, a false positive reaction of LAMP is less likely. It must be noted that the pap31 LAMP assay has a slightly lower LOD than that of qPCR. It is possible that the number of bacteria in those sand flies is too low to be detected by qPCR but not by LAMP. In addition, the presence of inhibitors such as hemoglobin, in blood fed sand flies may prevent Taq DNA polymerase in the qPCR assay from extending the DNA in the time allowed, adversely affecting the amplification efficiency but do not inhibit Bst DNA polymerase in LAMP assay [14], [31].
L. longipalpis is known to be an important vector of American visceral leishmaniasis in Latin America. This sand fly species has been shown to be very permissive of Leishmania spp. infection [10]. Even though Bartonella spp. has never been isolated from L. longipalpis, many bacterial infections have been observed from laboratory reared colonies of L. longipalpis [10]. Thus, we used this species to be experimentally infected with B. bacilliformis. To date, it remains unclear if B. bacilliformis is transmitted through surface contaminations or becomes an established infection in sand flies. If the sand flies can support the bacterial growth through the entire gonotrophic cycle, this would indicate a potential for biological transmission. Therefore, the detection of bacteria in sand flies after five days would support its susceptibility to B. bacilliformis infection. The initial pathogen load in the sand flies might affect the detection data for B. bacilliformis from D 1–9. However, we are convinced of sufficient pathogen load in the blood that fed on sand fly despite no culture or PCR testing on B. bacilliformis spiked blood. Firstly, previous study observed that B. bacilliformis had a high erythrocyte invasion rate for up to 80% in patients with Oroya fever which might be owing to a high motility of B. bacilliformis as compared to other Bartonella spp. [32]. In addition, our prior work (unpublished data), determining the RBC infection rate with B. bacilliformis by using flow cytometer showed that the infection rate for all of the blood samples tested (n = 8) was more than 73% (74–89%). Flow cytometry was used since it provided a near real time infection rate which was much more accurate than either culture or PCR testing[33]. Therefore, our results show that none of the flies after D3 were positive by either LAMP or qPCR should suggest that B. bacilliformis is incapable of propagating in artificially-fed L. longipalpis. While the authors recognize that L. verrucarum is the ideal sand fly species to work with, a colony of this species was not available. Future studies should evaluate this technique with L. verrucarum.
The ability of pap31 LAMP to detect even a single bacterium of B. bacilliformis suggests the capability of this method to identify B. bacilliformis in the sand fly vector, which is important for disease control and surveillance. This method could have the potential to support diagnosis of patients in the early stage of illness, thereby leading to prompt initiation of appropriate antimicrobial treatment. With the advantages of being a simple and rapid assay, the LAMP method could be useful for rapid detection of infected sand flies as well as early diagnosis of patients from low-resource settings. Yet, further studies to evaluate this method for the detection of B. bacilliformis in suspected patients are needed to further determine the performance of LAMP assay. Future studies should evaluate this assay with samples from patients with different clinical manifestation. Of interest, when this method was used in combination with a sensitive and specific antibody detection such as enzyme-linked immunosorbent assay (ELISA) using recombinant protein Pap31 (rPap31) [31], they will likely improve detection of all truly infected individual and provide a broader window of detection than was possible with either one assay.
|
10.1371/journal.pgen.1007665 | Hrg1 promotes heme-iron recycling during hemolysis in the zebrafish kidney | Heme-iron recycling from senescent red blood cells (erythrophagocytosis) accounts for the majority of total body iron in humans. Studies in cultured cells have ascribed a role for HRG1/SLC48A1 in heme-iron transport but the in vivo function of this heme transporter is unclear. Here we present genetic evidence in a zebrafish model that Hrg1 is essential for macrophage-mediated heme-iron recycling during erythrophagocytosis in the kidney. Furthermore, we show that zebrafish Hrg1a and its paralog Hrg1b are functional heme transporters, and genetic ablation of both transporters in double knockout (DKO) animals shows lower iron accumulation concomitant with higher amounts of heme sequestered in kidney macrophages. RNA-seq analyses of DKO kidney revealed large-scale perturbation in genes related to heme, iron metabolism and immune functions. Taken together, our results establish the kidney as the major organ for erythrophagocytosis and identify Hrg1 as an important regulator of heme-iron recycling by macrophages in the adult zebrafish.
| Total body iron stores in mammals is a composite of iron absorption from diet and iron recycled by macrophages from dying red blood cells (RBCs). Upon erythrophagocytosis of RBCs, the hemoglobin is degraded and heme is imported from the phagosomal compartment into the cytoplasm so that the iron can be released from heme. Defects in these pathways can lead to aberrant iron homeostasis. The Heme Responsive Gene-1 (HRG1, SLC48A1) was identified previously as a heme importer in the intestine of the roundworm, Caenorhabditis elegans. In cell culture studies, HRG1 was demonstrated to mobilize heme from the erythrophagosome of mouse macrophages into the cytosol. However, the in vivo function of HRG1 remains to be elucidated. The zebrafish is a powerful genetic animal model for studying vertebrate development and ontogeny of hematopoiesis. In zebrafish, the kidney marrow is the adult hematopoietic organ that is functionally analogous to the mammalian bone marrow. In this study, we show that Hrg1 plays an essential in vivo role in recycling of damaged RBCs, and that the kidney macrophages are primarily responsible for recycling heme-iron in the adult zebrafish.
| Heme is an iron-containing porphyrin that acts as an essential cofactor in numerous biological processes such as oxygen transport, miRNA processing, electron transfer, and circadian clock control [1]. In humans, total body iron store is a composite of dietary iron absorption, which accounts for around 10% of the daily iron requirement after compensating for daily iron losses, and heme-iron recycling through clearance of senescent red blood cells (RBCs) [2, 3]. RBCs contain approximately 70% of the overall body iron in the form of heme, and consequently heme-iron recycling is a significant contributor of systemic iron homeostasis [4]. Thus, a better understanding of heme recycling and transport is critical to understanding the role of heme and iron in red cell synthesis and turnover.
The lifespan of mature RBCs is limited in circulation, with approximately40 and 120 days for mouse and human, respectively [5, 6]. Furthermore, circulating RBCs can be subjected to damage under stress conditions leading to hemolysis. When RBCs become senescent or the number of damaged RBCs increases in the circulation, macrophages from the reticuloendothelial system (RES, spleen and liver) contribute to RBC clearance and subsequently promote heme-iron recycling through erythrophagocytosis (EP). Upon degradation of RBCs in the erythrophagosome, heme is imported into the cytoplasm for degradation by the heme-degrading enzyme heme oxygenase-1 (HMOX1) [7]. Defects in erythrophagocytosis (EP) lead to aberrant iron homeostasis, culminating in iron deficient anemia (IDA) or iron overload [8, 9].
The Heme Responsive Gene-1 (HRG1, SLC48A1) was identified previously as a heme importer in the intestine of Caenorhabditis elegans [10]. In ex vivo cultured mouse bone marrow derived macrophages (BMDMs), HRG1 was recruited to the erythrophagosome membranes where it colocalizes with NRAMP1, an iron transporter found on the phagolysosomal membranes, surrounding ingested senescent RBCs [11, 12]. HRG1 mRNA is upregulated during erythrophagocytosis (EP) in mouse bone marrow derived macrophages (BMDMs) and HRG1 protein is strongly expressed in macrophages of the reticuloendothelial system (RES) and specifically localizes to the phagolysosomal membranes [11]. Depletion of HRG1 by siRNA in bone marrow derived macrophages (BMDMs) causes defective heme transport from the phagolysosomal compartments and a failure to upregulate HMOX1 mRNA, demonstrating that HRG1 must mobilize heme from the erythrophagosome into the cytosol. However, whether HRG1 functions similarly in vivo remains to be elucidated.
The teleost fish, Danio rerio(zebrafish), is a powerful genetic animal model for studying vertebrate development and ontogeny of hematopoiesis [13]. In zebrafish, the kidney marrow (head kidney) is the adult hematopoietic organ and is functionally analogous to the mammalian bone marrow. Compared to the kidney marrow, the zebrafish spleen is not well characterized, although it has been proposed to function as a reservoir for RBCs where erythrocytes are stored and destroyed [14]. Histological analysis reveals that macrophages, although rare in the zebrafish spleen, contain phagosomes with erythrocytes and other cellular debris, suggesting that erythrophagocytosis (EP) could occur in the zebrafish spleen [15, 16]. However, direct experimental evidence identifying the tissues that are involved in heme-iron recycling in the zebrafish is lacking. In this study, we show that Hrg1 plays an essential role in the recycling of damaged RBCs and that the kidney macrophages are primarily responsible for heme-iron recycling in the zebrafish.
Owing to ancient whole-gene duplication in teleosts, the zebrafish genome can typically contain more than one orthologue per human protein-encoding gene [17, 18]. Indeed, two hrg1 paralogs are present—hrg1a (slc48a1b) and hrg1b (slc48a1a) are located on chromosome 6 and chromosome 23, respectively (Fig 1A). Both paralogs are phylogenetically related to C. elegans (CeHRG-1 and CeHRG-4), mouse (MmHRG1) and human (HsHRG1) HRG1 (Fig 1B and 1C). Sequence alignment reveal that the protein sequences of Hrg1a and Hrg1b share 73% identity and 86% similarity to each other (Fig 1C). Protein topological predictions show that Hrg1a and Hrg1b contain four transmembrane domains with a cytoplasmic N- and C-terminus and conserved amino acids that have been implicated in heme transport (Fig 1C, asterisk) [19].
To determine the spatiotemporal expression of hrg1, total mRNA was extracted from embryos at different stages, from single cell to 4 days post fertilization (dpf). RT-PCR analysis revealed that the temporal expression patterns of hrg1a and hrg1b are similar and that both hrg1a and hrg1b mRNAs are present in the one-cell embryo (Fig 1D, S1A Fig). As zygotic mRNA expression is turned-on around 3 hour post fertilization, mRNA detected before this stage is typically associated with maternal deposition during oogenesis [20]. Whole-mount in situ hybridization (WISH) confirmed the RT-PCR analysis and revealed that hrg1 is ubiquitously expressed throughout developing embryos, with high expression levels in the central nervous system (Fig 1E, S1B and S1C Fig). qRT-PCR from dissected adult zebrafish tissues showed that hrg1a and hrg1b are expressed at different levels in various organs (S1D and S1E Fig).
Ectopic expression of fluorescent-tagged zebrafish Hrg1a and Hrg1b in HEK293 cells revealed that Hrg1a and Hrg1b colocalize with LAMP1, a lysosomal marker, consistent with previous studies that showed that the worm and human HRG1 localize to endo-lysosome-related organelles [10]. Moreover, coexpression of fluorescent-tagged Hrg1a and Hrg1b showed both proteins colocalize to similar intracellular compartments (Fig 1F). These results are consistent with the presence of potential heme-binding ligands and sorting motifs in the C-terminus of Hrg1a and Hrg1b (Fig 1C) [10, 11, 19]. Expression of zebrafish hrg1a and hrg1b rescued growth of hem1Δ yeast at heme concentrations as low as 0.25 μM, comparable to the established heme transporters CeHRG-4 and CeHRG-1, demonstrating that Hrg1a and Hrg1b are heme transporters (Fig 1G) [10].
To further define the function of zebrafish Hrg1 in vivo, we generated hrg1a and hrg1b double knockout zebrafish with CRISPR/Cas9 genome editing. The hrg1aiq261 mutant allele contains a 61 nt deletion and a 7 nt insertion (-61, +7) in exon 2, resulting in loss of its original ATG translation start site (S2A Fig). Although an ATG site downstream in the hrg1a ORF could be used for alternative translation initiation, this would cause a truncation of 38 amino acids at the N-terminus. The hrg1biq361 mutant allele carries a -61 nt deletion in exon 3, resulting in a protein predicted to contain 28 less amino acids at the C-terminus and addition of 6 extra amino acids (S2A Fig).
To determine whether the truncated forms of Hrg1a and Hrg1b had residual function, we performed heme-dependent growth assays with the predicted mutant ORFs using the hem1Δ yeast mutant. Hrg1aiq261 and epitope-tagged Hrg1biq361 mutant constructs were expressed, as determined by immunoblotting (S2B Fig), but failed to rescue hem1Δ growth even in the presence of 5 μM exogenous heme (Fig 2A). hrg1 double knockout (DKO) zebrafish were generated by crossing hrg1aiq261 and hrg1biq361 mutant alleles (S2C Fig). Intercrossing of hrg1a+/iq261; hrg1b+/iq361 generated progeny with the expected Mendelian ratios (S2D Fig) (chi-square test, p > 0.05) and survival rates for DKO fish were comparable to wildtype Tü and hrg1aiq261/iq261 or hrg1biq361/iq36 single mutants. Immunoblotting of total membrane proteins obtained from 3 dpf embryos revealed no detectable Hrg1 protein in DKO fish (Fig 2B).
To evaluate whether hrg1aiq261/iq261, hrg1biq361/iq361 or DKO mutants have defects in red cell homeostasis, we first evaluated red cell synthesis by collecting embryos and processing them for o-dianisidine staining which interrogates RBC hemoglobin production. No defects in embryonic hemoglobinization were detected at 3 dpf embryos (Fig 2C). Quantification of GFP-positive RBCs, in hrg1 mutants crossed to the globinLCR-GFP transgenic fish, showed comparable numbers of RBCs demonstrating that neither globin expression nor RBC numbers were altered in the absence of hrg1 (Fig 2D, S2E Fig) (Two-way ANOVA, p>0.05). Correspondingly, WISH showed unaltered expression of gata1, a transcription factor required for erythropoietic lineage specification, and ae1, a marker for developing RBCs, further indicating normal erythropoietic differentiation in hrg1aiq261/iq261, hrg1biq361/iq361, and DKO mutant animals (Fig 2E) (n = 30). RBC morphology, as determined by May-Grünwald-Giemsa staining (Fig 2F and 2G), and hemoglobinization levels, as determined by o-dianisidine staining, were normal (S2F and S2G Fig) in RBCs from embryos and adults. Collectively, these results suggested that red cell differentiation and maturation in hrg1aiq261/iq261, hrg1biq361/iq361, and DKO mutants were unaffected.
Next, we determined whether Hrg1 plays a role in RBC degradation in the zebrafish. To analyze turnover, we had to first establish an in vivo model for hemolysis in the adult zebrafish and then locate the tissue involved in recycling damaged RBCs. To induce acute hemolysis and EP, phenylhydrazine (PHZ) was administrated to adult zebrafish. Histological analysis of the kidney, spleen, and liver conducted one day post-PHZ revealed infiltration of large-sized cells with irregular morphology in the kidney and spleen, but not the liver (Fig 3A, yellow arrows). qRT-PCR revealed that expression of the zebrafish hmox1a homolog was significantly upregulated in the kidney, spleen, and liver after PHZ treatment (Fig 3B). Perl’s Prussian blue showed iron-positive pigmentation only in the kidney macrophages but not in the spleen or liver (Fig 3C, yellow arrows), indicating that only the kidneys were responding to hemolysis for heme-iron recycling.
Since macrophage-specific antibodies are not available for the zebrafish, transgenic fish expressing gata1:gfp and mpeg1:gfp, which label the erythroid and macrophage cells respectively, were treated with PHZ to determine whether the large-sized cells were macrophages. Immunohistochemistry (IHC) with anti-GFP antibody detected GFP-positive cells in the kidney of PHZ-treated mpeg1:gfp zebrafish, indicating that the newly-populated cells are macrophages (Fig 3D). Furthermore, PHZ-treatment of gata1:gfp fish enhanced GFP staining within macrophages while GFP staining was restricted only to erythroid cells in untreated fish (Fig 3D). By contrast, GFP staining was absent in the livers of transgenic and WT (Tü) zebrafish (S3A and S3B Fig). Together, these results suggest that macrophages populate the kidney to phagocytose RBCs in response to hemolysis, and that these cells are the primary sites for heme-iron recycling in the adult zebrafish.
To assess whether hrg1 plays a role in red cell turnover, we compared hrg1 mRNA levels in tissues of adult zebrafish exposed to PHZ. qRT-PCR revealed that the levels of of hrg1a and hrg1b mRNA were significantly upregulated by PHZ in the kidney (Fig 4A). By contrast, only hrg1a was upregulated in the spleen, and neither were altered in the liver (Fig 4A). Consistent with these findings, IHC using anti-HRG1 antibodies showed increased Hrg1 staining after PHZ treatment in the kidney macrophages, while the signal was barely visible in the spleen and liver (Fig 4B).
DAB (3,3'-Diaminobenzidine)-enhanced Perl’s iron and Prussian blue staining detected iron accumulation in kidney macrophages of WT zebrafish at 1, 2 and 3 days post PHZ-treatment, but not in the kidneys of DKO zebrafish (Fig 4D, S4A Fig). In contrast to the WT kidneys, o-dianisidine staining revealed accumulation of heme in the kidneys of DKO zebrafish at 2 and 3 days after PHZ treatment (Fig 4E). No Perl’s iron staining above background could be detected in the spleens from WT or DKO zebrafish (S4B Fig). Importantly, the macrophage numbers or morphology in the kidney and spleen were unaltered in response to PHZ-induced hemolysis in the DKO mutants (S4C and S4D Fig, yellow arrows). These results further confirm that loss of Hrg1 results in the accumulation of heme in kidney macrophages due to defects in heme-iron recycling from damaged RBCs.
To determine whether heme and iron-dependent gene expression profiles were perturbed in hrg1 mutant fish, we performed an RNA-seq on total RNA extracted from dissected kidneys and spleens of WT and DKO fish with or without PHZ-treatment. MA-plots of all 24,220 genes annotated in zebrafish genome GRCz10 [21] showed large numbers of differentially expressed genes in kidney and spleen when pairwise comparisons were performed between DKO and WT samples (Fig 5A and S5A Fig). Gene ontology (GO) enrichment analysis between PHZ-treated versus untreated kidney samples revealed that genes related to porphyrin metabolism, heme metabolism, and intracellular sequestering of iron were significantly downregulated, and immune-related genes were upregulated in the DKO kidney after PHZ treatment (Fig 5B and 5C). Strikingly, enrichment analysis to identify connecting pathways and interactomes from the GO data identified hmox1a connecting the heme/porphyrin metabolism sub-network with iron homeostasis (Fig 5D). On contrast to the kidney, GO analysis of the spleen data showed genes related to blood regulation and hemostasis were downregulated and cell-division genes were upregulated (S5B and S5C Fig).
To classify iron metabolism genes in zebrafish, we compiled a list of 86 mammalian iron metabolism genes and performed BLAST homology to identify 124 potential orthologs in the zebrafish genome (S1 and S2 Table). Further examination of these genes identified 20 genes that were significantly downregulated (p<0.0001–0.038) in the DKO kidneys after PHZ treatment (Fig 5E). These genes included ones involved in heme degradation (hmox1a), iron-storage and transport (fth, ftl, slc40a1/fpn1), heme synthesis (alas2, fech), and systemic iron regulation (erfe) (Fig 5E). By contrast, 42 genes involved in iron metabolism regulation (hamp1, tmprss6), macrophage differentiation (spic), and inflammation (il6, il10, il22) were upregulated (p<0.0001–0.049) (Fig 5E). Comparison of the iron and heme metabolism genes from the kidney and spleen revealed 10 downregulated and 13 upregulated genes that were common to both datasets (Fig 5F, S5C and S5D Fig). These results demonstrate that hrg1 deficiency causes significant alterations in gene expression of heme and iron metabolism pathways in zebrafish.
In this study, we demonstrate the in vivo function of Hrg1 in a vertebrate model system. We show that the zebrafish kidney is the primary organ for heme-iron recycling during EP and that zebrafish Hrg1a and Hrg1b are heme transporters that are expressed and upregulated in kidney macrophages after PHZ-induced hemolysis. In agreement with this finding, genetic ablation of hrg1 results in aberrant heme-iron metabolism at the histological and transcriptomic level.
In mammals, the spleen and liver are major organs for heme-iron recycling. In zebrafish, the kidney is the organ where adult hematopoiesis occurs, comparable to the bone marrow in mammals. However, a role in RBC turnover has not been ascribed to the zebrafish kidney. It has been postulated that the zebrafish spleen is the major organ for EP, despite a lack of direct experimental evidence [15, 16]. This rationale is supported by the observation that the majority of the zebrafish spleen is a reservoir of RBCs, with identifiable lymphatic or myeloid cells plus red and white pulps that are comparable to mammals [15, 16]. However, our studies show that the zebrafish kidney is the major site for EP after hemolysis. Indeed, iron staining reveals active heme-iron recycling in the zebrafish kidney, but not the spleen and liver even though expression of hmox1a, which encodes for the heme degrading enzyme, is upregulated in all three tissues [22].
We have previously used pre-mRNA splice-blocking morpholinos, which targeted the boundaries of hrg1a intron 2 and exon 3 (hrg1a_I2E3_MO2) and exon 2 and intron 2 (hrg1a_E2I2_MO1). Both morpholinos resulted in anemic embryos which could be partially rescued with expression of hrg1. Short sequence BLAST searches showed that the morpholino is specific to its targeting site with low off-target sequences in the zebrafish genome. How is it possible for the hrg1a morphants to show anemia while germline CRISPR mutants do not? It is possible that compensatory pathways can buffer against deleterious mutations, an effect typically not observed in transient morpholino knockdowns. Indeed, Rossi et al showed that egfl7 mutants do not show any obvious phenotypes compared to morphants because Emilin2 and Emilin3 can compensate for loss of Egfl7 [23]. Another study showed that tmem88a-/- mutant embryos partially recapitulated but had a much milder phenotype compared to tmem88a morphants [24]. One way to recapitulate the morphant phenotype would be to delete the morpholino target site by removing intron 2. Although we attempted to delete the entire hrg1a and hrg1b locus using two CRISPR guide RNAs, only F0 chimeras were recovered with no germline transmission in the F1 progeny even after screening large numbers (>300) of F1 embryos. We speculate that the hrg1a locus may harbor genetic elements which might be essential for embryonic development.
In humans, heme-iron recycling from senescent RBCs contributes to more than 90% of daily iron requirement, while dietary iron accounts for the remaining 10% [2, 3]. In zebrafish, the individual contribution of iron/heme absorption versus recycling is poorly understood. The zebrafish is typically fed a diet of brine shrimp and dry food extracts every day that is likely to be iron-loaded, together with iron absorbed from surrounding water. While regulation of iron absorption by the Fpn1-Hamp1 axis is well-documented in mice, it is noteworthy that zebrafish can absorb iron by an alternate pathway via their gills [25]. Indeed, we measured hamp1 and fpn1 mRNA by qRT-PCR in isolated livers from zebrafish. Our results show that hamp1 levels are significantly elevated in the presence of PHZ but is equivalent in both, WT and DKO fish (S6 Fig). The upregulation of hamp1 in response to PHZ is consistent with published studies in zebrafish embryos [26]. However, changes in fpn1 expression is statistically not significant even though there is a trend of lower fpn1 levels in DKO fish exposed to PHZ. Whether the Fpn1-Hamp1 axis regulates iron absorption from the gills is currently unknown. Therefore, one possible explanation for why the hrg1 mutants lack overt erythropoietic phenotypes could be because zebrafish can obtain sufficient amounts of iron from dietary absorption or from the water [27].
Our RNA-seq results indicate that not only genes involved in heme-iron metabolism, but also immune-related genes are differentially expressed in PHZ-treated DKO zebrafish. One top hit of upregulated genes in DKO mutants is an unannotated gene named si:ch211-201o1.1, which is an NLRP3 homolog in zebrafish. It has been reported that an increase in extracellular heme activates NLRP3 expression, triggering inflammasome activation [28]. One possible explanation for this upregulation would be that the heme accumulating in the kidney macrophages in the absence of Hrg1 triggers an inflammatory response. It will be interesting to determine the immune response of Hrg1 mutant fish in the presence of pathogens and inflammatory agents that causes hemolysis.
Little is understood about red cell recycling and turnover in the adult zebrafish as the vast majority of studies related to red cell development and heme and iron metabolism have been confined to embryos. One major advantage of the zebrafish over other vertebrate models is the ex utero transparency of early-stage embryos which are easier to genetically and chemically manipulate. It is therefore noteworthy that hrg1a and hrg1b mRNA are both expressed in the one-cell embryo raising the possibility that Hrg1 and heme may play a hitherto underappreciated role in early embryonic development. Although there are no obvious phenotypes in the DKO mutant embryos, it is possible that these embryos may be more sensitive to maternal iron deficiency or disruption in heme biosynthesis. Here, generating transgenic zebrafish expressing genetically-encoded heme sensors/reporters [29, 30] or label-free imaging [31] to directly visualize heme trafficking at the maternal-embryonic interface in the transparent embryo, and within the hematopoietic organs during heme recycling in the adult fish will significantly influence our understanding of the role of heme and iron in vertebrate red cell development.
All zebrafish procedures were approved by University of Maryland College Park Animal Care and Use Committee (#R-NOV-17-52). The Zebrafish Tü strain was used as wild-type. The light providing cycle was maintained at 10 hr light off and 14 hr light on. Embryos were kept in embryo medium prepared following The zebrafish book (4th). To keep embryos transparent in early developmental stages, 0.003% PTU was added around 18-24hpf.
For genotyping of adult zebrafish, a small piece of tail was clipped and placed in 50μl 50mM NaOH at 95°C for 30min. For genotyping of embryos, either whole embryos or a small piece of tail was dissolved in 10μl 50mM NaOH at 95°C for 30 min [32]. The lysates were neutralized by adding 1/10 volume of 1M Tris-HCl pH 8.0 and 1 μl of crude lysate was used for PCR genotyping. The PCR fragment of hrg1aiq261 and hrg1biq361 alleles were genotyped with the following primers: for hrg1a, forward: 5’- GAATTATCAAGCTTCACATCACAGGCTCTTTCCGAG -3’, reverse: 5’- AAGCTACACTGCAGCACCGCTGTCTCCAGGTCAAACG -3’; for hrg1b, forward: 5’- actgcataGGATCCCCCTTTAAAGTGTGTTATCATGTG -3’, reverse: 5’- GCagactcctcgagCTTCCTACTACAGGGCCTGAATC -3’.
For PHZ treatment in adult fish, 2.5 μg/ml PHZ was prepared in system fish water. Adult fish was placed in fish water with PHZ for 25 min at 28°C. PHZ was then rinsed off with fresh fish water.
Adult zebrafish (8–12 months old) were anesthetized with 0.02% tricaine (MS-222, Western Chemical Inc.) in fish water. Adult fish were first fixed by 10% Neutral buffered formalin (NBF) after slitting along the ventral abdominal wall to allow the fixative to immerse the gastrointestinal organs. The volume of fixative was at least 10 times of fish volume. The fixed fish were either processed as whole-mount paraffin embedded sections or frozen sections embedded in OCT after dissecting the kidneys, spleens and livers.
Perl’s Prussian blue stain was performed to detect ferric iron zebrafish sections [33]. To perform DAB-enhanced staining, endogenous peroxidase activity was quenched by incubating embryos in 0.3% H2O2 (in methanol) for 20 min at RT. Following 3 times rinse in PBS, sections were incubated in DAB substrate kit (Pierce, Thermo Fisher) for 15 min. Ferric ferrocyanide catalyzes oxidation of DAB, producing a reddish-brown color.
Whole amount in situ hybridization (WISH) was performed following standard protocol as described [34]. To generate probes for hrg1a and hrg1b, 3’UTR regions were amplified using following primers: hrg1a probe: forward, 5’- gcagtcacctcgagACACACAGCAGCACACTAGTGTC -3’, reverse, 5’- gatctaggatccGTCTGAGCGCAGCTGACAGAC -3’; hrg1b probe: forward, 5’- gcagactcctcgagTTGGCTCCTTCAGCTCTAATGG -3’, reverse, 5’- gatctcggatccGACTTAAACTGTATATTATTTCC -3’. The amplified fragments were cloned to pCS2+ vector with BamH1 and Xho1 digestion. Probes were synthesized by using DIG RNA Labeling Kit (SP6/T7) (Roche).
Dissection of various adult zebrafish tissue was performed as previously described [35]. For qRT-PCR experiments, dissected tissue was immediately placed in TRIzol and flash-frozen.
The CRISPR gRNA was designed using Optimized CRISPR Design (http://crispr.mit.edu/). The gRNA target sequences for hrg1a (NM_200006.1) and hrg1b (NM_001002424.2) are listed: hrg1a exon 2: GGTGGATCTGACGACAGGAA TGG; hrg1b exon 3: GGCGGTAGCGGTAGGAGTAC AGG. The gRNA constructs were cloned using pT7-gRNA as backbone [36]. pCS2-Cas9 was used to produce Cas9 capped mRNA by in vitro transcription. Approximately ~300ng Cas9 mRNA and ~100ng gRNA were co-injected to embryos at 1-cell stage. Injected embryos were raised to adulthood as F0 chimeric founders. The founders were subjected to tail-clip genotyping to confirm indels at target sites. The positive chimeric founders were them crossed to WT zebrafish. The F1 embryos with indels at target sites were raised as stable mutant lines.
One microgram of total RNA was used for reverse transcription by iScript cDNA synthesis kit (Bio-Rad). The reaction without reverse transcriptase was used for a negative control. cDNA was diluted 2–5 times for following PCR reaction.qRT-PCR was performed with SsoAdvanced Universal SYBR Green Supermix (Bio-Rad). Each reaction was triplicated to avoid possible random variations.
O-dianisidine staining was performed to detect hemoglobin in RBCs of whole embryos or histological sections as previously described [37]. Heme catalyzes oxidation of o-dianisidine in the presence of H2O2, producing a dark brown color in hemoglobin-positive cells. Briefly, collected embryos or sections were placed in 1 ml staining solution (0.06 (w/v) O-dianisidine, 25% Ethanol, 10mM Sodium Acetate and 0.02% H2O2) for 20 min in dark conditions. Staining was stopped by rinsing with 70% ethanol.
FACS analysis was performed as described [38]. Embryos from globinLCR: GFP transgenic background were pooled. The cells were disaggregated and then sequentially filtered through 70 μm and 33 μm cell strainers. The percentage of GFP-positive cells in transgenic embryos was analyzed by FACSCantos II machine (BD Biosciences).
Dechlorinated zebrafish embryos or dissected adult tissues were disrupted using a Dounce Homogenizer in appropriate volume of homogenization buffer (10mM Tris-HCl, mM EDTA, 1mM PMSF, 1X protease inhibitor cocktail (Roche)). The homogenized solution was then centrifuged at 800 g at 4°C for 5 min. The supernatant was ultra-centrifuged at 100, 000 g at 4°C for 90 min. The supernatant is the cytosolic fractionation and the pellet was treated as crude membrane fraction. The pellet was collected and dissolved in lysis buffer (2% Triton-X100, 150mM NaCl, 50mM Tris-HCl, 20mM HEPES, 1mM PMSF, 1mM EDTA, 1X protease inhibitor cocktail). The membrane lysate was used Western blot experiments.
Polyclonal HRG1 antibody serum was generated in rabbit using the C-terminal 17 amino acid peptide sequence (YAHRYRADFADIILSDF) of human Hrg1 as antigen (Epitomics, Inc.). Since the C-terminal 17 amino acid sequence of human Hrg1 has high homology to zebrafish Hrg1a and Hrg1b (15/17), it cross-reacts with both Hrg1a and Hrg1b.
For western blot analysis of Hrg1 protein in zebrafish, total protein concentration in membrane fractionation lysate was measured using the Pierce BCA assay kit (Thermo Scientific). Equal amount of total protein was mixed with Laemmli sample buffer and were separated on 12% SDS-PAGE and transferred to a 45 μM nitrocellulose membrane with semi-dry transfer apparatus (Bio-Rad). The affinity purified Hrg1 antibody was used at a concentration of 1:1000, goat anti-rabbit HRP-conjugated secondary was used at 1: 30,000, and blots were developed in SuperWest Femto Chemiluminescent Substrate (Thermo Scientific).
Embryos were anesthetized in of 0.02% tricaine in embryos medium, 1% bovine serum albumin (BSA) in calcium- and magnesium-free PBS. The tails of approximately 30 embryos were cut with surgical scissors to allow red blood cells (RBCs) to flow into the tricaine solution. The tricaine solution containing RBCs was loaded into Shandon EZ Single Cytofunnels (Thermal scientific) and concentrated onto a slide (Thermal scientific) by centrifugation at 450 rpm for 3 mins using a Shandon Cytospin 4 cytocentrifuge (Thermal scientific) according to the manufactory’s instructions. Slides were air-dried prior to May-Grunwald Giemsa staining.
May-Grünwald staining solution (May-Grünwald solution (MG500, sigma-aldrich): methanol = 1:3) was gently added onto slides, incubated 5 min at RT and the stain rinsed off with distilled water. Subsequently, the slides were incubated with 1 ml Giemsa staining solution (Giemsa Stain (GS500, Sigma-Aldrich): water = 1:20) for 15 to 30 mins. The Giemsa staining solution was then washed off with distilled water and slides were air-dried before examining under microscope.
The S. cerevisiae strain W303 containing the hem1Δ mutation has been described previously [39, 40]. The mutant yeast cells were maintained at 30°C in yeast peptone dextrose (YPD) media supplemented with 250 μM δ-aminolevulinic acid (ALA) (Frontier Scientific). To generate yeast expression plasmids, the zebrafish hrg1a and hrg1b ORFs were cloned with primers containing BamHI and XbaI restriction sites into the pYES-DEST52 vector (Invitrogen). The mutant alleles of hrg1a and hrg1b were amplified with primers containing BamH1 and XbaI sites (with and without a C-terminal HA tag), and cloned into pYES-DEST52.
The dilution spot assays were performed as described previously [39, 40]. Plasmids containing potential heme transporters were transformed into hem1Δ yeast using the lithium method [41]. To assay aerobic growth exclusively, yeast was induced with 2% galactose in the of ALA, and then spotted onto plates containing indicated heme or ALA concentrations as well as 2% glycerol and 2% lactate as a carbon source. The western blotting experiments were performed as previously reported [19].
The kidneys from 5–6 month old adult zebrafish (Tü, DKO, non-PHZ and PHZ-treated) were dissected and flash-frozen in TRIzol before RNA extraction (To minimize variations from individual fish, tissues from three adult zebrafish were pooled for each of the three biological replicates. 3 fish as a cohort, 3 cohorts per genotype). Total RNA was extracted following TRIzol manual (Invitrogen). The extracted RNA was digested with RNase free-DNase to remove remaining genomic DNA and cleaned up using Qiagen RNA mini column (Qiagen, Germany). Quality and quantity of total RNA were checked by Agilent Bioanalyzer 2100. One microgram of total kidney RNA and 100 ng of total spleen RNA were used for RNA-seq library construction. Purified mRNA was prepared from total RNA following the manufactory’s manual of NEBNext Poly-A) mRNA Magnetic Isolation Module (E7490S, New England Biolabs). RNA-seq libraries was constructed with NEBNext Ultra RNA Library Prep Kit for Illumina (E7530L, New England Biolabs). The RNA-seq libraries and fragment size was roughly qualified by Agilent Bioanalyzer 2100. RNA-seq libraries were quantified using NEBNext Library Quant Kit for Illumina (E7630S, New England Biolabs).
RNA-seq was performed using Illumina’s HiSeq-2500. Total of 24 samples with single-end 50 base reads were sequenced, with triplicate libraries of spleens and kidneys. Bioinformatics quality control was done using FastQC, version 0.11.5. The reads were aligned to zebrafish GRCz10reference genome using STAR, version 2.5.2b. The numbers of reads mapped to genes were counted using HTSEQ, version 0.6.1p1. Finally, differentially expressed genes were identified via DESeq2, version 1.12.3 with the cutoff of 0.05 on False Discovery Rate (FDR). R version 3.3.2 (2016-10-31) was used, and Bioconductor version 3.4 with BioInstaller version 1.24.0 were used. For gene annotation, we used Ensembl GRCz10, release 87. False Discovery Rate (FDR) by Benjamini-Hochberg was used to determine the statistical significance with the cutoff value of 0.05. GO enrichment and network analysis were performed using R package clusterProfiler [42]. All the sequencing data including read counts per gene were deposited to GEO with the accession number of GSE109978.
|
10.1371/journal.pgen.1003677 | Mediator Directs Co-transcriptional Heterochromatin Assembly by RNA Interference-Dependent and -Independent Pathways | Heterochromatin at the pericentromeric repeats in fission yeast is assembled and spread by an RNAi-dependent mechanism, which is coupled with the transcription of non-coding RNA from the repeats by RNA polymerase II. In addition, Rrp6, a component of the nuclear exosome, also contributes to heterochromatin assembly and is coupled with non-coding RNA transcription. The multi-subunit complex Mediator, which directs initiation of RNA polymerase II-dependent transcription, has recently been suggested to function after initiation in processes such as elongation of transcription and splicing. However, the role of Mediator in the regulation of chromatin structure is not well understood. We investigated the role of Mediator in pericentromeric heterochromatin formation and found that deletion of specific subunits of the head domain of Mediator compromised heterochromatin structure. The Mediator head domain was required for Rrp6-dependent heterochromatin nucleation at the pericentromere and for RNAi-dependent spreading of heterochromatin into the neighboring region. In the latter process, Mediator appeared to contribute to efficient processing of siRNA from transcribed non-coding RNA, which was required for efficient spreading of heterochromatin. Furthermore, the head domain directed efficient transcription in heterochromatin. These results reveal a pivotal role for Mediator in multiple steps of transcription-coupled formation of pericentromeric heterochromatin. This observation further extends the role of Mediator to co-transcriptional chromatin regulation.
| DNA is packaged into chromatin structure, which is important for various genome functions such as gene expression and maintenance of genetic information. Heterochromatin is a condensed chromatin structure and involved in epigenetic regulation of gene expression through repression of transcription. Heterochromatin at the pericentromeric repeats in fission yeast is assembled by two distinct mechanisms, RNAi-dependent and Rrp6, a component of the nuclear exosome, -dependent mechanisms. In addition, heterochromatin spreads into neighboring regions in an RNAi-dependent manner. Both mechanisms are coupled with the transcription from the target loci by RNA polymerase II, but the molecular nature of the coupling is not understood. Here we showed that the multi-subunit complex Mediator, which directs initiation of RNA polymerase II-dependent transcription, functions in the coupling between transcription and heterochromatin assembly. Mediator is required for Rrp6-dependent heterochromatin assembly and contributes to the RNAi-dependent spreading of heterochromatin via enhancement of production of siRNA by RNAi machinery. These observations highlight the multi-functions of Mediator in the transcription-coupled processes.
| Heterochromatin is a silent higher-order chromatin structure that is associated with various genome functions such as transcriptional regulation, chromosomal segregation, suppression of recombination and repression of selfish elements. The fission yeast Schizosaccharomyces pombe provides a good model system for investigating heterochromatin formation. In fission yeast, heterochromatin is preferentially enriched across large chromosomal domains at the pericentromeres, subtelomeres and the mating-type locus. These regions are rich in methylation of histone H3K9 (H3K9me), which is catalyzed by the histone methyltransferase Clr4, a homolog of mammalian SUV39h [1], [2]. The modification of H3K9me is critical for the binding of HP1 proteins [2], [3], which recruit various factors for the assembly of repressive chromatin and associated various functions [4], [5].
Several distinct pathways promote heterochromatin assembly in fission yeast. At the pericentromere, RNAi machinery plays essential roles in heterochromatin formation [6], [7]. Pericentromeric heterochromatin is assembled on the outer repeat (otr) region (containing of dg and dh repeats), and the outer portion of the innermost repeats (imr), which surround the central core (cnt) domain, the site of kinetochore assembly [8]. The repeats are transcribed by RNA polymerase II (RNAPII) to produce non-coding RNAs (ncRNAs) during S-phase [9], [10], [11]. Transcribed ncRNAs give rise to double-strand RNA via the RNA-dependent RNA polymerase complex (RDRC), comprised of Rdp1, Cid12 and Hrr1, and are processed into small interfering RNAs (siRNAs) by the RNase III helicase Dicer (Dcr1). The siRNAs are then loaded into an RNA-induced transcriptional silencing (RITS) complex composed of Ago1, Tas3 and Chp1 [7]. siRNAs target the RITS complex to cognate nascent transcripts, resulting in the recruitment of additional factors, including RDRC and ultimately Clr4, to methylate histone H3K9. Generation of siRNAs and heterochromatin assembly are interdependent processes that form a self-enforcing loop [12], [13]. Importantly, RNAPII appears to couple transcription at the target loci with the generation of siRNAs. This was shown by the fact that a specific mutation in RNAPII results in a decrease in heterochromatic histone modifications, accumulation of pericentromeric transcripts, and accompanying loss of siRNAs, which are effects that were observed previously in RNAi mutants [10].
Heterochromatin, once established, spreads into neighboring region, which is typically shown by the heterochromatin formation and silencing of the genes inserted into heterochromatin. This process depends on RNAi system and probably couples with transcription [14], [15].
Nuclear RNA is monitored by a nuclear RNA surveillance system involving exosomes with 3′-5′ exonuclease activity, and a portion of the ncRNA at the pericentromere has been shown to be degraded by the nuclear exosome [16], [17]. In addition to RNA degradation, 3′-5′ exonuclease Rrp6, a component of the nuclear exosome, was shown to mediate heterochromatin formation in parallel with RNAi, which is demonstrated by the cumulative increase and decrease of H3K9me at the pericentromere in the double null-mutant of ago1 and rrp6 [18]. Since the amount of siRNA is not affected by depletion of Rrp6, Rrp6-dependent heterochromatin formation occurs via a pathway that is distinct from that of RNAi-dependent siRNA generation [16]. The molecular basis of the Rrp6-dependent pathway is not yet clear.
The cenH sequence, which shows 96% homology to centromeric dg and dh repeats, is present at the silent mating-type (mat2/3) locus and serves as an RNAi-dependent heterochromatin nucleation center [6], [19]. In parallel with the RNAi-dependent pathway, the ATF/CREB family DNA-binding proteins, Atf1 and Pcr1, participate in heterochromatin nucleation with a histone deacetylase, Clr3 [20].
Mediator, which is a well-conserved protein complex consisting of at least 20 subunits, was first identified as a factor that mediates DNA transcription factors binding at regulatory sequneces and RNAPII at promoters for the efficient start of transcription [21], [22] and has been shown to be required for transcription of almost all protein-coding genes in vivo [23], [24], [25]. Structural analysis indicates that this complex consists of four distinct structural domains: head, middle, tail and kinase. The head domain is responsible for extensive interaction with RNAPII, and the Med18/Pmc6-Med20 heterodimer, which is a portion of the head domain, binds to the core head domain through the C-terminal helix of Med8 [26]. The head domain stabilizes the connection between RNAPII and TFIIH, which facilitates the transition from initiation complex to elongation complex [27]. In addition to the promotion of general transcription from protein-coding genes, recent studies have revealed a new function of Mediator. In Arabidopsis thaliana, Mediator directs the transcription of ncRNA genes by recruiting RNAPII to their promoters [28]. In mammalian cells, a specific subunit of Mediator functions as an interaction site for alternative mRNA splicing or transcription elongation factors [29], [30]. These data suggest that Mediator might play roles in both transcription elongation and the subsequent processing of transcripts as a platform for the recruitment of various factors.
Since both Rrp6-dependent heterochromatin formation and RNAi-dependent heterochromatin formation are coupled with transcription, we assumed that the factor(s) that interacts with RNAPII directs the coupling. Therefore, we assessed the role of several RNAPII-interacting factors in pericentromeric heterochromatin assembly. We found that the disruption of Med18 and Med20, non-essential subunits of the Mediator head domain (MHD), compromised both RNAi-dependent and Rrp6-dependent heterochromatin assembly at the pericentromere. In addition, the head domain is required for transcriptional activation in heterochromatin. Therefore, we propose that Mediator links transcription of ncRNA and its processing by RNAi and exosomes for the formation of centromeric heterochromatin.
To investigate whether Mediator is involved in heterochromatin assembly, each gene encoding a non-essential subunit of mediator was disrupted in a strain possessing marker genes in the pericentromeric heterochromatin (otr1R::ade6+ and imr1L::ura4+) to monitor heterochromatic silencing (Figure 1A) [31]. Since the otr1R::ade6+ and imr1L::ura4+ genes are repressed by heterochromatin, the wild-type strain formed red colonies on a solid medium containing a limiting amount of adenine (Low Ade) and was resistant to 5-fluoroorotic acid (5-FOA), a counter-selective drug for ura4+ expression. By contrast, heterochromatin mutants, such as clr4Δ, formed white or pink colonies on the Low Ade plate and showed sensitivity to 5-FOA (Figure 1B). Among the eight non-essential subunits of mediator tested (med1/pmc2, med27/pmc3, med18/pmc6, med20, med19/rox3, med12/srb8, med13/srb9 and cdk8/srb10), only disruption of med18 (also known as pmc6) and med20 resulted in the formation of pink colonies and increased sensitivity to 5-FOA (Figure 1B), which is in accordance with growth on plates lacking uracil or adenine (Figure S1A). Closer examination revealed that both med18Δ med20Δ cells formed a mixture of white and pink colonies on Low Ade plates. In addition, point mutants of med8 (med8-K9) and med31 (med31-H1), which were isolated by the screening of heterochromatic mutants (Figure S1B; Kato et al. submitted), also formed a mixture of pink and white colonies (Figure 1C, Figure S1C).
The variegation in the color of colonies by the mutation of Mediator subunits suggested that silencing of the otr1R::ade6+ gene was variegated in the mutant cells and that distinct levels of otr1R::ade6+ silencing were epigenetically inherited. To test the stability of the pink and white phenotype, pink and white colonies of each mutant were selected, cultured in YES media overnight and re-spotted onto Low Ade and 5-FOA plates. Re-spotting of the cells from white colonies (med18Δ-w, med20Δ-w and med8-K9-w) and pink colonies (med18Δ-p med20Δ-p and med8-K9-p) produced predominantly white colonies and pink colonies, respectively (Figure 1D, Figure S1C). This indicated that the white and pink phenotypes were epigenetically inherited through generation but exchangeable, which is further confirmed by the measurements of the conversion rates between white and pink epiclones (Figure S2). The conversion rates are different in each mutants, but in all mutants, conversion rates from pink to white is higher than those of white to pink, showing that white-epiclones, in which heterochromatin is compromised, are more stable. Hereafter, we designate the epigenetic clones derived from white colonies and red colonies as white and pink “epiclones”, respectively. med18Δ-w and med20Δ-w showed greater sensitivity to 5-FOA than med18Δ-p and med20Δ-p (Figure 1D), indicating that silencing at imr1L::ura4+ was also compromised more severely in the white epiclones and that the silencing defect at otr1R::ade6+ is connected with that at imr1L::ura4+. This suggested that the white phenotype reflected silencing defects of the entire pericentromeric heterochromatin. It should be noted that it was difficult to separate the white epiclones from the pink epiclones of med31-H1 cells (med31-H1-w and med31-H1-p in Figure S1C) because of frequent variegation between the two (Figure S2).
The loss of heterochromatic gene silencing was confirmed in Mediator mutants by measuring the accumulation of transcripts from the pericentromeric repeats (dg and dh) and inserted marker genes. Strand-specific northern analysis showed a large increase in those transcripts in both med18Δ- and med20Δ-w cells (Figure 1E and Figure S3), which was consistent with the observed silencing defects (Figure 1D), while only marginal accumulation was observed in med18Δ-p and med20Δ-p cells. Both point mutants of the other Mediator subunits (med8-K9 and med31-H1) also showed accumulation of the transcripts (Figure S3). Accumulation of heterochromatic transcripts from dh repeats was also demonstrated by strand-specific RT-PCR (Figure 1F). These results showed that the Mediator subunits Med8, Med18, Med20 and Med31 were involved in silencing of pericentromeric heterochromatin. Med18 and Med20 form a heterodimer that associates with head domain core complex through the C-terminal linker region of Med8 [26], and the connection between the Med18/Med20 heterodimer and head domain appears to be lost in the med8-K9 mutant (Figure S1B). In addition, Med31, a component of the middle domain, is located close to the head domain. Because these findings indicate that Mediator functions in pericentromeric heterochromatin via the head domain, Med18 and Med20 were selected for closer examination.
Both RNAi- and Rrp6-dependent heterochromatin formation, which occur in parallel at the pericentromere, appear to be coupled with the transcription of ncRNA at the pericentromeric repeats [10], [18]. We, hence, assume that Mediator contributed to pericentromeric heterochromatin formation directly through the transcription of ncRNA and/or processing of ncRNA. If this assumption was true, Mediator should localize to the transcribed region in heterochromatic repeats. To test this possibility, the localization of Med20-5Flag and RNAPII to the transcribed regions of dh repeats was examined by Chromatin immunoprecipitation (ChIP) assay. Since heterochromatic ncRNA is mainly transcribed during G1/S-phase [11], cell cycle was synchronized using the cdc25 temperature-sensitive mutation. The results show that RNAPII accumulated during G1 to early S-phase, followed by the accumulation of transcripts (, C, D). Med20-5Flag showed a similar oscillating pattern, but the peak disappeared slightly earlier than the Pol2 peak (Figure S4B). This is consistent with the speculation that Mediator is involved in the heterochromatic ncRNA transcription.
To gain further insight into the roles of Mediator during heterochromatin organization, the occupancy of H3K9me2, Swi6 and RNAPII at the centromeric heterochromatin was assessed by ChIP assay. At the inserted marker gene (otr1R::ade6+), the levels of histone H3K9me and Swi6 were decreased and RNAPII occupancy was increased in the white epiclones of med18Δ and med20Δ (Figure 2A–C). This indicated that heterochromatin structure at the marker gene was disrupted in the white epiclones. By contrast, in the pink epiclones of med18Δ and med20Δ, the decrease in H3K9me/Swi6 and increase in RNAPII were less prominent than those in white epiclones. This reflected the difference in silencing defects in each epiclone (Figure 1D). At the heterochromatic repeats, dh, H3K9me/Swi6 and RNAPII were also decreased and increased in the Mediator mutants, respectively, but the differences between the white and pink epiclones were less prominent than at otr1R::ade6+. These results showed that the accumulation of transcripts from heterochromatic repeats and marker genes is, at least in part, due to an increase in transcription induced by the disruption of heterochromatin structure. These results also confirm that Mediator is required for heterochromatin formation at the pericentromere.
There are two distinct pathways for heterochromatin formation at the pericentromeric repeats: RNAi-dependent and Rrp6-dependent pathways. In RNAi mutants such as dcr1Δ, H3K9me is diminished at the inserted marker genes but substantially retained at the pericentromeric repeats, while disruption of rrp6 did not affect H3K9 me at the marker genes [18]. The distribution of H3K9me in the white epiclones of the Mediator mutants resembled that observed in dcr1Δ cells; the level of H3K9me at the marker genes was lower than that at heterochromatic repeats. Thus, We speculated that Mediator is involved in the RNAi-dependent pathway. To confirm this, a med18Δ dcr1Δ double mutant was established and used to examine heterochromatin silencing and the amount of H3K9me and Swi6 at dh repeats and imr1:: ura4+ (Figure 3A, B). Note that since med18Δ Δdcr1Δ cells did not exhibit the variegated phenotype observed in the med18Δ single mutant (Figure 3A), white and pink epiclones of med18Δ cells were not separated in the following experiments (Figure 1B, C and D). If Mediator functions in the RNAi-dependent pathway, the med18Δ dcr1Δ double mutants would retain H3K9me and Swi6 to the level similar to those in single mutant. However, in the double mutant, the retained H3K9me/Swi6 at the dh repeats was significantly decreased compared to each single mutant (Figure 3B), suggesting that, contrary to our speculation, Med18 functions in a pathway distinct from the RNAi-dependent pathway.
Because the results in Figure 3B were reminiscent of the results reported by Reyes-Trucu et al., in which the amount of H3K9me retained at the centromeric repeats in ago1Δ cells was significantly decreased by further disruption of rrp6 [18], we speculated that Mediator functioned in an Rrp6-dependent heterochromatin formation pathway. To confirm this, single, double and triple mutant cells of dcr1, med18 and rrp6 were used to measure the amount of H3K9me at the dh repeats (Figure 3C, left panel). Each single mutant, as well as the med18Δ rrp6Δ double mutants, retained similar amounts of H3K9me. By contrast, combination of dcr1Δ with rrp6Δ caused a substantial decrease in H3K9me, which was consistent with the previous proposal that both RNAi- and Rrp6-dependent pathways contribute to heterochromatin formation at the pericentromeres [18]. Similarly, the combination of dcr1Δ with med18Δ also caused a significant decrease in H3K9me, while med18Δ rrp6Δ cells maintained a level of H3K9me comparable to each single disruptant. The H3K9me retained in med18Δ rrp6Δ cells was also decreased by the introduction of dcr1Δ. The amount of Swi6 in each mutant reflects the amount of H3K9me (Figure 3C, left panel). These results clearly indicate that Med18 functions in the same heterochromatin formation pathway as Rrp6 at the dh repeats. Therefore, H3K9me in dcr1Δ cells was retained by the Rrp6/Med18-dependent pathway, while H3K9me in each of the med18Δ and rrp6Δ cells was maintained by the RNAi-dependent pathway.
Details of the Rrp6-dependent pathway are not clear yet; even the localization of Rrp6 at heterochromatin has not been examined. We, thereby, analyzed the localization of Rrp6 tagged with myc epitope at heterochromatin (dh) as well as euchromatin (act1 and fbp1) (Figure. 3D). Compered with no-tag control, Rrp6-myc was enriched at both dh and euchromatic genes to the same extent. Depletion of clr4 did not affect the localization of Rrp6-myc, while deletion of dcr1 caused a slight increase at all loci. The enrichment of Rrp6 at dh increased in med18Δ-w and med18Δ-p epiclones, and also in the med18Δ dcr1Δ double mutant, while the enrichment at euchromatic genes was marginally affected in those mutant cells. This suggests that Mediator functions in a step after association of Rrp6 on chromatin for heterochromatin formation.
Both dcr1Δ and rrp6Δ cells retained similar levels of H3K9me at centromeric repeats (Figure 3C). By contrast, deletion of rrp6 does not affect H3K9me at the marker genes inserted in centromeric repeats [18], whereas deletion of dcr1 caused the loss of H3K9me on the marker genes (Figure 3B), indicating that the spreading of H3K9me into the inserted marker genes occurs via an RNAi-dependent mechanism [14], [15]. While H3K9me at otr1R::ade6+ was severely decreased in the white epiclones of med18Δ and med20Δ, it was substantially retained in the pink epiclones (Figure 2A), indicating that the spreading of H3K9me in the marker genes was variegated in the Mediator mutants. In addition, introduction of, dcr1Δ to med18Δ cells abolished a variegated phenotype (Figure 3A). These data confirm that the loss of Med18/Mediator results in the variegation of RNAi-dependent heterochromatin spreading. In other words, Med18/Mediator is also involved in the RNAi-dependent heterochromatin pathway.
To examine the involvement of Mediator in RNAi, siRNA derived from dg and dh repeats was analyzed in the Mediator mutants. siRNAs corresponding to the pericentromeric repeats were not detected in dcr1Δ cells (Figure 4A). In the white epiclones of the Mediator mutants, a marginal amount of siRNA from the dg and dh repeats was detected (Figure 4A and Figure S5A, B). The marginal amount of siRNA was diminished by introduction of dcr1Δ (Figure S5B), showing that the siRNA observed in med18Δ cells are produced through RNAi pathway. In the pink epiclones, reduced but significant amounts of siRNAs (approximately 10–50% of that of wild-type cells) were detectable. Note that the amount was varied because of the state of variegation. Since the structure of heterochromatin (H3K9me and Swi6) at the repeats was substantially maintained in both white epiclones (Figure 2A, B) and the maintenance was dependent upon RNAi-pathway as shown above (Figure 3C), the small amount of siRNA synthesized in the white epiclones appears to be sufficient to maintain heterochromatin structure at the repeats. A similar reduction of siRNA was observed in the white and pink epiclones of med8-K9 cells and med31-H1 cells (Figure S5). Note that when siRNA derived from dg and dh repeats was analyzed separately, each siRNA was found to be reduced in med18Δ cells (Figure S5). These data indicate that Mediator is required for siRNA generation at pericentromeric heterochromatin and the defect of the MHD causes variegation of the spreading of H3K9me into the marker genes.
Since RNAi machinery localizes on heterochromatin for processing of ncRNA into siRNA [3], [6], [12], [32], the requirement of Med18 for the recruitment of RNAi factors to heterochromatin was investigated. Binding of the components of the RITS complex (3Flag-Ago1 and Chp1-13myc) and of RDRC (Rdp1-5Flag) to pericentromeric repeats was examined by ChIP assay (Figure 4 B–D). As reported, 3Flag-Ago1 bound to dh repeats in a heterochromatin- and/or RNAi-dependent manner [3], [12], [32], as evidenced by the finding that the binding of Ago1 was reduced to a level comparable to that of the no-Flag-tag control in clr4Δ and dcr1Δ cells (Figure 4B left panel). By contrast, a substantial amount of 3Flag-Ago1 was retained in med18Δ cells that formed a mixture of pink and white epiclones (Figure 4B left panel). Binding of Chp1-13myc to dh repeats was abolished by deletion of clr4, while reduced but significant Chp1-13myc localization was observed at the dh repeats in dcr1Δ cells, representing the binding of the chromo-domain of Chp1 to H3K9me that was retained at the pericentromeric repeats in these cells (Figure 4C, right panel). By contrast, the binding of Chp1-13myc was not affected by the deletion of med18. Even in white epiclones, in which H3K9me is reduced to the same level as in dcr1Δ cells (Figure 4C, right panel), Chp1-13myc binds to dh repeats at the same level as in wild-type cells (Figure 4C, left panel). Association of Rdp1-5Flag in each mutant was similar to that of 3Flag-Ago1 in that it was almost abolished in clr4Δ and dcr1Δ cells, but significantly retained in both med18Δ-w and med18Δ-p cells (Figure 4D, left panel). All together, the RITS complex and RDRC associated with heterochromatin even in med18Δ-w cells, probably because the small amount of siRNA synthesized in med18Δ cells was sufficient for the association of RITS with heterochromatin. Together with the data on the accumulation of ncRNA and reduction of siRNA in Mediator mutants, these data indicate that Mediator is not required for the association of the RITS complex and RDRC to heterochromatin but is required for efficient siRNA production from the ncRNA by heterochromatin-bound RNAi machinery.
It has been previously reported that the tethering RITS to ura4 RNA induces RNAi- and heterochromatin-dependent gene silencing of the ura4 gene, indicating that binding of the RITS complex to ncRNA is a key step in the RNAi-directed formation of heterochromatin by inducing H3K9 methylation and conversion of ncRNA to siRNA [33]. Tethering of RITS is achieved by the fusion of Tas3, a subunit of the RITS complex, to the λN protein, which binds to the 5BoxB sequence inserted at the 3′ UTR region of ura4 RNA (Figure 4E). To determine whether Med18 or Med20 is required for Tas3-λN-induced silencing of the ura4-5boxB gene, the effect of the deletion of these subunits on silencing induced by artificial tethering of the RITS complex was examined. Disruption of med18 or med20 resulted in the loss of ura4-5BoxB silencing (Figure 4F), similar to the effect of clr4 disruption. This result showed that Med18 and Med20 are required for Tas3-λN-induced silencing of the ura4-5BoxB locus and that Mediator plays a role in the step following the binding of the RITS complex to target RNA.
Since Mediator regulates general transcription in euchromatin, it is possible that it also regulates the transcription of heterochromatic non-coding RNA. Indeed, recent reports suggest a negative role of MHD subunits (Med18 and Med20) in heterochromatic transcription, based on the observation of an increase in the transcription of pericentromeric ncRNA in Mediator mutants [34], [35]. However, it is difficult to state conclusively whether the observed increase is due to the direct effects of the absence of Mediator because it is also possible that the deletion of Mediator subunits causes disruption of the heterochromatin, which secondarily induces an increase in transcription. To avoid this dilemma, heterochromatin at the mating-type locus was selected for examination (Figure 5A) because the RNAi-dependent pathway is dispensable for the maintenance of heterochromatin here due to the existence of another pathway mediated by the DNA-binding proteins Atf1and Pcr1 [20]. Thus, mutation of Mediator would not be expected to affect the mating-type locus heterochromatin, making it possible to directly measure the effect of the mutation on transcription activity in heterochromatin.
First, a ChIP assay was performed for H3K9me and Swi6 to examine heterochromatin structure at the mating-type locus in various mutants using specific primers for the cenH sequence (Figure 5B). As expected, high levels of H3K9me2 and Swi6 were maintained at the cenH sequence at the mating-type locus and the inserted ura4+ gene (kint2::ura4+) in dcr1Δ and dcr1Δ med18Δ mutants. In med18Δ cells, the level of H3K9me was decreased to half of that of wild-type cells for an unknown reason, but the level of Swi6, which is essential for transcriptional gene silencing in heterochromatin, was maintained, consistent with the observation that silencing at this locus was not affected by med18Δ (Figure 5B). Hence, the Atf1/Pcr1-dependent pathway probably retains heterochromatin structure and silencing without Med18 function.
Next, the effect of deletion of dcr1 and med18 on the silencing of kint2::ura4+ was examined (Figure 5C). While the wild-type strain was able to grow on a 5-FOA-containing plate but not on a uracil-lacking (-Ura) plate, the clr4Δ strain was hypersensitive to 5-FOA but grew well on an –Ura plate (Figure 5C), showing that kint2::ura4+ was silenced and expressed, respectively, in each strain. By contrast, dcr1Δ cells, like the wild-type cells, hardly grew on 5-FOA containing media, while some cells were able to grow on an –Ura plate, suggesting that silencing is only weakly compromised in dcr1Δ cells. However, med18Δ cells showed a phenotype similar to that of wild-type cells, showing no silencing defect at the mating-type locus. Introduction of med18Δ to dcr1Δ cells suppressed the silencing defect detected on the –Ura plate. RT-PCR analysis of transcripts from cenH and kint2::ura4+ was consistent with the silencing assay; more than 100-fold, approximately 40-fold, and approximately 10-fold accumulation of transcripts from cenH and kint2::ura4 were observed in clr4Δ, dcr1Δ and med18Δ cells, respectively (Figure 5D). In addition, introduction of med18Δ into dcr1Δ cells caused a decrease in transcripts to a level similar to that of med18Δ cells.
The accumulation of RNA in dcr1Δ cells and med18Δ cells could be explained by defects in RNA degradation by RNAi and/or the exosome (Noma et al., 2004, Buhler et al., 2007), or by an increase in heterochromatic transcription. To examine the latter possibility, localization of RNAPII at cenH and kint2::ura4+ was examined by ChIP assay (Figure 5E). Unexpectedly, RNAPII was significantly increased in dcr1Δ cells at both loci in spite of the maintenance of heterochromatin in this strain, suggesting that Dcr1 negatively regulates heterochromatin transcription. Note that RNAi machinery has been shown to interact with RNAPII and modulate transcription in other organisms [36], [37]. Thus, an increase in transcription and prevention of processing of RNA into siRNA could cause the observed accumulation of transcripts in dcr1Δ cells (Figure 5D). By contrast, the level of RNAPII in med18Δ cells was comparable to that of wild-type cells (Figure 5E). Similar results of RNAPII localization were obtained with ChIP assay using the antibody against RNAPII-C-terminal repeats phosphorylated at the second serine, which represents elongating RNAPII (Figure S6). These indicated that Med18 does not repress transcription in heterochromatin. The approximately 10-fold accumulation of cenH and kint2::ura4+ RNA observed in med18Δ cells might be due to a defect in exosome-dependent degradation of RNA [16], which would indicate that significant transcription took place in the absence of Med18. Importantly, introduction of med18Δ into dcr1Δ cells caused a decrease in RNAPII to the level of wild-type cells, suggesting that Mediator is required for efficient transcription in heterochromatin in dcr1Δ cells.
To analyze the role of Med18 on the transcription in the absence of heterochromatin, we compared RNAPII occupancy at centromeric repeats of dcr1Δ rrp6Δ cells with those of dcr1Δ rrp6Δ med18Δcells (Figure S7). Note that both strains showed similarly low levels of H3K9me at dh repeats (Figure 3C). Introduction of med18Δ caused the moderate increase of RNAPII. This suggested that Med18 negatively regulates transcription in the compromised heterochromatin.
From these data, we suggest that in the fully assembled heterochromatin, Med18/Mediator does not negatively regulate pericentromeric transcription; rather, it might be required for efficient transcription in heterochromatin.
The effect of med18Δ and med20Δ on euchromatic gene expression was further examined using microarray. Pink and white epiclones of med18Δ and med20Δ cells were separated and the expression pattern of each epiclone was compared. Analysis of the genes that showed ≥1.5-fold increase (Up) or decrease (Down) in expression between the epiclones revealed that a common set of genes were affected in both med18Δ cells and med20Δ cells, irrespective of the state of heterochromatic silencing (Figure 6A). Therefore, a clear difference between the white and pink epiclones was observed at pericentromeric ncRNA expression. Indeed, in the white epiclones, no euchromatic genes showed stronger induction than centromeric ncRNA; the most strongly increased euchromatic gene showed an approximately 14-fold and 26-fold increase in med18Δ-w and med20Δ-w cells, respectively, which is much weaker than the increase in centromeric ncRNA (which increased by more than 100-fold). Comparison of med18Δ-w and med20Δ-w, or med18Δ-p and med20Δ-p, showed that both subunits shared a common set of targets (Figure S8). This is consistent with the fact that both Med18 and Med20 formes heterodimers submodule in the MHD.
When the expression pattern in the white epiclones of the Mediator mutants (med18Δ-w and med20Δ-w) was compared with that in dcr1Δ cells, it was evident that the expression of a common set of genes was upregulated (Figure 6B, upper panels). Similar sharing of target genes was observed between the pink epiclones of the Mediator mutants (med18Δ-p and med20Δ-p) and dcr1Δ cells (Figure 6B, lower panel). Interestingly, gene ontology analysis showed significant enrichment of terms pertaining to stress responses in the shared target genes. For example, the top GO terms included “cellular response to stimulus” (GO: 0033554, med18Δ-w vs. dcr1Δ P = 5.35×10−3, med18Δ-p vs. dcr1Δ P = 2.2×10−6, med20Δ-w vs. dcr1Δ P = 2.19×10−6, med20Δ-p vs. dcr1Δ P = 1.27×10−4). These results suggest that some euchromatic genes, including stress response genes, are repressed by the RNAi/Mediator system, which may function via a mechanism partly similar to that of RNAi-mediated heterochromatin.
In this study, we showed that the specific subunits of Mediator, Med18, Med20, Med8 and Med31 were involved in pericentromeric heterochromatin formation. The Med18-Med20 heterodimer is a component of the head domain of Mediator [26]. Because the med8-K9 mutation causes truncation of the C-terminal domain that interacts with the Med18/Med20 heterodimer (Figure S1B), it resulted in the loss of the heterodimer. Importantly, Med31 belongs to the middle domain but is located close to the head domain [38], suggesting that the head domain does not function alone in heterochromatin formation, but rather as a part of Mediator. Therefore, we suggest that Mediator specifically plays multiple roles in the formation of pericentromeric heterochromatin via the MHD (Fig. 7).
Two distinct mechanisms, the RNAi-dependent and Rrp6-dependent pathways, function in heterochromatin formation at pericentromeric repeats, while the spreading of H3K9me onto marker genes mainly depends on the RNAi pathway [14], [15]. We found that at the pericentromeric heterochromatin, in the absence of MHD, the Rrp6-dependent pathway is compromised and H3K9me is largely maintained by the RNAi-dependent pathway (Fig. 3). The finding that the amount of siRNA produced in the white epiclones of MHD mutants decreased to 3–20% of that in wild-type cells (Fig. 4A and Figure S5) indicates that only a small amount of siRNA is necessary to maintain heterochromatin at the pericentromeric repeats. The remaining H3K9me at the repeats in MHD mutants spreads onto the marker genes by an RNAi-dependent mechanism. This process was also compromised by the decrease of siRNA caused by the absence of MHD, resulting in variegation of the level of H3K9me at the marker genes, which ultimately caused the appearance of white and pink epiclones. The spreading process appears to require more efficient siRNA production than the maintenance of heterochromatin at the pericentromeric repeats because med18Δ-p cells that produced more siRNA showed more efficient spreading of H3K9me and silencing of marker genes than med18Δ-w cells. In contrast, dcr1Δ cells did not show the variegation of silencing because of loss of siRNA production. It is noteworthy that the variegated phenotype was metastable, which suggests that once heterochromatin was spread into the marker gene, it could be maintained in an MHD-independent manner, probably through the small amount of siRNA produced in MHD mutants.
Many processes involved in RNA processing, such as RNA splicing and RNA transport, are coupled to transcription by RNAPII. siRNA production is also coupled with RNAPII-dependent transcription [10]. Our results showed that the mutation of MHD resulted in a large decrease in siRNA. By contrast, rrp6Δ cells, which have levels of H3K9me and Swi6 at the pericentromeric repeats similar to med18Δ cells (Fig. 3), produced the same amount of siRNA as wild-type cells [16]. These results indicate that MHD is somehow involved in siRNA production after transcription of ncRNA. MHD is localized at the transcribed region in pericentromeric repeats (Figure S4; [34], [35] and is required for transcription in heterochromatin, suggesting that it directly functions in the coupling of transcription of heterochromatic ncRNA by RNAPII and processing of the siRNA by RNAi machinery. Retention of RNAi factors at the pericentromeric repeats in med18Δ cells and the requirement for MHD in heterochromatin formation via the artificial tethering of the RITS complex to RNA suggest that MHD functions after RITS associates with heterochromatic repeats and/or target RNA. Recently, we showed that RNAi factors are assembled into an siRNA amplification compartment that includes transcriptionally active heterochromatin [39]. Thus, MHD might be involved in the formation of this compartment. Although we were not able to detect a stable interaction between MHD components and RNAi factors, such as Ago1, by co-immunoprecipitation experiments (data not shown), it is still possible that MHD recruits factors required for siRNA generation to transcriptionally active heterochromatin through direct or indirect interactions. In any case, further experiments are necessary to clarify the molecular function of Mediator in the RNAi pathway.
In contrast to the RNAi pathway, little is known about the Rrp6-dependent heterochromatin formation pathway. As deletion of rrp6 marginally affects H3K9me and silencing at the inserted marker genes [16], [18], the Rrp6-dependent pathway mainly functions at the pericentromeric repeats. Our genetic experiments showed that rrp6 and med18 were epistatic in the formation of pericentromeric heterochromatin (Fig. 3), indicating that MHD functions in Rrp6-dependent heterochromatin formation. Rrp6 is an exonuclease that is a subunit of the nuclear exosome involved in RNA-quality control [40]. A functional relationship between Mediator and the nuclear exosome has not been reported. We found that Rrp6 associates with both heterochromatin (dh) as well as euchromatin (act1 and fbp1) and deletion of med18 did not affect the localization, suggesting Mediator acts in a step after the association of Rrp6/exsosome with chromatin. This is analogous to the function of Mediator in the RNAi-dependent pathway; MHD functions after recruitment of RITS complex and RDRC to chromatin. Considering the co-transcriptional nature of RNA-quality control [41], [42] and recruitment of RNA-splicing factors to transcripts by Mediator [29], we speculate that MHD plays a role in the co-transcriptional function of chromatin-associated Rrp6 and/or other co-factors to promote heterochromatin formation. Alternatively, given that the RNAi-independent heterochromatin nucleation pathway and Mediator functionally interact with RNAPII processivity factors [18], [30], Mediator may promote Rrp6-dependent heterochromatin formation by affecting elongation by RNAPII through interaction with these processivity factors. It is also possible that the same mechanism is also involved in RNAi-dependent heterochromatin formation through MHD.
Recently, two reports showed that MHD was important for heterochromatin formation at pericentromeres [34], [35]. However, there are several discrepancies between their data and ours. Firstly, the decrease in H3K9me and Swi6 in med20Δ cells was much more severe than that in ours. Secondly, Carlsten et al. claimed that siRNA from the dh repeat in med20Δ cells was diminished but that siRNA from the dg repeats was comparable to that in wild-type cells. Thirdly, both papers assert that Mediator negatively regulates heterochromatic transcription. The first two discrepancies could be caused by the variegated phenotype of MHD mutants. If this variegated phenotype was overlooked or disregarded, the results would be affected by which epiclones were used in the experiments. In addition, since the amount of H3K9me/Swi6 varies depending on the position in the repeats, the discrepancies between their results and ours might reflect a difference in the sites used for ChIP analysis. The third discrepancy could be explained by the use of pericentromeric transcription for their analysis. As described in the Results section, it is hard to argue definitively for the direct influence of MHD mutants on transcription in pericentromeric heterochromatin because it is difficult to determine whether the observed increase is due to the direct effect of depletion of MHD or a secondary effect resulting from the disruption of heterochromatin. Our data using the mating-type locus heterochromatin showed that disruption of Mediator did not cause increased transcription in heterochromatin, rather it caused a decrease in transcription enhanced by deletion of dcr1 (Fig. 6). Interestingly, when heterochromatin was compromised, Mediator appears to negatively regulate transcription, which might also explain the discrepancy.
Recently, Dcr1 was shown to repress a set of genes, including stress response genes, through the degradation of target RNA [43]. We identified a similar set of euchromatic genes that were up-regulated in med20Δ and dcr1Δ cells, suggesting that MHD functions in co-transcriptional degradation of euchromatic RNAs in collaboration with Dcr1. Note that previous transcriptome analysis of the mediator mutants also showed that a similar set of genes were up regulated in med18Δ and med20Δ cells but not in med12Δ cells, supporting the collaborative function of MHD and Dcr1 [44]. In addition, Rrp6-dependent heterochromatin formation was observed at several meiotic genes [45]. Moreover, the exosome and RNAi are shown to regulate a set of genes, including retrotransposons and developmental genes [46]. Therefore, it is also possible that Mediator functions at these loci to silence genes by regulating both RNAi and exosomal machineries.
Emerging evidence shows that Mediator works as a platform for various factors that function in transcription and RNA processing, using a distinct subunit for particular interactions with the factors [29], [30]. Our results further extend the range of Mediator function to include regulation of higher-order chromatin structure in the genome. It is now widely accepted that RNAPII transcribes almost all of the genome. Mediator might not only mediate transcription factors and RNAPII at each gene, but also mediate RNAPII and genome-wide regulation of higher-order chromatin structure.
The S. pombe strains used in this study are described in Table S1. The media and genetic methods used in the study were essentially as described previously [47]. Yeast cells were cultured in YES at 30°C. For deletion or epitope-tagging of the target genes, the PCR-based module method [48] was used.
Silencing assays were conducted from overnight unsaturated cultures grown in 10 ml YES. A 5-fold dilution series of cells was spotted on N/S plates (YES in all spot figures, except Figure 5B), 5-FOA plates (N/S plates with the addition of 1 g/l 5-fluoroorotic acid), and Low Ade plates (N/S plates including limited amount of adenine). The plates were then incubated at 30°C for 3 days.
ChIP was performed as described previously [39] with the following changes: crosslinking with formaldehyde was performed for 30 min and digestion with Proteinase K was carried out for 1 h at 42°C. The following antibodies were used: anti-H3-K9-me2 monoclonal (a gift from T. Urano, Shimane University), anti-Swi6 (produced in-house), anti-RNA polymerase II (8WG16, Abcam), anti-FLAG (M2, Sigma), or anti-myc (4A6, Millipore). Primer sequences are shown in Table S2.
Total RNA was isolated from logarithmically-growing S. pombe (in YES media) using the hot phenol method [49]. For northern blotting of centromeric and mat RNA, 50 µg of total RNA was electrophoresed on a 1% agarose gel containing 1× MOPS and 1% formaldehyde. RNA was transferred to positively-charged nylon membranes (Amersham Biosciences) in 10× SSC by standard capillary blotting. Following UV crosslinking of the RNA to the nylon filter, prehybridization and hybridization were carried out at 42°C in UltraHyb-Oligo buffer (Ambion). For hybridization, 50 pmol oligos were end-labeled with [γ32P]dATP (3000 Ci/mmol) using T4 Polynucleotide Kinase (TOYOBO). After hybridization for 24 h, membranes were washed four times in 2× SSC/0.1% SDS for 10 min at 42°C before exposure to an imaging plate for 1–2 days. For re-probing, probes on the membrane were stripped by boiling in 200 ml of 0.5× SSC/0.1% with shaking. Detection of siRNA was performed as described previously [39]. Oligonucleotides used as probes are shown in Table S2.
For RT-PCR analysis, total RNA was cleaned up and treated with Recombinant DNase I (RNase-free) (TaKaRa) according to the manufacturer's instructions. RT-PCR was performed using PrimeScript Reverse Transcriptase (TaKaRa) according to the manufacturer's instructions. Primer sequences are shown in Table S2.
qPCR was performed using SYBR premix Ex-Taq (TaKaRa) and the Thermal Cycler Dice Real time system TP800 (TaKaRa). Primer sequences are shown in Table S2.
cdc25-22 cells were grown at 25°C to a concentration of 2×106 cells/ml and then shifted to 36°C for 4 hr and 15 min to stop the cell cycle at the G2/M phase. Samples for ChIP assay were collected every 30 min for 300 min after shifting the cells back to 25°C to release cell cycle block. ChIP assay was performed as described in the Experimental Procedures. To prepare RNA for RT-PCR, the input fractions of ChIP were adjusted to 0.25% SDS and 0.25 mg/ml proteinase K and incubated for 45 min at 45°C and then at 65°C for more than 4 hours to reverse crosslinking. Samples were extracted once with phenol-chloroform. After ethanol precipitation, the samples were resuspended in a suitable volume of DEPC-treated distilled water. RT-PCR was performed as described in the Experimental Procedures.
Microarray analysis for gene expression was performed as described previously [50] using FY2002 as a parental strain. White and pink epiclones of med18Δ and med20Δ were analyzed separately. The sequences of the probes and original data from the microarray experiments were deposited to GEO (http://www.ncbi.nlm.nih.gov/geo) with accession number GSE43543.
Methods used in the supplemental information are described in Text S1.
|
10.1371/journal.ppat.1003153 | Plasmodium falciparum Malaria Elicits Inflammatory Responses that Dysregulate Placental Amino Acid Transport | Placental malaria (PM) can lead to poor neonatal outcomes, including low birthweight due to fetal growth restriction (FGR), especially when associated with local inflammation (intervillositis or IV). The pathogenesis of PM-associated FGR is largely unknown, but in idiopathic FGR, impaired transplacental amino acid transport, especially through the system A group of amino acid transporters, has been implicated. We hypothesized that PM-associated FGR could result from impairment of transplacental amino acid transport triggered by IV. In a cohort of Malawian women and their infants, the expression and activity of system A (measured by Na+-dependent 14C-MeAIB uptake) were reduced in PM, especially when associated with IV, compared to uninfected placentas. In an in vitro model of PM with IV, placental cells exposed to monocyte/infected erythrocytes conditioned medium showed decreased system A activity. Amino acid concentrations analyzed by reversed phase ultra performance liquid chromatography in paired maternal and cord plasmas revealed specific alterations of amino acid transport by PM, especially with IV. Overall, our data suggest that the fetoplacental unit responds to PM by altering its placental amino acid transport to maintain adequate fetal growth. However, IV more profoundly compromises placental amino acid transport function, leading to FGR. Our study offers the first pathogenetic explanation for FGR in PM.
| Malaria infection during pregnancy can cause fetal growth restriction and low birthweight associated with high infant mortality and morbidity rates. The pathogenesis of fetal growth restriction in placental malaria is largely unknown, but in other pathological pregnancies, impaired transplacental amino acid transport has been implicated. In a cohort of Malawian women and their infants, we found that placental malaria, especially when associated with local inflammation, was associated with decreased expression and activity of an important group of amino acid placental transporters. Using an in vitro model of placental malaria with local inflammation, we discovered that maternal monocyte products could impair the activity of amino acid transporters on placental cells. Amino acid concentrations in paired maternal and cord plasmas revealed specific alterations of amino acid transport by placental malaria, especially with local inflammation. Overall, our data suggest that, more than malaria infection per se, the local inflammation it triggers compromises placental amino acid transport function, leading to fetal growth restriction. Greater understanding of the mechanisms involved, combined with interventions to improve fetal growth in malaria, are important priorities in areas of the world where the co-existence of malaria and maternal malnutrition threatens the health and lives of millions of young babies.
| Pregnant women living in malaria endemic regions are highly susceptible to malaria, especially in first pregnancies [1], [2]. Malaria in pregnancy is characterized by placental malaria (PM), the selective accumulation of Plasmodium-falciparum infected erythrocytes (IE) in the maternal intervillous blood space of the placenta, in direct contact with the nutrient-transporting epithelium, the syncytiotrophoblast. When placental malarial infection is poorly controlled, chemokine release results in the recruitment of maternal immune cells, predominantly monocytes, to the intervillous blood spaces [3]. The resultant inflammation is termed intervillositis (IV) [4]. In comparison to PM without local inflammation, PM with IV is associated with significant decreases in birthweight and an increased prevalence of low birthweight (LBW) deliveries, primarily due to fetal growth restriction (FGR) [1], [2], [5], [6].
Recent studies have begun to shed light on the pathogenetic mechanisms linking PM and FGR (reviewed in [7]). Inadequate maternal nutrition and placental insufficiency have been proposed. In Congolese women studied by serial ultrasound examinations, FGR associated with PM was 2–8 times more common in undernourished than in well-nourished mothers [8]. The same undernourished mothers with PM had increased uterine artery resistance (Griffin et al. submitted), which is associated with placental insufficiency. A decreased fetal/placental weight ratio is one manifestation of placental insufficiency found in primigravid women with PM [9].
It has previously been suggested [9], [10] that FGR and LBW associated with PM could be caused by impaired capacity of the placenta to transport maternal nutrients, especially amino acids, to the growing fetus. Although this postulate has never been formally tested, it is supported by observations in idiopathic FGR showing that the activities of various placental nutrient transporters are selectively altered [11], [12]. Among the nutrient transporters affected is system A, a group of Na+-dependent neutral amino acid transporters that actively transfer small, neutral amino acids and thereby enables the establishment of high intracellular amino acid concentrations, which are then used to exchange for extracellular essential amino acids via system L [13], [14].
In the placenta, system A activity is mediated by three Na+-dependent neutral amino acid transporter (SNAT) isoforms belonging to the SLC38 gene family; SNAT1 (SLC38A1), SNAT2 (SLC38A2) and SNAT4 (SLC38A4). All isoforms are expressed on the microvillous plasma membrane (MVM) of the human syncytiotrophoblast [15]. A reduced system A amino acid transporter activity in MVM has been consistently observed in placentas of pregnancies associated with FGR [16]–[18], and the reduction in system A activity in MVM correlates well with the severity of FGR [19]. Further, various animal studies have suggested that reduced system A activity may be causally related to the etiology of FGR [20]–[22].
Pro-inflammatory cytokines produced by monocytes have been shown to decrease system A activity. IL-1β reduces system A activity in trophoblast cells [23] and acute exposure to TNF-α resulted in diminished maternofetal transfer of a system A analogue in a rat model of FGR [24]. These cytokines have been associated with LBW in PM, especially when associated with IV [25]–[27], and their production could be caused by PM, either through activation of monocytes by IE [28] or by direct effects of IE on syncytiotrophoblast leading to secretion of cytokines and chemokines [29], [30]. This suggests a link between PM, IV and altered placental amino acid transport, with impacts on fetal growth and development.
In the current study, we hypothesized that the release of soluble mediators triggered by IV associated with PM impairs placental transport of amino acids across the syncytiotrophoblast, contributing to the pathogenesis of FGR and LBW. We found that PM, especially with IV, was associated with decreased placental amino acid uptake and dysregulated maternofetal amino acid balance, likely to alter the transfer of amino acids to the fetus, and to contribute to the pathogenesis of PM-associated FGR.
Characteristics of the individuals who participated in the various aspects of the study are summarized in Table 1.
SLC38A1 transcript levels were reduced (p = 0.008) in the syncytiotrophoblast of infected placentas with IV compared to that of uninfected placentas, while levels in syncytiotrophoblast of infected placentas without IV were intermediate. A similar trend was observed for SLC38A2 transcript levels (Fig. 1A). SLC38A1 (p = 0.017) but not SLC38A2 (p = 0.39) transcript levels were lower in the syncytiotrophoblast of placentas of LBW infants compared to normal birthweight infants (Fig. 1B).
Figure 2A reveals that Na+-dependent MeAIB uptake by MVM vesicles from infected placentas either with or without IV was lower (p≤0.015) than uptake by vesicles from uninfected placentas. Na+-dependent MeAIB uptake was similar between groups with PM (p = 0.65). Birthweight was positively associated with Na+-dependent MeAIB uptake by MVM vesicles from all placentas (Rho = 0.26, p = 0.07; Fig. 2B).
In response to P. falciparum infection, monocytes elicit a pro-inflammatory response including the secretion of IL-1β [28]. IL-1β has been previously reported to decrease Na+-dependent MeAIB uptake by placental trophoblast cells [23]. We therefore investigated whether PM with IV was associated with increased IL-1β concentration and if conditioned medium from a monocyte/IE co-culture could impair Na+-dependent MeAIB uptake.
IL-1β plasma concentration in maternal blood harvested from placentas with PM and IV was higher (p = 0.017) compared to uninfected controls, and comparable (p = 0.1) to the PM without IV group (Fig. 3A). Within the group of PM with IV, IL-1β concentration was negatively correlated with birthweight (Rho = −0.52; p = 0.04; Fig. 3B).
Because IL-1β is produced by monocytes in response to IE [28] and because Na+-dependent MeAIB uptake by MVM vesicles was lowest in PM with IV (Fig. 2A), we speculated that system A activity impairment could be attributable to products generated by monocytes in response to IE.
Medium collected from a monocyte/IE co-culture was used to mimic the intervillous space milieu in cases of PM with IV. This medium was applied to human placental choriocarcinoma BeWo cells to investigate its effect on Na+-dependent MeAIB uptake. Cell viability was monitored in all subsequent experiments and was unaffected by any of the treatments (data not shown).
An inhibition of Na+-dependent MeAIB uptake by BeWo cells when exposed to monocyte/IE co-culture conditioned media was consistently observed (Fig. 4A). An IL-1β blocking antibody was used to investigate the role of IL-1β in mediating this effect (Fig. 4B). At the concentration used, the blocking antibody was effective in abolishing recombinant IL-1β-mediated reduction in Na+-dependent MeAIB uptake. In contrast, addition of IL-1β blocking antibody had no effect on the inhibition of Na+-dependent MeAIB uptake observed with monocyte/IE conditioned media. This indicated that IL-1β was not a major factor in the reduction in Na+-dependent MeAIB uptake observed with monocyte/IE conditioned media.
Treatment of BeWo cells with uninfected erythrocytes (UE) or with lysed or intact IE that had or had not been opsonized by human Ig did not alter Na+-dependent MeAIB uptake by BeWo cells compared to media control (p≥0.12; data not shown). This suggests that the reduction in Na+-dependent MeAIB uptake by BeWo cells relies on monocytes' response to IE more than effects of IE per se.
We next investigated the potential effect of altered system A activity, indicated by the reduced Na+-dependent MeAIB uptake observed in PM with IV, on fetal amino acid levels by measuring free amino acid concentration in paired maternal and cord plasma samples (Table 2).
A low fetal/placental weight ratio is a marker of placental insufficiency and has been associated with malaria [9], [31], [32] and PM with IV cases had lower fetal/placental weight ratio than uninfected controls (p = 0.036). For a number of amino acids, cord concentration was positively correlated with fetal/placental weight ratio, either among all PM cases or for those with IV (Table 3). Among babies with PM and IV, cord concentrations of several neutral, branched chain amino acids, transported by system L [13] were positively associated with fetal/placental weight ratio. This suggests that intervillositis may lead to placental insufficiency in part through impaired transplacental amino acid transport. There was no positive correlation between cord concentration of amino acids and birthweight (p≥0.29).
Understanding the pathogenesis of PM-associated FGR is critical to the design of novel interventions to decrease its burden [33]. In this study, we identified an impaired placental amino acid uptake and dysregulated maternofetal amino acid balance in PM, especially with IV, providing a pathogenic mechanism for PM-associated FGR through altered transfer of amino acids to the fetus.
The activities of various transport mechanisms in the plasma membranes of the syncytiotrophoblast are dysregulated in idiopathic FGR [34], [35]. In particular, the activity of system A amino acid transporters has often been reported to be downregulated in FGR [17]–[19], and animal studies demonstrate that a reduction in system A activity precedes development of FGR [22].
We observed reduced transcription of the system A transporters SLC38A1 (SNAT1), and, to a lesser extent SLC38A2 (SNAT2), within the syncytiotrophoblast in PM with IV, which is compatible with the reduction we observed in system A activity, and consistent with the involvement of these two SNAT subtypes in the downregulation of system A activity associated with PM with IV. Previous studies of idiopathic FGR have not demonstrated altered SLC38A1 or SLC38A2 transcription in whole placental lysates [36], whereas here we have specifically measured SLC38A1 and SLC38A2 transcript levels from the syncytiotrophoblast. Our data suggest that in cases of PM with IV, the syncytiotrophoblast responds to infection and inflammation by down-regulating the transcription of these SNAT isoforms. As activity of SNATs is partly regulated at the level of transcription [37], this suggests that PM with IV may decrease SNAT-mediated placental amino acid transport.
To investigate this further, we studied system A activity ex vivo. System A activity was decreased in PM alone, and to a greater extent in PM with IV. The relationships observed between PM or birthweight with system A activity and SNAT transcript levels suggest that system A makes important contributions to fetal growth, and that these contributions are compromised by PM, especially PM with IV.
To understand how PM, especially with IV, might impair placental amino acid transport we developed an in vitro model to examine trophoblast cell responses to factors present in the placental intervillous blood space. Regardless of the way they were presented to placental cells, IE alone did not induce a significant decrease in system A-mediated MeAIB uptake. This is in accord with our ex vivo data, and with clinical observations that PM without IV is not associated with FGR [6]. In contrast, conditioned media from monocyte-IE co-cultures, which mimic the intervillous milieu in PM with IV, significantly reduced MeAIB uptake by trophoblast cells, indicating that monocytes participated in eliciting this response. Our evidence suggests that IE activate monocytes to release factors that inhibit system A activity.
A number of factors have been shown to modulate system A activity in placental cells or BeWo layers including cytokines such as IL-1β, IL-6 and TNF-α [23], [38] which are increased in PM [25]–[27] and produced by monocytes in response to IE [28]. Despite IL-1β concentrations being raised in maternal blood of the intervillous space in PM with IV and negatively correlating with birthweight in this group, the decreased Na+-dependent MeAIB uptake by BeWo cells in our in vitro model was not substantially mediated by IL-1β, as illustrated by the inability of blocking antibody to IL-1 β to counteract the inhibitory effect of monocyte-IE co-culture supernatants; the mediator(s) responsible are at present unknown and could either be a factor(s) consumed by malaria-stimulated monocytes or a factor(s) secreted by these cells.
The cause of the decreased amino acid uptake observed ex vivo in MVM from women with PM and IV is not known. As discussed above, it may be mediated by monocyte-derived factors, but these remain to be conclusively identified. Hormones that stimulate system A-mediated amino acid uptake including IGFs [39] and leptin [40] are decreased in PM [41], [42], and these may contribute in part to the decreased system A activity demonstrated in patient samples. In vitro, supernatants from co-cultures (rather than monocytes themselves) inhibit amino acid uptake, suggesting that local depletion of available amino acids by activated monocytes is not a significant contributing factor.
We next assessed whether the observed decrease in system A activity in PM, or effects of PM on other amino acid transport systems, resulted in altered amino acid concentrations in maternal and cord blood. In normal pregnancy, delivery of some amino acids, particularly essential amino acids, is only just sufficient to meet fetal requirements [43], [44]. In pregnancies compromised by severe FGR, maternofetal transfer of amino acids may be reduced [45], [46]. In PM without IV, maternal concentration of a number of amino acids (Asn, Asp, Cys and Trp) was increased compared to uninfected controls, possibly due to reduced uptake of these amino acids by the syncytiotrophoblast, resulting in increased maternal concentrations. In PM with IV, maternal concentration of all amino acids except Ala was either unchanged or elevated compared to uninfected controls, consistent with observations in idiopathic FGR [46]. Cord Ala levels were also lower compared to uninfected controls and positively correlated with maternal levels in PM with IV (Rho = 0.56; p = 0.002). Thus, malaria-related inadequate maternal concentrations of amino acids were not responsible for changes in fetal amino acid concentrations.
We did not see widespread decreases in cord amino acid concentration in the group with PM and IV, as have been described in idiopathic FGR [19], [45]. This lack of widespread impact of PM on cord amino acid concentration could be explained by the degree of severity of the FGR in our study compared to the idiopathic FGR studies. In the latter, birthweight was dramatically decreased, by ∼600 g to ∼1550 g. In contrast, in our cohort, birthweight of control infants was relatively low, and birthweight only differed by ∼200 g between control infants and those with PM and IV (in keeping with larger epidemiological studies in this population [5], [27]). In resource-poor settings such as Malawi, obstetric care is limited, and women with at risk pregnancies due to highly compromised placental function and severe FGR may not be identified for intensive management, but may instead experience pregnancy loss. Malaria is a common cause of stillbirth and miscarriage in such settings [47], [48], and our study design may have resulted in malaria-affected pregnancies with severe FGR being under-represented in our cohorts. Longitudinal studies of at-risk pregnancies may be useful in quantifying the risks and manifestations of severe FGR in malaria-affected pregnancies further.
Differences in cord blood amino acid concentrations between groups suggest that placental transport and/or metabolism of a number of amino acids is altered in PM. In the PM without IV group, the neutral and anionic classes of amino acids were most notably affected, suggesting a selective effect on placental handling of these amino acids. In infected women without IV but not in the group with PM with IV, there was a particularly striking increase in the fetal concentration of the anionic amino acids Asp and Glu, which are taken up into the placenta from the maternal and fetal circulations respectively by system XAG−. The physiological significance of this is unclear at present and it could suggest that IV restored placental transport of these amino acids. However, it is known that Glu uptake from the fetal circulation plays a crucial role in fetoplacental Glu-Gln cycling and may also serve to protect the fetus against Glu neurotoxicity [49]. The similar trend observed for Asn, Gly, Met (system A substrates) and both Met and Leu (system L substrates) also implicates these systems as being affected differentially in PM without IV as against PM with IV. Whether this reflects altered amino acid transport and/or placental amino acid utilization or production has yet to be established. The decrease in system A activity could also indirectly impair the activity of other systems that depend on the gradient of amino acid concentration established by system A for their own activity.
Trp was the only amino acid for which there was a failure to concentrate in cord blood as compared to maternal, in the PM with IV group. In idiopathic FGR, Trp concentration occurs [45] implying that the failure of Trp to concentrate in our cases was related to the presence of PM rather than FGR per se. Other essential amino acids (Ile, Leu, Phe, Thr, Val), which, like Trp, are transported predominantly by system L [50], were concentrated in cord blood as compared to maternal, suggesting that there was no global impairment of system L activity. In placental infections, Trp is catabolised through the kynurenine pathway, notably by the enzyme indoleamine 2,3 dioxygenase [51], and we speculate that similarly increased placental catabolism of Trp in PM contributes to the lack of Trp placental concentrative capacity we observed. Cord Trp concentration was increased in PM without IV, suggesting that Trp placental catabolism may be reduced with acute infection; a change that is then blocked by inflammatory cells in chronic infection.
Taken together, our data suggest that PM alters placental function through effects on multiple amino acid transporter systems, and that these effects are selective for certain amino acids; PM may also increase placental amino acid metabolism. In PM with IV, the dysregulation of maternofetal amino acid concentrations is more pronounced, possibly because the monocytes accumulating in the intervillous space release inflammatory mediators that alter the activity of amino acid transporters in the syncytiotrophoblast.
We have shown effects of PM on one amino acid transport system, system A, in vitro and ex vivo, and found clues for the dysregulation of other placental amino acid transport systems. Amino acid transport is highly complex, with overlapping and interdependent pathways. Although the defects we observed in amino acid transporter activity did not translate directly into lower fetal amino acid concentrations in women with PM and IV, we did observe important correlations in women with PM, or PM and IV, between low fetal/placental weight ratio, an index of placental insufficiency, and low cord levels of critical amino acids. Our evidence calls for studies to further characterize the effects of PM and IV on the activity of system A as well as investigating the activities of systems XAG− and L in the placenta. Such studies should also capture whether PM with IV alters transplacental transport of glucose [52] or lipids which, together with amino acids, form essential substrates for fetal growth [53], [54].
In order to counteract the decrease in placental nutrient transport, nutrient supplementation interventions [55] could be implemented, but should ideally be combined with further research to ensure that such interventions correct, and do not exacerbate [56], defects in transport and fetal growth [57].
Greater understanding of the mechanisms by which PM affects placental nutrient transport, combined with possible interventions to improve fetal growth in malaria, are important priorities in areas of the world where the co-existence of malaria and maternal malnutrition threaten the health and lives of millions of young babies.
From 2001–2006, pregnant women delivering a live singleton newborn in the labor ward of Queen Elizabeth Central Hospital, Blantyre, Malawi were recruited into a case-control study. Cases were defined by the presence of P. falciparum asexual parasites on placental blood smear. For each case identified, two uninfected, age (±2 years) and gravidity-matched controls, negative for malaria parasites by both peripheral and placental smears, were then enrolled. Inclusion and exclusion criteria have been described elsewhere [41]. The College of Medicine Research Ethics Committee, University of Malawi, approved the study and written informed consent was obtained from all participants. Immediately after delivery maternal and cord venous blood were collected and separated by centrifugation. Plasma was stored at −80°C.
One set of placental biopsies was snap-frozen in liquid nitrogen for MVM purification, another set was embedded in optimal cutting temperature (OCT) medium before being frozen at −80°C for laser capture microdissection (LCM) of the syncytiotrophoblast and a last set was fixed in 10% neutral-buffered formalin for malaria infection grading. Placental tissue sections were examined by light microscopy for presence of malaria infection. In infected placentas, 500 randomly-selected intervillous space maternal blood cells were counted as previously described [5], to derive estimates of placental parasite density and monocyte counts (expressed as percentage of all maternal intervillous cells).
Samples used in the study were selected from the cohort of participants based on tissue availability, after assessment of placental histology as described below. Presence of IE in the intervillous space of the placenta defined PM cases. These were sub-grouped into PM with or without IV. IV was defined as a monocyte count ≥5% of all intervillous cells counted [5]. Uninfected placentas were defined as showing no signs of malaria infection or intervillositis.
RNA was extracted from laser-captured syncytiotrophoblast as previously described [58]. Briefly, tissue cryosections immobilized on SuperFrost PLUS slides (Fisher) were air-dried and fixed in acetone. After rehydration, sections were stained with methyl green (Sigma-Aldrich) and dehydrated. Material captured by laser microdissection using a MicroBeam microscope (P.A.L.M. Microlaser Technologies) was catapulted directly into RNA extraction buffer (RLT buffer with β-mercaptoethanol; Qiagen), and RNA extracted using an RNeasy Micro Kit (Qiagen), according to the supplier's recommendations. Purified RNA was eluted and kept at −80°C. RNA (10 ng) was reverse transcribed using Superscript III enzyme mix (Invitrogen) with random hexamers. Transcript levels for SLC38A1 (SNAT1), SLC38A2 (SNAT2) and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide (YWHAZ) were quantified using 1∶4 dilution of cDNA for all samples. Primer sequences are shown in Table 4. Real-time quantitative PCR was performed using previously reported thermal cycling conditions at an annealing temperature of 60°C with SYBR Green 1 (Applied Biosystems) [59]. Transcript levels were quantified against a standard curve generated from a pool of all placental cDNA samples. Preliminary studies confirmed YWHAZ transcript levels were comparable between groups and YWHAZ was used to normalize target gene transcript levels.
Isolation of MVM vesicles from placental biopsies (7.2±1.6 g) was performed using magnesium precipitation and differential centrifugation based on the method of Glazier et al. [60] as described previously [61] with modifications according to Jimenez et al. [62] to allow simultaneous recovery of the basal plasma membrane for other studies.
Frozen placental biopsies were thawed and homogenized in 250 mM sucrose, 10 mM HEPES-Tris, pH 6.95 (Buffer D; 2.5 volumes biopsy weight) and a sample of the homogenate (1 ml) was retained for further analysis. The remaining homogenate was centrifuged at 10000 g for 15 min at 4°C and the supernatant retained. The pellet was resuspended in buffer D (1.5 volumes of initial biopsy weight) and the centrifugation step was repeated. The supernatants were pooled and centrifuged at 125000 g for 30 min at 4°C. The pellet was resuspended in buffer D and 12 mM MgCl2 added and stirred on ice for 20 min. The suspension was centrifuged at 2500 g for 10 min at 4°C. The supernatant (containing MVM) was centrifuged at 125000 g for 30 min at 4°C. The pellet was resuspended in 300 mM sucrose, 20 mM Tris-maleate, pH 7.4 and loaded onto a discontinuous 25%–37%–45% sucrose gradient. After centrifugation at 90000 g for 6 h at 4°C, the MVM fraction at the 37%–45% interface was recovered and centrifuged at 110000 g for 30 min at 4°C. The resultant MVM pellet was resuspended in intravesicular buffer (290 mM sucrose, 5 mM HEPES, 5 mM Tris-HCl, pH 7.4; 3 volumes pellet weight) and repeatedly passed through a 25-gauge needle to vesiculate the MVM fragments to form vesicles. MVM vesicles were stored at −80°C.
Protein concentration of placental homogenate and MVM vesicles was determined by the Lowry method [63]. Purity of MVM vesicle preparations was assessed by enrichment of alkaline phosphatase activity as described previously [60]. Alkaline phosphatase enrichment factors (mean ± SD) were not different (p = 0.39) between uninfected (16.8±8.6; n = 18), PM (11.5±6.0; n = 14) and PM with IV (12.9±6.7; n = 21) groups, suggesting comparable MVM purity between groups.
14C-MeAIB (NEC-671; PerkinElmer,) was used as a well-characterized, non-metabolizable amino acid analogue substrate to measure the activity of system A amino acid transporter [15], [17].
Placental blood was aspirated from an incision made in the basal plate at a pericentral site of the placenta. Plasma was separated and frozen at −80°C. IL-1β was measured by ELISA (DuoSet, R&D) in undiluted plasma samples according to the manufacturer's instructions.
Samples from primigravidae were preferentially selected for the amino acid analysis, as they are at highest risk of malaria in pregnancy [2]. In paired maternal and cord blood plasma samples, free forms of the common 20 amino acids were analyzed by reversed phase ultra performance liquid chromatography (RP-UPLC) with pre-column derivatization with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate (AccQ Tag Ultra; Waters Corporation). Standards (Amino Acid Standard H (Pierce) together with Gln, Trp and Asn (Sigma)) were prepared with norvaline (Sigma) included as an internal standard. For assay, 100 µl plasma was mixed with an equal volume of 200 µM norvaline and deproteinated by ultrafiltration through a 10 kDa MWCO spin filter (Millipore) at 4800 g for 60 min at 10°C. 10 µl filtrate was derivatized and analyzed on a Waters Acquity UPLC over a 20 min gradient using a 2.1×150 mm, 1.7 µm i.d., BEH C18 column (Waters Corporation) flowing at 0.6 ml/min at 60°C. Detection was via UV (260 nm) and data was collected and analyzed using the Waters Empower2 software.
Non-normally distributed data are presented as box plots showing the median, 25th/75th and 10th/90th centiles unless described otherwise. Values out of the 10th centiles were included in statistical analyses but not represented in graphs. Normally distributed values are presented as mean and standard deviation. Non-normally distributed data were normalized by log-transformation prior to statistical analysis. Data were then compared between 3 groups using one-way ANOVA. When the p value of the ANOVA test was lower than 0.1, two-group comparisons were made using a 2-tailed T-test. Correlations were assessed using Pearson's correlation test. Trends across ordered groups were tested using Cuzick's test. The ability of the placenta to concentrate amino acids in cord blood was examined by testing if the cord to maternal concentration ratio of each amino acid was different from 1 using a one-sample T-test.
The College of Medicine Research Ethics Committee, University of Malawi, approved the study and written informed consent was obtained from all participants prior to inclusion in the study.
|
10.1371/journal.pbio.1000196 | Loss of Mitogen-Activated Protein Kinase Kinase Kinase 4 (MAP3K4) Reveals a Requirement for MAPK Signalling in Mouse Sex Determination | Sex determination in mammals is controlled by the presence or absence of the Y-linked gene SRY. In the developing male (XY) gonad, sex-determining region of the Y (SRY) protein acts to up-regulate expression of the related gene, SOX9, a transcriptional regulator that in turn initiates a downstream pathway of testis development, whilst also suppressing ovary development. Despite the requirement for a number of transcription factors and secreted signalling molecules in sex determination, intracellular signalling components functioning in this process have not been defined. Here we report a role for the phylogenetically ancient mitogen-activated protein kinase (MAPK) signalling pathway in mouse sex determination. Using a forward genetic screen, we identified the recessive boygirl (byg) mutation. On the C57BL/6J background, embryos homozygous for byg exhibit consistent XY gonadal sex reversal. The byg mutation is an A to T transversion causing a premature stop codon in the gene encoding MAP3K4 (also known as MEKK4), a mitogen-activated protein kinase kinase kinase. Analysis of XY byg/byg gonads at 11.5 d post coitum reveals a growth deficit and a failure to support mesonephric cell migration, both early cellular processes normally associated with testis development. Expression analysis of mutant XY gonads at the same stage also reveals a dramatic reduction in Sox9 and, crucially, Sry at the transcript and protein levels. Moreover, we describe experiments showing the presence of activated MKK4, a direct target of MAP3K4, and activated p38 in the coelomic region of the XY gonad at 11.5 d post coitum, establishing a link between MAPK signalling in proliferating gonadal somatic cells and regulation of Sry expression. Finally, we provide evidence that haploinsufficiency for Map3k4 accounts for T-associated sex reversal (Tas). These data demonstrate that MAP3K4-dependent signalling events are required for normal expression of Sry during testis development, and create a novel entry point into the molecular and cellular mechanisms underlying sex determination in mice and disorders of sexual development in humans.
| In mammals, whether an individual develops as a male or female depends on its sex chromosome constitution: those with a Y chromosome become males because of the development of the embryonic gonad into a testis. The Y-linked sex determining gene SRY regulates this process by initiating a pathway of gene and protein expression, including the expression of critical autosomal genes such as SOX9. We identified a mouse mutant that causes embryonic gonadal sex reversal: the development of ovaries in an XY embryo. This mutant, which we called boygirl (byg), was shown to contain an early stop codon that disrupts the autosomal gene encoding MAP3K4, a component of the mitogen-activated protein kinase (MAPK) signaling pathway. Analysis of embryonic XY gonads suggests that sex reversal is caused by delayed and reduced expression of the sex-determining gene SRY. Our data indicate, for the first time, a requirement for MAPK signaling in the developing XY gonad in order to facilitate normal expression of SRY and the downstream testis-determining genes and also suggest that reduced dosage of MAP3K4 may be the cause of a previously described autosomal sex-reversing mutation in the mouse. We predict that loss of MAP3K4 or other MAPK components may underlie disorders of sexual development (DSD) in humans as well.
| Sex determination is the process by which an embryo develops into a male or female, namely, the formation of testes in an XY embryo and ovaries in an XX embryo. In the mouse, this process begins with commitment of cells of the bipotential genital ridge to either the testicular or ovarian fate at 11.5 d post coitum (dpc) [1]. In mammals such as mice and humans, this commitment depends on the presence or absence of the Y-linked testis-determining gene, SRY [2]–[4].
During the search for the elusive mammalian testis-determining factor, it was a criterion of correct identification that any candidate gene be associated with mutations that cause pure (gonadal) XY sex reversal: the development of an ovary in an XY individual. Such mutations in SRY were readily discovered in mice [5] and humans [6] exhibiting sex reversal, and this link with sex reversal has been a constant theme in the subsequent identification of novel, mostly autosomal, genes functioning in sex determination.
Instances of XY sex reversal in the mouse associated with single gene mutations remain relatively uncommon. Excluding Sry, they include targeted mutations of Sox9 [7],[8], Dax1 [9], Fgf9 [10], Fgfr2 [11],[12], Gata4/Fog2 [13],[14], Cbx2 (M33) [15], and Wt1(+KTS) [16]. Mice harbouring targeted mutations in three members of the insulin-receptor signalling pathway also exhibit XY sex reversal [17]. In several of these cases, variability exists in the degree of sex reversal observed, depending on genomic context. The C57BL/6J background often biases gonadal development in favour of ovarian tissue in mutant XY embryos and this “B6 sensitivity” increases still further if the AKR/J Y chromosome (YAKR) is present [14]. Additional genes have been identified that disrupt testis development, affecting testis cord formation or the differentiation of testis-specific cell lineages. These include Dhh [18]–[20], Pdgfra [21], Pod1 [22], Arx [23], Wnt4 [24], and Spry2 [25]. The contribution of other protestis genes to sex determination, such as Sf1 [26], Dmrt1 [27], and Sox8 [7], can be difficult to discern owing to functions of such genes earlier in gonad development or functional redundancy.
In addition to the contribution of specific genes, other autosomal loci have been reported to control sex determination in the mouse. Such loci have been identified on the basis of genetic segregation in cases of sex reversal observed when the Y chromosome of the C57BL/6 strain is replaced by that of Mus domesticus poschiavinus [28], or on the basis of their modifying the phenotypic effect of another sex determining locus [29],[30].
The search for novel sex determining genes has been driven in recent years by the transcriptional properties of candidate genes identified by expression profiling [31]–[35]. However, such gene-driven approaches have not yielded a significant number of novel sex reversal phenotypes or abnormalities of gonadal differentiation that could act as important models for the investigation of the molecular genetic basis of sex determination. Notable exceptions to this general observation include the genes Cyp26b1 [36],[37] and Pgds [38],[39], whose roles in germ cell and somatic cell development, respectively, were established partly on the basis of earlier observations on their male-specific expression derived from systematic expression screens.
As an alternative to expression-based screens, we have employed a forward genetic approach to identifying loci controlling sexual development in the mouse. Using N-ethyl-N-nitrosourea (ENU) mutagenesis and a three-generation (G3) breeding scheme, we screened for abnormalities of the developing gonads in embryos homozygous for induced mutations. In one mutant pedigree, RECB/31, we identified XY embryos exhibiting abnormal testis cord development and, in some cases, gonadal sex reversal. We have named this mutant line boygirl (byg). Genetic mapping placed byg on the proximal region of mouse Chromosome 17 and molecular studies revealed that the byg phenotype is caused by a point mutation in the Map3k4 (Mekk4) gene. Embryos doubly heterozygous for both the Map3k4byg allele and a targeted null allele of Map3k4 (Map3k4tm1Flv) exhibited neural tube defects and XY gonadal sex reversal, confirming that Map3k4 is the causal gene. Map3k4 encodes a mitogen-activated protein kinase (MAPK) kinase kinase, demonstrating for the first time a role for MAPK signalling in mammalian sex determination. We describe molecular and cellular studies on the byg mutant that demonstrate a requirement for mitogen-activated protein kinase kinase kinase 4 (MAP3K4) in regulating XY gonadal growth, mesonephric cell migration, and the expression of Sry, and hence Sox9, during XY gonad development. We also describe genetic experiments that suggest that loss of Map3k4 is responsible for a previously reported autosomal sex reversal phenomenon, T-associated sex reversal (Tas) [40],[41].
Line 31 (RECB/31) was identified in a forward genetic (phenotype-driven) screen for embryonic gonad abnormalities after ENU mutagenesis (see Materials and Methods for details). Embryos homozygous for ENU-derived mutations were isolated and examined for a variety of morphological abnormalities. One RECB/31 embryo, dissected at 13.5 dpc, exhibited spina bifida, mild oedema, and also contained gonads shaped like normal testes but with no visible testis cords (Figure 1A and 1B). A second, independent RECB/31 litter contained an embryo with spina bifida and testes that had fewer cords than normal with an irregular morphology (Figure 1C). Having identified these individuals, subsequent RECB/31 embryos were examined and gonads were collected for sexing and wholemount in situ hybridisation (WMISH). In this manner, another XY individual was identified in which the gonads were morphologically ovarian at the same stage (Figure 1D). WMISH analysis of gonads from these three abnormal embryos using the Sertoli cell marker Sox9 confirmed the disruption to testis development and its variable severity as described above (Figure 1B–1D). In each case, Sox9 expression was still prominent. However, in the case of the XY gonad with an ovarian appearance, expression was restricted to the central portions of the gonad and absent from the poles. This observed phenotypic variability, and that of subsequent mutants identified in the RECB/31 pedigree, is likely due to the mixed genetic background of the embryos examined. All embryos with abnormal XY gonads exhibited failure of neural tube closure, either spina bifida or exencephaly (unpublished data). Embryonic death of homozygous mutants was commonly observed after 15.5 dpc. Because of the observed gonadal abnormalities and apparent XY gonadal sex reversal, this mutant line was named boygirl (byg).
During subsequent generations of backcrossing onto C3H/HeH the gonadal phenotype was still robust, although the majority of RECB/31 XY gonads had the appearance of ovotestes, in which the central portion of the gonad shows evidence of cord formation, but the poles are ovarian in both appearance and marker expression (Figure 1E–1N). No overt abnormalities were observed in XX byg/byg gonads in these marker studies.
Identification of additional affected XY gonads permitted mapping of the byg mutation. Abnormal embryos (n = 9) were typed with a genome-wide panel of 55 SNP markers in order to identify chromosomal regions that were consistently homozygous for the C57BL/6-derived allele. Only one region, on proximal mouse Chromosome 17, showed this feature of genetic association with byg. This initial linkage was refined by subsequent backcrossing of byg carrier males with C3H/HeH females and intercrossing of carrier progeny, identified by SNP haplotype analysis. Additional SNPs were then used to identify a critical region, in which the byg mutation must reside, between 9.66 Mb (rs3665053) and 15.32 Mb (rs13482889) on Chromosome 17. We used an informatics-based approach to identify candidate genes in the byg critical region. One such candidate was the gene Map3k4 (also known as Mekk4, GenBank [http://www.ncbi.nlm.nih.gov/Genbank] number NM_011948), which encodes a MAPK kinase kinase [42],[43]. Mice lacking this gene, which were generated previously by gene targeting, are associated with perinatal lethality on the C57BL/6 background [44]. Because homozygous Map3k4 mutant embryos also exhibit neural tube defects and because Map3k4 is expressed in most embryonic tissues between 9.5 and 15.5 dpc [42],[44],[45], including the gonads (Figure 2A and 2B), we examined the sequence of Map3k4 in affected byg/byg embryos. A single nucleotide substitution at nucleotide position 1,144 of the Map3k4 open reading frame was identified in the homozygous form in two independent byg/byg mutants (Figure 2C and 2D). This substitution replaces an arginine with a premature stop codon at amino acid position 382 of the 1,597 amino acid MAP3K4 protein. The predicted truncated protein lacks the critical kinase domain of MAP3K4 and, therefore, lacks any MAPKKK function (Figure 2E). Absence of full-length (180 kDa) MAP3K4 protein in byg homozygous mutants was confirmed by Western blotting with an anti-MAP3K4 antibody (Figure 2F). A kinase-inactive allele of Map3k4 has previously been shown to have very similar phenotypic consequences to the null allele [45]. Thus, because of the effect of the premature stop codon causing loss of the kinase domain, we conclude that the Map3k4byg allele is a null allele. The entire colony of byg mice was typed for the presence of the mutation in Map3k4 and all known byg carriers were heterozygous for the mutation. The mutation was not found in any wild-type C57BL/6J or C3H/HeH mice. We concluded, therefore, that the gonadal phenotype in mutant byg embryos is caused by loss of MAP3K4 function.
To confirm this, and discount the possibility that a second, closely linked mutation in an unrelated gene was responsible for the gonadal phenotype, we studied a second Map3k4 mutant allele (Map3k4tm1Flv) generated by gene targeting [44]. Embryos homozygous for the Map3k4tm1Flv allele exhibit neural tube defects and die perinatally, although there have been no descriptions of sexual development in these individuals. Embryos doubly heterozygous for both Map3k4byg and Map3k4tm1Flv were examined at 14.5 dpc and exhibited neural tube defects (unpublished data). All XY embryos contained gonads with an overt ovarian appearance (Figure 2G), and these failed to express Sox9 at significant levels (Figure 2H). The absence of any overt testicular tissue in these XY embryonic gonads is likely due to the increased contribution from the C57BL/6J genome in these individuals. Embryos homozygous for the targeted allele, which was maintained on the C57BL/6J background, also exhibited gonadal sex reversal, with affected XY embryos containing gonads with ovarian morphology, lacking Sox9 and expressing Wnt4, a marker of ovarian differentiation (Figure 2I and 2J). Thus, these data confirm that Map3k4 is the gene disrupted in the byg mutant and that MAP3K4 is required for testis determination in mice.
Evidence exists that the C57BL/6J background is sensitised to disruptions to XY gonad development, and this conclusion appeared to be supported by the increased severity of the XY gonadal phenotype observed in embryos heterozygous for both Map3k4byg and Map3k4tm1Flv, and homozygous for Map3k4tm1Flv, in which the contribution from C57BL6/J was greater. Thus, we performed a detailed examination of embryos homozygous for Map3k4byg after backcrossing to C57BL6/J for at least two generations. We examined cell proliferation, mesonephric cell migration, and cellular differentiation in mutant and wild-type gonads because all these processes are required for normal testis development [1],[46],[47].
Cellular proliferation is an important component of the organogenetic programme of testis development [48],[49]. Gonadal cell proliferation was examined at 11.5 dpc (17–18 tail somites [ts]), 12.0 dpc (20–22 ts), and 12.25 (24 ts) in the coelomic region of gonads from byg/byg and control littermates using immunostaining with an antibody for the mitotic marker phosphorylated histone H3 (pHH3). Somatic cell proliferation in XY byg/byg gonads appeared reduced in the coelomic region in comparison to wild-type XY embryonic gonads at all stages examined (Figure 3; Table 1). Moreover, at the 22- and 24 ts stages (12.0–12.25 dpc), the coelomic region of control XY gonads was thickened and contained a larger number of somatic cells, in contrast to byg/byg XY gonads, which had fewer cells in this region and resembled wild-type XX gonads of the same stage (Figure 3). We conclude that cellular proliferation, and thus gonadal growth, in the coelomic region is severely compromised in byg/byg XY gonads at an early stage.
Increased levels of apoptosis have previously been described in the neural tube of mice lacking Map3k4 [44]. For this reason we examined levels of apoptosis in the byg/byg XY gonad and controls at 17 ts using an antibody to cleaved caspase 3. We observed very few positive cells in mutant and control gonads, although large numbers of apoptotic cells were observed in a positive control (interdigital mesenchyme of the developing limb) using this assay (unpublished data). Thus, we cannot attribute impaired gonadal growth in XY byg/byg embryos to increased levels of apoptosis.
Testis cord formation in the mouse requires cell migration from the associated mesonephros into the XY gonad in a male-specific fashion [50]–[53]. To examine mesonephric cell migration into the XY byg/byg gonad we first examined development of mutant gonads when explanted from the embryo at 11.5 dpc and cultured in vitro. Control gonads (wild-type and byg/+ littermates) formed clear testis cords after 2 d of culture and expressed the Sertoli cell marker, Sox9 (n = 3) (Figure S1A). In contrast, we did not observe any testis cords in cultured XY gonads from byg/byg embryos (n = 3) (Figure S1B and S1D). WMISH analysis revealed that these cord-free XY gonads failed to express Sox9 (Figure S1B), indicating a failure to execute the program of testis differentiation. In contrast, high levels of Wnt4 expression in the mutant XY gonads indicated an activation of the ovarian pathway (Figure S1D).
To examine whether the severe disruption to cord formation in the byg/byg gonad was associated with any loss of mesonephric cell migration, we performed recombination experiments in which mesonephroi ubiquitously expressing green fluorescent protein (GFP) were cultured adjacent to a byg/byg XY gonad from 11.5 dpc. Cell migration into control XY gonads was prominent after 48 h of culture (Figure S1E). In contrast, little or no cell migration was detected in cultured byg/byg gonads (Figure S1F). These data suggest that two early cellular processes associated specifically with XY gonad development, cell proliferation in the coelomic growth zone and mesonephric cell migration, are disrupted in the absence of MAP3K4.
In order to address the molecular basis of these defects, we next investigated the expression of key male- and female-determining genes and gene-products between 11.5 and 14.5 dpc, stages of gonad development between which the male and female fates are established and the programme of sexually dimorphic morphogenesis is executed. Several molecules have been shown to be required for normal testis determination, including SRY [5], fibroblast growth factor 9 (FGF9) [10], FGFR2 [11],[12], and SRY-like HMG box 9 (SOX9) [7],[8]. Current understanding suggests that SRY, in concert with SF1, acts to up-regulate Sox9 expression in the XY gonad at 11.5 dpc [54],[55]. Sox9 expression is then maintained at a high level by a positive feedback loop with FGF9/FGFR2, and acts to antagonise function of the ovary-determining gene Wnt4 [56]. A role for prostaglandin D2 in the regulation of Sox9 expression has also been proposed [12],[39],[57],[58]. Downstream of SOX9, genes such as Amh [59] and Vanin-1 [31],[32],[60], with male-determining effects, are transcriptionally activated, and germ cell fate is established by modulation of retinoic acid signalling [36],[37]. These molecular events are associated with precise spatial (cellular and subcellular) and temporal expression profiles of genes and their protein products, often in a sexually dimorphic manner.
Given its central role in testis development we began our study with an analysis of Sox9 expression. From 11.5 dpc onwards Sox9 transcription in control XY gonads is prominent, initially in pre-Sertoli cells and subsequently in Sertoli cells of the seminiferous cords/tubules. However, analysis of byg/byg homozygotes revealed dramatically reduced levels of Sox9 transcript (Figure 4). At 14.5 dpc the byg/byg XY gonad resembles an ovary morphologically and no significant Sox9 transcription was detectable (Figure 4A). This loss of a Sertoli cell marker in mutant XY gonads was accompanied by elevated expression of two known female-specific markers at the same stage, Stra8 and Wnt4 (Figure 4B and 4C). Expression of these genes indicates that the ovarian pathway of development, including entry of germ cells into meiosis, is activated in vivo in the absence of MAP3K4.
At 11.5 dpc, the sex-determining stage of gonadogenesis, little or no Sox9 transcript was observed (Figure 4E), and this loss of expression was confirmed by immunostaining of mutant and control gonads at the same stage with an anti-SOX9 antibody (Figure 4G–4I). However, Wnt4 expression was prominent in the XY byg/byg gonad at 11.5 dpc, in contrast to wild-type controls (Figure 4F). Interestingly, Sox9 transcription at 11.5 dpc in mutant gonads on the C3H/HeH background was reduced in comparison to controls (Figure 4D), but not to the same degree as the C57BL/6J-derived mutant gonads at the same stage. Loss of Sox9 expression is associated with XY sex reversal in a number of genetic contexts, and mice homozygous for a loss-of-function allele of Sox9 targeted to the developing XY gonads by Cre-mediated excision exhibit immediate, complete gonadal sex reversal, as evidenced by the expression of female-specific markers and the absence of testis cord formation [8]. Thus, loss of Sox9 expression is sufficient to explain the failure of male-specific events in XY byg/byg homozygotes, such as enhanced coelomic region growth, mesonephric cell migration, and testis cord formation.
We next analysed expression of several other important markers of male and female gonad development around 11.5 dpc (16 to 19 ts) using immunohistochemical staining of transverse sections. SF-1 (NR5A-1) is thought to mediate up-regulation of Sox9 transcription in the early XY gonad by acting on a specific gonadal enhancer (TESCO) in synergy with SRY [55]. We observed no significant difference in the expression of SF-1 between wild-type and byg/byg XY gonads at this stage, with large numbers of somatic cells exhibiting nuclear staining in both genotypic classes (Figure S2A and S2B). FGFR2, a gonadal receptor for FGF9, has been reported to exhibit a sexually dimorphic profile of expression in the gonads at 11.5 dpc, with somatic cells in the body of the XY gonad exhibiting nuclear localisation of the protein and XX somatic cells, in contrast, exhibiting a cytoplasmic localisation [12],[61]. We also observed nuclear localisation of FGFR2 in somatic cells of control XY gonads at 11.5 dpc (Figure S2F and S2G), but in XY byg/byg gonads, although FGFR2 expression was still prominent, its localisation was cytoplasmic, resembling XX control gonads at the same stage (Figure S2H–S2J).
Next, we examined the early expression of FOXL2, a protein required for normal ovary development [62]–[65]. Foxl2 transcription has been reported to be up-regulated in the developing mouse gonad around the time of sex determination [66] and restricted to somatic cells [62],[63]. In newborn mice FOXL2 protein is expressed in the nuclei of pregranulosa cells [63]. We observed expression of FOXL2 in the nuclei of somatic cells in wild-type XX gonads at 11.5 dpc (Figure S2E), but negligible expression was observed in wild-type XY gonads (Figure S2C). However, striking up-regulation of FOXL2 was observed in the nuclei of somatic cells of byg/byg XY mutants (Figure S2D). Together with prominent expression of Wnt4 transcript in mutant gonads at the same stage (Figure 4F), these data suggest that the ovarian determining pathway is activated at an early stage in the gonads of XY byg/byg embryos lacking MAP3K4.
Absence of a number of molecules has been reported to cause reduction or loss of Sox9 expression in mutant mouse gonads, including FGF9 [56], FGFR2 [12], WT1 [16],[67], and DAX1 [9],[29]. Recently, it has been shown that SRY and SF-1 cooperatively bind a specific enhancer element (TESCO) to up-regulate Sox9 transcription during XY gonad development and that SOX9 subsequently acts to maintain its own expression by binding to the same enhancer [55]. Because of this central role for SRY in regulation of Sox9 expression, we investigated the expression of Sry in XY byg/byg gonads (Figure 5). Sry transcription reaches a peak at 11.5 dpc (17–18 ts) in XY mouse gonads, and so we studied expression at this stage using in situ hybridisation. At 17 ts we observed Sry transcripts in wild-type XY gonads using WMISH. However, no significant Sry transcription was observed in XY mutant gonads at the same stage (Figure 5A). At the 19 ts stage, Sry transcription is reduced in the wild-type gonads and still absent from mutant (Figure 5B). We utilised quantitative real time-PCR (qRT-PCR) to confirm this reduction in Sry expression in mutant gonads at 11.5 dpc (Figure 5C). This qRT-PCR study revealed an almost 3-fold reduction in Sry transcript levels in XY byg/byg gonads. Sf1 transcript levels did not differ significantly between mutant and control gonads, in line with our immunohistochemistry data. Fgf9 transcript levels appeared to be reduced in XY byg/byg gonads, although this difference was not statistically significant. We then studied the expression of SRY protein in mutant and control gonads at the same stage using a specific antibody to SRY [39],[68]. Expression of SRY was observed in somatic cells of the developing gonad at 11.5 dpc in control XY gonads (Figure 5D and 5F). In contrast, very few SRY-positive cells were detected in XY byg/byg gonads, which resembled XX control gonads at the same stage of development (Figure 5E, 5G, and 5H). High magnification examination of XY byg/byg gonads at these stages also revealed that those cells that did express SRY did so at a greatly reduced level (Figure 5I and 5J). In contrast to wild-type controls, no SRY-positive cells were detected at 11.0 dpc (Figure 5K and 5L). These studies suggest that appropriate expression of Sry in XY gonads, at both the transcript and protein level, is dependent on the presence of MAP3K4. In the absence of MAP3K4, Sry expression is delayed and, at 11.5 dpc, severely reduced. Reduced or delayed expression of Sry is known to be a cause of XY gonadal sex reversal [69],[70].
MAP3K4 activity results in activation of the p38 and JNK MAPK pathways as part of a three-kinase phosphorelay module [71]. This signalling module is thought to regulate, amongst other things, the cell's response to stress including ultraviolet radiation, heat shock, and osmotic stress [72]. MAP3K4 regulates the MAPKs p38 and JNK via the phosphorylation of the MAP2Ks MKK3/MKK6 and MKK4/MKK7, respectively [42],[43]. A reduction in the number of cells positive for activated MKK4 activity has been reported in the neuroepithelium of embryos lacking MAP3K4 [44]. Therefore, we assayed for the presence of activated MKK4 in wild-type XY gonads at 11.5 dpc using antibodies specific for the phosphorylated form of this protein (pMKK4). pMKK4-positive cells were observed in the gonad, but these were primarily found in the coelomic region of the gonadal periphery (Figure 6A and 6B), a profile reminiscent of pHH3-positive mitotic cells (Figure 3A). A similar distribution was observed when pMKK7-positive cells were imaged (Figure 6H). Given these observations, we assayed directly for co-expression of pMKK4 and pHH3 in the gonad at 11.5 dpc using immunostaining of sections. pMKK4-positive cells were found to be positive for pHH3 too, both in the gonad and adjacent mesonephros (Figure 6B–6D). We then assayed for the presence of activated p38 (pp38) and pMKK7 in the same tissue sections, and observed a similar pattern of pp38- and pMKK7-positive cells at the gonadal periphery, which were also positive for pHH3 (Figure 6E–6J). The co-expression of pMKK4 and pHH3 was also observed in XY byg/byg gonads at the same stage. In the case of pMKK4, pMKK7, pp38, and pJNK, cells positive for these activated proteins were still detectable in XY byg/byg gonads at 11.5 dpc (Figure S3), consistent with residual pMKK4 expression in the neural tube of embryos lacking MAP3K4 [44]. These data suggest that MAPK signalling is active in the developing XY gonad at early stages, and is associated with proliferating cells of the coelomic growth zone, but that alternative pathways exist for MAPK activation in the gonad in the absence of MAP3K4. Moreover, given our observation that mitotic somatic cells in the coelomic region are those cells with activated MKK4/7 and activated p38, the reduction in the number of proliferative cells in the XY byg/byg gonad (Figure 3) corresponds to a reduction in the number of pMKK4/7- and pp38-positive cells. Whilst it is possible that a gonadal somatic cell activates the MAPK pathway only once it enters mitosis, it is more consistent with the known role of MAPK signalling in cell proliferation to conclude that male-enhanced proliferation in the coelomic region is a MAPK-dependent process. The reduction of coelomic region growth in the XY byg/byg gonad at 11.5 dpc is thus explicable by a reduction in the number of cells exhibiting MAP3K4-mediated phosphorylation of MKK4/7, p38, and possibly other MAPK signalling components.
In order to address whether disruption to components of MAPK signalling can disrupt testis development in vitro, we cultured wild-type embryonic XY gonads from 11.5 dpc for 48 h in the presence of highly selective inhibitors of the MAPKs extracellular signal-related kinase (ERK) (U0126) and p38 (SB202190) [73]. We then assayed for the expression of Sox9 using WMISH (Figure 6K–6M). Similar experiments to address the role of JNK were not performed because of the unavailability of highly specific small molecule inhibitors. We observed little affect on Sox9 expression in gonads treated with ERK inhibitor when assayed by WMISH, although testis cord formation did not occur in treated samples with the same efficiency as samples cultured in vehicle control (Figure 6K). These data are consistent with other reports that MEK1/ERK inhibitors fail to significantly disrupt testis development in vitro [74]. In contrast, culturing in the presence of p38 inhibitor resulted in dramatic reduction of Sox9 expression, including an almost complete loss of signal in 50% of treated samples (n = 8) (Figure 6L and 6M). Examination of Wnt4 expression in SB202190-treated cultured explants (n = 3) also revealed robust expression of this ovarian marker in contrast to vehicle controls (Figure 6N), suggesting that at least partial gonadal sex reversal was occurring during culture of XY explants because of abrogation of p38 activity. In this context, it is interesting to note that human SRY has been recently identified as a possible target of p38 MAPK signalling in cultured keratinocytes [75].
Given the importance of two components of the FGF signalling pathway, FGF9 and FGFR2, in XY gonad development, we next studied whether byg/byg gonads exhibited defects in this pathway by determining whether FGF9 was able to activate Sox9 transcription in XY byg/byg gonads. It has previously been reported that FGF9 is capable of activating Sox9 transcription in developing XX gonads if they are cultured in the presence of beads coated in this growth factor [56]. In an attempt to address the question of which upstream, extracellular signals employ MAP3K4-dependent phosphorylation during XY gonad development, we determined whether FGF9-mediated activation of Sox9 transcription was abrogated in MAP3K4-deficient gonads. XY gonads from byg/byg and control embryos were cultured at 11.5 dpc for 48 h in the presence of FGF9-coated beads (or beads coated in bovine serum albumin [BSA]) and were then analysed for the presence of Sox9 transcripts in cells contacting the bead using in situ hybridisation. BSA-coated beads did not induce Sox9 transcription in any samples. In contrast, Sox9 transcripts were clearly detected in the vicinity of beads in both cultured wild-type XX gonads and in XY byg/byg gonads (Figure 7A and 7B). These data suggest that MAP3K4 is not an obligatory component of signal transduction pathways employed by FGF9 to activate transcription of Sox9 in the developing gonad. However, failure of normal SRY, and thus SOX9, expression in byg/byg XY gonads may result in failure to establish the positive feedback loop between SOX9 and FGF9/FGFR2 [56].
A locus on mouse Chromosome 17 associated with XY sex reversal and ovotestis formation has previously been described [40]. This mutation, known as Tas, was identified in a mouse stock carrying the hairpin-tail (Thp) deletion whilst being crossed to the C57BL/6J background. The presence of an AKR/J-derived Y chromosome is also required for the development of ovarian tissue in XY C57BL/6J Thp/+ individuals. It has been hypothesized that the Tas mutation resides within the region of the t complex deleted in Thp and hemizygosity for the relevant locus causes varying degrees of sex reversal when on the C57BL/6J YAKR background [41]. This genetic background is known to be very sensitive to disturbances in the early events of testis development induced by gene mutation [14], and thus one potential explanation for the Tas phenotype is haploinsufficiency for a t complex locus that is ordinarily testis determining. Map3k4 is located on proximal mouse Chromosome 17 in the region of the t complex and, in the form of the previously anonymous DNA marker D17Rp17 (still a synonym of Map3k4, see http://www.informatics.jax.org/searches/accession_report.cgi?id=MGI%3A1346875 and GenBank entry NM_011948), has been shown to map within the Thp deletion [76]. Given this map position and the gonadal phenotype of XY embryos lacking functional Map3k4 on C57BL/6J, we hypothesized that haploinsufficiency for this gene might be, at least partially, responsible for the Tas gonadal sex reversal phenotype.
We tested this model in two ways. First, we generated embryos doubly heterozygous for the byg mutation and the Thp deletion. If Map3k4 resides within the Thp deletion these embryos will lack Map3k4 function because of failure of complementation and will recapitulate the phenotype of byg/byg homozygous embryos. Figure 8 shows that XY byg/+, Thp/+ embryos exhibited abnormalities of testis development. XY gonads dissected from doubly heterozygous embryos at 13.5/14.5 dpc showed disruption to cord morphology or gonadal sex reversal, in which Sox9 transcription is lost (Figure 8A) and Wnt4 transcription is activated (Figure 8B). Doubly heterozygous mutants also exhibited neural tube defects (unpublished data). We performed this cross on the C3H/HeH background because this strain has not previously been associated with sensitisation to events disrupting testis development, even given the presence of the YAKR chromosome [41]. We confirmed, therefore, that Map3k4 resides in the Thp deletion and that this deletion, combined with a loss-of-function allele of Map3k4, causes varying degrees of disruption to XY gonad development even in the absence of any other predisposing genetic factors.
Secondly, we performed a cross to test directly whether Map3k4 haploinsufficiency might account for the development of ovarian tissue in XYAKR Thp/+ C57BL/6J individuals. We generated XYAKR mice after backcrossing of YAKR to C57BL/6J for six generations. These males were then crossed with females heterozygous for the targeted null allele of Map3k4 (Map3k4tm1Flv), also on C57BL/6J, to generate XYAKR Map3k4tm1Flv/+ heterozygotes on a C57BL/6J background. Nine of these individuals were generated in five litters and seven were scored as normal males based on examination of the external genitalia. However, two were scored as phenotypic females on the basis of external genitalia morphology. Examination of these sex-reversed individuals revealed the presence of ovaries and uterine horns. Histological examination of the ovaries from one of these individuals showed them to be smaller than controls and lacking clearly discernible follicles or ova (unpublished data). Examination of four other heterozygous males at approximately 11 wk of age revealed that these had testes of reduced size (ranging from 0.03 g to 0.08 g, mean = 0.06 g±0.015), in contrast to wild-type controls (n = 6, ranging from 0.08 g to 0.11 g, mean = 0.093 g±0.009). Small testes are sometimes an indication of earlier ovotestis development. To test this possibility, we performed timed matings in order to examine gonadal morphology in XYAKR Map3k4tm1Flv/+ embryos at 14.5 dpc. Of four XYAKR Map3k4tm1Flv/+ embryos examined, one contained gonads with an overt ovarian morphology, whilst three contained ovotestes identified by morphology and the familiar variegated expression of Sox9 and Wnt4 (Figure 8C). On the basis of the XY gonadal sex reversal, complete and partial, observed in adult and embryonic Map3k4tm1Flv/+ individuals on C57BL/6J-XYAKR, we conclude that haploinsufficiency for Map3k4 is a major contributory factor to male-to-female sex reversal observed in XYAKR C57BL/6J Thp/+ individuals.
Here we describe evidence demonstrating, for the first time to our knowledge, an in vivo role for the phylogenetically ancient MAPK signalling cascade in mammalian sex determination. XY embryos lacking functional MAP3K4 on a predominantly C57BL/6J background exhibit embryonic gonadal sex reversal associated with failure of a number of cellular and molecular events, paramount amongst these being failure to transcriptionally up-regulate Sry and, presumably as a consequence, Sox9 in the developing gonad at 11.5 dpc. Previous studies, often involving analyses of Mus domesticus-derived Sry alleles on a C57BL/6 background, have suggested that the testis determining pathway is exquisitely sensitive to levels and timing of Sry: if a threshold level is not met in a critical time window, ovary development is likely to ensue [69],[70],[77]. Thus, attention is naturally focussed on the possible explanation for reduced Sry expression, at the transcript and protein levels, in XY byg/byg gonads. Three potential explanations exist: (i) that a transcriptional regulator (or regulators) required for transcription of Sry in pre-Sertoli cells does not function appropriately because of, either directly or indirectly, the absence of MAP3K4-mediated signalling; (ii) that insufficient numbers of pre-Sertoli cells are established in the XY byg/byg gonad; (iii) a combination of both of the above effects. With respect to the second hypothesis, the coelomic epithelium is thought to be a source of pre-Sertoli cells in the early XY gonad (prior to 11.5 dpc) [78]. Thus, the reduction in cell proliferation and gonadal growth in the coelomic region of XY byg/byg mutant embryos might be considered evidence of a wider range of defects in the developmental potential of the mutant coelomic epithelium and associated mesenchyme, perhaps extending to a reduction in the provision of pre-Sertoli cells, or the provision of pre-Sertoli cells competent to activate transcription of Sry. This hypothesis is consistent with the active MAPK signalling that we report in the coelomic region at 11.5 dpc in XY gonad. However, it should be noted that in other genetic contexts in which cell proliferation in the coelomic region of the developing XY gonad is disrupted, such as in gonads lacking Fgf9 [10],[61], Sry transcription is reported to be unaffected [56]. Thus, there is no established mechanistic link between prior proliferative defects in the early gonad and subsequent loss of Sry expression. However, given the reported role of FGF9 in promoting gonadal cell proliferation [61], it is possible that loss of MAP3K4 results in an inability of coelomic region cells to efficiently transduce FGF9 signal produced by initial SRY-positive pre-Sertoli cells. This, in turn, would result in failure to establish a positive feedback mechanism by which cell proliferation and SRY expression mutually promote each other, causing insufficient provision of pre-Sertoli cells. This model would explain the reduced numbers of SRY-positive cells detected in XY byg/byg gonads between 11.0 and 11.5 dpc (Figure 5). In order to establish whether there is a paucity of cells migrating into the XY byg/byg gonad at around 11.2–11.4 dpc to populate the pre-Sertoli cell niche, it will be necessary to perform single-cell labelling experiments similar to those used to establish the role of the coelomic epithelium in this process [78]. However, establishing whether a marked cell was undergoing, or had undergone, active MAPK signalling of the appropriate sort would be technically daunting.
With respect to the first hypothesis, little is known about the transcriptional control of Sry, although several potential activators have been described including M33, WT1(+KTS), GATA4/FOG2, and SF1 [79]. This hypothesis is supported by the presence of a few SRY-positive cells in the XY byg/byg gonad at 11.5 dpc that exhibit a significant reduction in the intensity of the SRY signal, and also the existence of FOXL2-positive cells in the XY byg/byg gonad at 11.5 dpc, since this lineage is arguably the ovarian equivalent of the pre-Sertoli cell lineage of the testis. Evidence already exists for MAPK-dependent phosphorylation of SF1 [80],[81] and GATA4 [82] in other contexts, as a means of increasing their transcriptional activation potency. It is also noteworthy that SRY, which is phosphorylated in humans [83], has recently itself been proposed to be a target of p38-mediated signalling pathways on the basis of cell line studies in vitro [75]. We are currently attempting to identify reduced phosphorylation of candidate testis-determining proteins in MAP3K4-deficient embryonic gonads. However, we cannot rule out the possibility that previously uncharacterised molecules are the key effectors of MAPK-mediated events during gonadogenesis. Moreover, MAP3K4-mediated events required for normal Sry transcription may occur in the progenitors of pre-Sertoli cells, in the form of programming, rather than pre-Sertoli cells themselves. In conclusion, the data suggest that the third hypothesis may best explain the observations concerning SRY expression.
The similarity in the phenotypes of mice lacking the Map3k4 gene [44] and those merely lacking a functional kinase domain of the same gene [45], strongly argues that MAP3K4 functions primarily to regulate MAPK signalling through its kinase domain. Thus, although we cannot formally exclude additional functions, we conclude that loss of functional MAP3K4 in the byg mutant results in disrupted MAPK signalling during gonad development. Although ours is the first report of a requirement for MAPK signalling in sex determination in vivo, one previous report has implicated a MAPK scaffolding protein, Vinexin-γ, in regulation of Sox9 transcription during gonad development [84]. However, the fetal gonads of both XX and XY embryos lacking Vinexin-γ are morphologically normal and adult mice of the same genotypes are viable and fertile. Moreover, Sox9 transcript levels in Vinexin-γ −/− XY gonads at 12.5 dpc are 75% that of Vinexin-γ −/+ gonads, suggesting that any modulation of Sox9 transcription by Vinexin-γ is relatively modest. These data appear to be consistent with reported organ culture studies in which the MAPK inhibitor PD98059 did not significantly inhibit testis cord formation in XY gonad explants [74]. In contrast to the Vinexin-γ studies, we observe an almost complete absence of Sox9 at the sex determining stage of gonad development (11.5 dpc) in C57BL/6J XY embryos lacking MAP3K4 and a complete failure of testis cord formation at later stages.
One possible explanation of the apparent discrepancy in these observations with respect to the role of MAPK signalling in testis development is the focus in other studies on the MEK-ERK pathway of MAPK signalling, sometimes called the classical MAPK cascade [73]. It has been proposed that Vinexin-γ mediates its effects on Sox9 transcription in vitro via male-specific activation of the MAPK, ERK [84], and PD98059 is a specific MEK-ERK inhibitor [73],[85]. The focus on MEK-ERK in other studies is likely a consequence of the inviting similarities between requirements for Sox9 up-regulation during gonad development and chondrogenesis. FGF-mediated activation of Sox9 transcription during chondrogenesis has been shown to be blocked by the MAPK inhibitor U0126 [86]. U0126 is also a specific MEK-ERK inhibitor [73],[85]. Given that MAP3K4 is thought to act ultimately by activation of the MAPKs p38 and JNK [42],[43], the focus on ERK activation and the consequences of its disruption as a means of determining the role of MAPK signalling during gonad development may have been overly restrictive and resulted in misleading conclusions. Our studies utilising specific small molecule inhibitors of MAPK signalling in organ culture assays corroborate previous observations that MEK-ERK inhibition does not significantly disrupt Sox9 expression in vitro. However, in contrast, they do suggest a possible role for p38 in gonadal Sox9 transcriptional regulation and testis cord formation. The significance of these in vitro observations for the possible role of p38 in the aberrant phenotype of the MAP3K4-deficient gonad is unclear, given that Sry transcription is already at its peak at 11.5 dpc, the approximate stage at which gonadal explants were employed for in vitro culture experiments. Inhibition of p38 at these stages may disrupt testis-determining events downstream of regulation of Sry transcription, perhaps related to regulation of Sox9 expression, in a manner analogous to that reported for the IL-1β-dependent induction of SOX9 expression in human articular chondrocytes [87], or disruption of SOX9 function itself. Mice constitutively lacking the alpha isoform of p38 die at around 10.5 dpc, before gonadogenesis can be fully examined [88]. For this reason, it is important to remain open-minded about how many distinct steps in testis development require MAPK-dependent events. Teasing these out genetically will require a conditional null allele of Map3k4 (and genes encoding other MAPK signalling elements) and inducible, cell-type-specific Cre lines. It will also be important to determine whether disruption to individual MAP2Ks and MAPKs also results in abnormal gonad development in vivo, or whether loss of MAP3K function is disruptive to a broader range of MAPK signalling events, including potential compensatory ones, and thus more likely to result in phenotypic abnormalities.
In addition to downstream events mediated by MAP3K4, it is not yet clear which upstream signals employ MAP3K4 for their transduction. Analogies with chondrogenesis, as described above, have tended to focus attention on the role of FGF signalling and its use of MAPK for its transduction. Moreover, FGF9 is known to be required for the male-specific elevated proliferation rate in the gonadal coelomic region at around 11.5 dpc [61]. However, we have demonstrated that the ability of exogenous FGF9 to activate Sox9 transcription during gonad culture remains unaltered in the absence of MAP3K4. These data do not definitively demonstrate that FGF9 does not employ MAP3K4-mediated signal transduction during regulation of Sox9 expression during male gonad development in vivo, but they do suggest that such a pathway is not obligatory. Moreover, initial up-regulation of Sox9 transcription, along with Sry transcription, proceeds as normal in embryonic gonads lacking FGF9 [56]. It is, rather, the maintenance phase of Sox9 transcription in developing male gonads that is disrupted in the absence of FGF9. Taken together, these observations suggest that we should look at other pathways, in addition to FGF, for the activating signals that require MAP3K4 for their transduction. Although activation of MAPK is a widespread phenomenon, ligand binding to receptor tyrosine kinases (RTK) is commonly associated with activation of this intracellular signalling cascade [89]. Interestingly, the insulin receptor tyrosine kinase gene family (Ir, Igf1r, and Irr) has previously been shown to be required for testis determination through its regulation of Sry expression [17], and a number of reports describe a requirement for MAPK in signal transduction through this family of receptors in different biological contexts [90],[91]. Similarly, loss of another RTK, PDGFRα, also disrupts testis development [21] and PDGF signalling is reported to employ MAPK [92]. Finally, in addition to RTK activity, prostaglandin D2 (PGD2) has been shown to influence Sertoli cell differentiation and SOX9 activity [39],[57],[58], presumably through its G-protein coupled receptors, DP and CRTH2 [93], although this is not established. PGD2 signalling in other contexts has been shown to require MAPK [94],[95]. Although the details of MAPK activation in these disparate systems vary, they are all potentially relevant to the phenotype of MAP3K4-deficient gonads because evidence suggests cross-talk between distinct MAPK pathways [75].
Despite the above observations, we cannot rule out the possibility of a role for a hitherto unrecognised growth factor or other extracellular signal in the employment of MAP3K4 during testis development. One virtue of invoking a requirement for MAP3K4 in FGF9-mediated signalling during gonadogenesis in vivo is that this model does not predict a requirement for sexually dimorphic expression of MAP3K4, consistent with Map3k4 expression data. We observed near ubiquitous expression of Map3k4, including male and female gonads, although higher levels were detected in particular cell types. Because of a lack of the relevant antibodies, we were unable to assay for the presence of activated MAP3K4 specifically in XY gonads, although such activation is predicted by the existence of MAP4Ks [96]. It should also be noted that the same explanatory virtue applies to invoking a requirement for MAP3K4 in activation of Sry transcription.
We also report here data indicating that haploinsufficiency for Map3k4 is sufficient to account for Tas [40], a phenomenon that has remained unexplained at the molecular level since its discovery more than 20 y ago. XY embryos heterozygous for the Map3k4tm1Flv mutation on the C57BL/6J-YAKR background exhibit testicular abnormalities, including XY ovary and ovotestis development, reminiscent of XYAKR C57BL/6J Thp/+ embryos [40]. Moreover, two adult XY Map3k4tm1Flv/+ individuals developed as phenotypic females, and both contained ovaries. Four others exhibited testicular hypoplasia, which is associated with prior ovotestis development. It is unclear, however, despite the role for Map3k4 haploinsufficiency established here, whether additional testis-determining genes exist in the region deleted in Thp, or whether chromosome deletions themselves predispose XY embryos to sex reversal by inhibitory effects on fetal growth [97].
Significantly, it has been demonstrated that, on the appropriate genetic background, the loss of a single copy of a male-determining gene can result in XY gonadal sex reversal [14]. It has been proposed that such phenotypic effects in mice caused by a single disrupted allele mimic the more common situation in humans, where loss of a single functional copy of genes such as SF1, SOX9, or WT1 can result in the development of XY females [98],[99]. Our findings suggest that the loss of a single copy of Map3k4, caused by the Thp deletion or targeted gene deletion, is another example of such a case. Thus, we propose that haploinsufficiency of MAP3K4 could be the cause of previously unassigned cases of XY gonadal dysgenesis in humans [100]. A second, independent case of Tas on the C57BL/6J XYAKR background (B6-TAS) is caused by the T-Orleans deletion (TOrl), which overlaps with the hairpin tail deletion and also includes Map3k4/D17Rp17 [41]. Interestingly, it has been proposed that B6-TAS in TOrl/+ XYAKR mice is due to biologically insufficient levels of Sry expression [101]. An analogous explanation of the mechanism underlying the Thp/+ XYAKR phenotype is consistent with the report here of delayed and reduced levels of Sry transcription in byg/byg XY gonads at 11.5 dpc. Levels of Sry transcription in XY embryonic gonads of Map3k4byg/+ or Map3k4tm1Flv/+ heterozygotes on the C57BL/6J-YAKR background have yet to be determined, but this experiment will form part of a more extensive analysis of gonadogenesis in these individuals.
Our data have opened a novel entry point into the molecular genetic control of mammalian sex determination and, in particular, the regulation of Sry expression. We know of no other higher organisms in which MAPK signalling is thought to regulate sexual development, although pheromone response during mating in yeast and other fungi is known to require a highly related pathway of kinase activity [102]–[104]. We are currently investigating the role of other proteins required for MAPK signalling in mouse gonad development, utilising in vivo and in vitro methods. The ultimate aim of these studies is to clarify the pathway of MAPK signalling that operates during gonadogenesis and determine precisely how it interacts with the molecular events constituting sex determination. Finally, our study suggests that forward genetic screens in the mouse should be considered as another important tool for identifying vertebrate sex determining genes.
We have previously described the mutagenesis and screening methodology employed here [105]. Briefly, a three-generation (G3) recessive mutagenesis screen was used in which C57BL/6J males were injected with ENU and outcrossed to C3H/HeH females; F1 (founder) males were used to establish pedigrees by mating to C3H/HeH and F2 female offspring were backcrossed to their father. Using this breeding scheme it is expected that approximately one in eight embryos in a pedigree will be homozygous for any given ENU-induced mutation. G3 embryos were examined at 13.5 and 14.5 dpc for developmental abnormalities. Examination of pedigree RECB/31 (byg) revealed several embryos with abnormal male gonad development. Affected embryos were used for genetic mapping with a 55-marker genome wide SNP panel (sequences available on request). byg was maintained by backcrossing to C3H/HeH and, following identification of the Map3k4 mutation, genotyped for the mutant SNP by pyrosequencing. Timed matings were used to generate embryos at specific stages. Breeding pairs were set up at approximately 3 pm and vaginal plugs were checked the following morning. Noon on the day of the plug was counted as 0.5 dpc. Embryos were typed for chromosomal sex as previously described [106].
Genotyping for the byg mutation was performed using a PCR-based pyrosequencing assay using the following primers: Forward PCR primer: 5′-AGGACTATGAACGGTACGC-3′; Reverse PCR primer: 5′-Bio-CGCAGCTTCTGATTTAGATC-3′; Sequencing primer 5′-GCCAAGGACTTTGAGG-3′. byg was backcrossed to C3H/HeH and C57BL/6J. Analysis of byg/byg embryos on C57BL/6J was performed between generations n = 2 to n = 5.
The generation and maintenance of mice lacking Map3k4 has been previously described [44],[107]. Map3k4-deficient mice utilised here were maintained on a C57BL/6J background. Hairpin tail (Thp) mice, originally archived on a mixed genetic background, were rederived using independent in vitro fertilisation (IVF) with both C57BL6/J and C3H/HeH oocytes. Thp was maintained on both C57BL/6J and C3H/HeH. XY sex reversal was observed on the former, but not the latter, genetic background. Thp carriers were identified by the shortened tail [108]. Confirmation of the presence of the AKR-derived Y chromosome was performed by using a PCR assay based on that described in [109], which exploits a Zfy-2 polymorphism between M. domesticus and M. musculus.
WMISH to explanted gonads was performed as previously described [31],[106]. The following probes were used for WMISH: Sox9 [110]; Oct4 [111]; 3β-HSD [112]; Wnt4 (IMAGE clone 40044945), Sry [113], Stra8 (IMAGE clone 40045823), Map3k4 (IMAGE clone 5705378).
Total RNA and protein were extracted from individual 11.5 dpc (17–18 ts) mouse urogenital ridges (comprising gonad and mesonephros) using the Nucleospin RNA/protein isolation kit (MACHEREY-NAGEL) following manufacturer's instructions. The quantity and quality of the RNA was assessed using the Nanodrop ND1000 (Isogen Life Science) and by gel electrophoresis. A two-step real-time analysis approach was taken. First, cDNA was synthesised using the AB High Capacity cDNA Reverse Transcription Kit using 1 µg of total RNA. The following TaqMan assays (Applied Biosystems [AB]) were used: Sf1 (Mm00496060_m1); Fgf9 (Mm00442795_m1); Sry (Mm00441712_s1); Hprt1 (Mm01545399_m1). For each assay, reactions were performed in triplicate using AB Fast Mastermix in a final volume of 20 µl (5 µg of cDNA added). Real-time amplification was performed on an AB 7500 Fast machine, using the manufacturer's recommended program for Fast Mastermix. Analysis of the results was performed using AB software, employing a ddCt method with the gene Hprt1 as the endogenous control. For each assay four biological replicates and three technical replicates were performed. Statistical analysis was performed using a non-paired t-test on the average dCt values calculated for the three technical replicates of each independent sample (biological replicate).
The following antibodies were used in this study: SRY [39]; SOX9 [39]; FGFR2 (Santa Cruz number sc-122); SF1, a kind gift from K. Morohashi; FOXL2: antibodies were raised in rabbits against the peptides MMASYPEPEDAAGAALL and WDHDSKTGALHSRLDL, previously utilised in [114]. Antibodies were affinity purified and tested prior to use: platelet/endothelial cell adhesion molecule (PECAM) (BD Bioscience number 553708); phospho-histone H3 (pHH3, Sigma number HH908 or Upstate number 06-570); phospho-MKK4 (Cell Signalling number 9151); phospho-MKK7 (Cell Signalling number 4171); phospho-p38 (Cell Signalling number 4631); phospho-JNK/SAPK (Cell Signalling number 9251); cleaved caspase-3 (Cell Signalling number 9661S); MAP3K4 (Sigma m7194). Wholemount immunohistochemistry was performed as previously described [106]. Section immunohistochemistry was performed on the basis of protocols described in [39]. Wholemount samples were imaged using a Leica TCS SP5 confocal microscope. Sections were visualised using a Zeiss Axiophot 2.
After immunostaining with anti-PECAM and anti–pHH3 (Upstate, number 06-570) and nuclear counterstaining with aqueous DAPI, the central third of each gonad was imaged using a Leica TCS SP5 confocal microscope (40×). A Z-stack series (10 µm steps) was generated for each sample and then three central sections were chosen for cell counts in the coelomic region (pHH3-positive cells and DAPI-stained nuclei). Sections were separated by 20 µm to ensure that no cell was counted twice. Differences between samples were assessed using a two-tailed t-test.
Culturing of embryonic gonads and recombination experiments between subdissected gonads and marked mesonephroi were performed based on methodologies described in [50] and [106]. Briefly, XY urogenital ridges (UGRs), consisting of gonad and attached mesonephros, were collected at 11.5 dpc (16–19 ts stage) and cultured to establish conditions under which testis cords formed reliably after 48 h culture. Samples were incubated on 1.5% agar blocks at 37°C/5% CO2 in Dulbecco's Minimal Eagle's Medium (DMEM)/10% fetal calf serum (FCS)/50 µg/ml ampicillin/200 mM L-glutamine in the presence of MAPK inhibitors or vehicle control. For recombination cultures, 11.5 dpc XY male UGRs from byg/byg mutant embryos were subdissected into component gonad and mesonephros in PBS. The gonads were recombined with mesonephroi from XY Tg(GFPU)5Nagy/J embryos (ubiquitously expressing GFP) and cultured for 48 h, as above. Migration from the marked mesonephros into the attached gonad was imaged using a Leica TCS SP5 confocal microscope. No migration was observed into control XX gonads during these experiments.
The following MAPK signalling inhibitors were used: SB202190 (p38 inhibitor, Sigma) and U0126 (ERK [Mek1] inhibitor, Sigma). SB202190 was used at a final concentration of 25 µM in culture medium, in line with previously reported in vitro studies employing this inhibitor [115]–[117]. U0126 was also used at a final concentration of 25 µM [86],[118],[119].
To examine the effects of exogenous FGF9 expression in XX gonad development we employed the methodology described in [56]. Briefly, agarose beads (Sigma-Aldrich) were incubated in culture medium containing 50 ng/ml FGF9 protein (R&D Systems), or 0.1% BSA, in a humidified chamber at room temperature for at least 5 h. Beads were then placed adjacent to gonads (n = 3 for each genotypic class) and cultured for approximately 42 h.
Animal procedures employed in this study were authorized by UK Home Office Project License PPL 30/2381.
|
10.1371/journal.pgen.1002050 | Loss-of-Function Mutations in PTPN11 Cause Metachondromatosis, but Not Ollier Disease or Maffucci Syndrome | Metachondromatosis (MC) is a rare, autosomal dominant, incompletely penetrant combined exostosis and enchondromatosis tumor syndrome. MC is clinically distinct from other multiple exostosis or multiple enchondromatosis syndromes and is unlinked to EXT1 and EXT2, the genes responsible for autosomal dominant multiple osteochondromas (MO). To identify a gene for MC, we performed linkage analysis with high-density SNP arrays in a single family, used a targeted array to capture exons and promoter sequences from the linked interval in 16 participants from 11 MC families, and sequenced the captured DNA using high-throughput parallel sequencing technologies. DNA capture and parallel sequencing identified heterozygous putative loss-of-function mutations in PTPN11 in 4 of the 11 families. Sanger sequence analysis of PTPN11 coding regions in a total of 17 MC families identified mutations in 10 of them (5 frameshift, 2 nonsense, and 3 splice-site mutations). Copy number analysis of sequencing reads from a second targeted capture that included the entire PTPN11 gene identified an additional family with a 15 kb deletion spanning exon 7 of PTPN11. Microdissected MC lesions from two patients with PTPN11 mutations demonstrated loss-of-heterozygosity for the wild-type allele. We next sequenced PTPN11 in DNA samples from 54 patients with the multiple enchondromatosis disorders Ollier disease or Maffucci syndrome, but found no coding sequence PTPN11 mutations. We conclude that heterozygous loss-of-function mutations in PTPN11 are a frequent cause of MC, that lesions in patients with MC appear to arise following a “second hit,” that MC may be locus heterogeneous since 1 familial and 5 sporadically occurring cases lacked obvious disease-causing PTPN11 mutations, and that PTPN11 mutations are not a common cause of Ollier disease or Maffucci syndrome.
| Children with cartilage tumor syndromes form multiple tumors of cartilage next to joints. These tumors can occur inside the bones, as with Ollier disease and Maffuci syndrome, or on the surface of bones, as in the Multiple Osteochondroma syndrome (MO). In a hybrid syndrome, called metachondromatosis (MC), patients develop tumors both on and within bones. Only the genes causing MO are known. Since MC is inherited, we studied genetic markers in an affected family and found a region of the genome, encompassing 100 genes, always passed on to affected members. Using a recently developed method, we captured and sequenced all 100 genes in multiple families and found mutations in one gene, PTPN11, in 11 of 17 families. Patients with MC have one mutant copy of PTPN11 from their affected parent and one normal copy from their unaffected parent in all cells. We found that the normal copy is additionally lost in cartilage cells that form tumors, giving rise to cells without PTPN11. Mutations in PTPN11 were not found in other cartilage tumor syndromes, including Ollier disease and Maffucci syndrome. We are currently working to understand how loss of PTPN11 in cartilage cells causes tumors to form.
| Cartilage tumor syndromes are characterized by multiple cartilaginous bone tumors that develop in childhood, often causing significant morbidity and predisposing to chondrosarcoma. Tumors can form as exostoses (on the surface of bone), as in the autosomal dominant, multiple osteochondroma (hereditary multiple exostoses) syndromes (MO; MIM 133700 and 133701), or as endosteal tumors (within bone), as in the sporadically occurring multiple enchondromatosis disorders (MIM 166000) Ollier disease and Maffucci syndrome. In MO, mutations in EXT1 or EXT2, which encode heparan sulfate glycosyltransferases, affect chondrocyte orientation in the growth plate [1]. A small percentage of patients with Ollier syndrome have mutations in PTH1R, which encodes the receptor for parathyroid hormone and parathyroid hormone-related protein, causing altered chondrocyte differentiation in the growth plate [2]. The cause of Maffucci syndrome is unknown [3]. Patients with MO do not develop endosteal tumors, and patients with Ollier disease or Maffucci syndrome do not develop exostotic tumors [1], [3], [4].
Patients with metachondromatosis (MC; MIM 156250) form exostotic and endosteal tumors (Figure 1). Fewer than 50 cases of MC have been published since Maroteaux's initial description in 1971 [5]. Exostotic lesions in MC occur frequently in the digits, involve metaphyses and epiphyses, and tend to grow toward the joint; in contrast, exostotic lesions in MO occur frequently in the long bones, involve only the metaphyses, and tend to grow away from the joint [6]–[11]. MC exostotic lesions can also spontaneously decrease in size and completely regress [6], [7], [9], [12]. Endosteal lesions in MC are common in the metaphyses of long bones and in the pelvis [7]–[11]. Avascular necrosis of the femoral head, due to endosteal tumors, has been a frequent complication in patients with MC [7], [8], [13]–[15]. Hand deformity due to endosteal tumors is uncommon in patients with MC, whereas it is often a significant problem for patients with Ollier disease and Maffucci syndrome [3]. Finally, malignant transformation has only been reported in one patient with MC, whereas it has been more frequently reported in patients with MO, Ollier disease, and Maffucci syndrome [3], [4], [16]. The distinct distribution and clinical behavior of lesions in patients with MC, suggest that MC is pathophysiologically distinct from these other cartilage tumor syndromes. We therefore sought to better characterize MC and to determine its genetic basis.
We diagnosed participants as having MC based upon the presence of both multiple exostotic and endosteal cartilaginous lesions as previously described [5]–[10], [15]. We excluded from analysis participants with solitary lesions, contiguous endosteal lesions suggestive of Ollier disease, soft tissue lesions suggestive of Maffucci syndrome, or radiographs suggestive of MO. We included participants who had clinical and radiographic features of MC, even if they lacked a positive family history. For each patient, the clinical history and radiographs were reviewed by at least 3 authors. MC patients from 17 unrelated families from 9 countries were identified (Supplementary Table 1). All participants gave their informed consent following the guidelines of each referring institution. In 10 families disease segregation is consistent with autosomal dominant inheritance. In 7 families, the disease is suspected to have arisen de novo. Immediate family members of patients with sporadically occurring MC were interviewed and examined, although detailed imaging was not performed. For 8 familial cases, blood or DNA was available from additional family members.
Figure 1 depicts features seen in affected participants with MC. No phenotypic differences could be found between sporadic or familial cases of MC. Radiographs identify exostotic and endosteal lesions of the digits (Figure 1A–1C) and long bones (Figure 1E–1H), along with degenerative hip disease secondary to endosteal lesions in the femoral neck (Figure 1I). Spontaneous regression of exostotic lesions is seen in radiographs obtained 10 years apart in the same patient (Figure 1F,1G). Also depicted in Figure 1 are histopathologic features that distinguish exostoses in patients with MO from those in patients with MC, based upon a comparison of 30 exostoses excised from children with MO and 15 exostotic lesions excised from 3 affected individuals with MC. Exostoses in children with MO have cartilage caps with endochondral bone growth immediately beneath the cap (Figure 1J). In contrast, exostoses in children with MC have a predominantly fibrous cap and a core of disorganized cartilage surrounded by trabecular bone (Figure 1K, 1L). In all MC cases, the lesions were bilateral and not obviously confined to a single body segment as in Ollier or Maffucci patients.
We performed linkage analysis in the largest family (Family A, Figure 2A) to identify a genetic locus for MC. Raw genotype data were generated using Affymetrix 6.0 SNP arrays and multipoint parametric linkage analysis of the autosomal genome was performed using MERLIN [17]. Because non-penetrance and non-ascertainment are potential confounding factors in the diagnosis of MC, we analyzed only founders and affected individuals (Figure 2A). Although this limited the maximum attainable LOD score to 2.7, which is lower than the genome-wide significance threshold of 3.3, the dense marker set ensured a reasonable probability that only one large interval that achieved the maximum LOD score would be observed, with the remainder of the genome being excluded.
We identified a single interval on chromosome 12, from 121.5 to 132.3 cM, that attained the maximum LOD score of 2.7 (Figure 2B). No other autosomal interval >1 cM yielded a peak LOD score >−1.9. Several intervals <1 cM attained LOD scores >0 (Figure S1), but we considered these unlikely to be candidate intervals and instead assumed they represented either unfiltered genotyping errors or short ancestral population haplotypes, rather than familial haplotypes inherited from a common ancestor.
We next performed array-based capture, followed by Illumina GAII sequencing, using bar-coded DNA libraries created from 16 individuals in 11 families. We sequenced 1 affected individual in 8 families, 2 affected individuals in 1 family, and 2 affected and 1 unaffected individual in 2 families (Table S1). We prepared bar-coded genomic DNA libraries, having an average insert size of 150 bp, using sheared DNA from each of the 16 individuals (Figure S2). We performed array-based targeted capture by pooling each DNA library and hybridizing the pooled sample to an Agilent Technologies 1M SureSelect DNA capture array containing 973,952 probes targeting 844,339 bp within the 8.6 Mb candidate interval, including 88.4% and 98.6% of UCSC exons and CCDS coding sequence, respectively. After hybridization and elution, the captured DNA was PCR amplified, purified to remove primer dimers, and sequenced on two lanes of an Illumina GA II.
We obtained 50 million, 80 bp, single-end reads. Novobarcode software was used to sort the reads according to their 3 bp barcode, and Novoalign was used to align the reads to the reference genome (hg19). We obtained between 1 and 6 million reads for each individual. Among individuals, an average of 61% (±2%) of the aligned reads mapped to regions targeted by the capture array (Figure S3A). Of the bases targeted by the capture array, 75% (±7%) had a read depth of at least 5×, which diminished to 55% (±14%) after the removal of PCR duplicates (Figure S3B). Twenty percent of targeted bases were not captured. In the 3 families for which pairs of affected family members were sequenced, total filtered variants in the candidate interval (388–1499 per individual) were analyzed to find variants shared by both affected individuals from the same family (Table 1). In all 3 families, frameshift mutations in exon 4 of PTPN11 were the only novel coding variants present in both affected family members and, for Families A and B, absent in the unaffected individual. Family A had a 5 bp deletion, Family B had a more complex deletion/insertion, and Family C had a 2 bp deletion (Table S1). In the remaining 8 families for which only 1 affected individual per family was sequenced, there were 18 novel coding variants present in ≥3 reads, one of which was a nonsense mutation in exon 13 of PTPN11 (p.Q506X) (Figure S4). We used Sanger sequence analysis of PCR amplimers to demonstrate that affected family members from these 4 families had PTPN11 mutations, and that unaffected family members lacked PTPN11 mutations.
Sanger sequence analysis of the 15 coding exons of PTPN11 in the 7 families for whom we had not found mutations by array capture and Illumina-sequencing detected a 1 bp deletion in exon 11 in 1 of the 7 families (Family D). This deletion was within a 98 bp segment that had been targeted but not captured in any of the DNA samples. Another family (F) had a splice-acceptor site mutation (AG>CG) in intron 5 in 2 affected siblings, but not in either parent. The siblings' mother was clinically affected with MC, although less severely than her children. The mother was the first in the family to have MC and was the only member of the family who was included in the Illumina sequencing. The site of the splice-site mutation identified in her children was covered 25× in her DNA sequence and was always wild-type, as were her Sanger sequence results. These data suggest the mother is mosaic for a PTPN11 mutation and that the family's mutation would have been found by Illumina sequencing had we initially sequenced her children's DNA.
We subsequently collected DNA from an additional 6 MC families. Sanger sequence analysis revealed a nonsense mutation involving exon 3 (p.K99X) in one family (I), and a splice site mutation in intron 9 (c.1093-1G>T) in another family (G). In total, we found PTPN11 mutations in 10 of 17 families. Five mutations were frameshift, 2 were nonsense, and 3 disrupted a splice-acceptor site (Figure 3A, Table S1). Each family had a different mutation and mutations were scattered across the gene (Figure 3A). In two families without mutations we had performed aCGH and did not find evidence of PTPN11 intragenic deletions or duplications (Table S1). We did not have other family members' DNA samples from the one familial MC patient who lacked a PTPN11 mutation to be able to test for locus heterogeneity by linkage analysis.
To test whether the patients without PTPN11 coding mutations had noncoding mutations in PTPN11 or had mutations in other genes, we designed a second Agilent 1M capture array. Firstly, we included probes to target the entire PTPN11 gene, excluding Alu repeats. Secondly, we included probes targeting the exons of 74 genes that function in the same pathways as PTPN11, including the Ras/MAPK and PI3K/Akt pathways (Table S2). Thirdly, we included probes targeting the exons of the MO genes EXT1 and EXT2, to determine whether any of our patients lacking mutations and classic radiographic features of MC had been misdiagnosed. Barcoded genomic libraries for an individual from each of the 7 MC families without PTPN11 coding mutations and from 2 families originally referred with MC, but whose radiographic features were more consistent with MO, were pooled and hybridized to the capture array. The captured DNA was then sequenced using two lanes of Illumina GAII 42 bp single end sequencing. For each barcoded sample, 4.3 (±1.2) million reads were obtained, of which 37% (±7%) mapped to regions targeted by the array (Figure S3A). Of the bases targeted by the array, 85% (±6%) were covered by a read depth of at least 10×, which dropped to 83% (±8%) after the removal of PCR duplicate reads (Figure S3B).
Identified variants with a quality score >20 were filtered to remove SNPs listed in the SNP database (version 132) and the 1000 genomes project (Nov. 2010 release). No PTPN11 coding mutations were found. We then analyzed the noncoding regions and identified one 3′ UTR mutation, and 7 intronic mutations, of which 6 were in LINE elements or other repetitive regions (Table S3). All intronic mutations were at least 300 bp from the nearest exon and, using an online splice prediction tool (http://genes.mit.edu/GENSCAN.html), were not predicted to alter splicing. We then analyzed variants identified in the exons of the 76 other genes included in the capture array (Table S2). No nonsense, frameshift or splice site mutations were identified. Of the missense mutations that were present in more than 4 independent sequencing reads, 4 were nonsynonomous (ERBB2 p.S1050L, MTOR p.P1408S, MVP p.R49S, SOS2 p.D952N) and 4 were synonomous (PIK3C2B p.D478D, RAF1 p.L351L, MAP2K2 p.D140D, MVP p.T199T). Further experiments will be needed to determine if any of the novel noncoding PTPN11 mutations or novel variants in the other genes are disease causing.
We next analyzed the sequencing read depth across the PTPN11 locus to detect deletions or duplications. In one individual (Patient S), we identified an ∼15 kb region spanning exon 7 that contained half as many reads as would be expected based upon the read depths of the other patients included in the capture array (Figure 3B). As expected, PCR primers that flank this 15 kb region failed to produce amplimers when wild-type genomic DNA was used as template. However, PCR amplification using genomic DNA from Patient S yielded an ∼700 bp PCR product and Sanger sequence analysis of this product indicated that 14,629 bp of genomic sequence (chr12:112,897,487–112,912,115) had been replaced with a single CA dinucleotide (Figure 3B). In addition, PCR amplification and sequencing of PTPN11 in peripheral blood cDNA from this patient, using a forward primer in exon 6 and a reverse primer in exon 8, detected a mutant cDNA that lacked exon 7 (data not shown). The loss of exon 7 results in a frameshift with introduction of a premature stop codon (T253LfsX54).
The read depth of the remaining 76 genes targeted by the array was also analyzed to detect deletions or duplications. Two patients initially included in the study, but on radiographic review were felt more likely to have MO than MC, were found to have deletions involving EXT1 (Figure S5). In one patient (Q), the first exon of EXT1 contained half as many reads over its 1.8 kb as expected (Figure S5A). In a second patient (N), all exons of EXT1 had half as many reads as expected (Figure S5A). In additional to skeletal lesions, this patient has developmental delay, microcephaly and mild dysmorphism, suggesting a possible contiguous gene deletion syndrome. EXT1 deletions were confirmed in both patients by multiplex ligation-dependent probe amplification (MLPA) (Figure S5B, S5C).
Our finding of nonsense, frameshift, and splice-site mutations in multiple exons, as well as a large deletion, suggests that MC-causing PTPN11 alleles are loss-of-function. We tested this hypothesis by performing Western blots on whole protein extracts from white blood cells and from an excised exostotic lesion in a patient (B-IV-7) with a PTPN11 frameshift mutation in exon 4. An anti-SHP2 antibody that recognizes an epitope amino-terminal of the polypeptide encoded by the frameshifted exon detected only full-length, wild-type SHP2 protein (Figure S6).
We next determined whether MC exostoses arise from a “second hit,” similar to what has been observed in autosomal dominant MO [18]. We looked for a second hit in cells of the cartilage core of an MC lesion (e.g., Figure 1K) by performing microdissection, PCR amplifying the mutation containing exon, and Sanger sequencing the amplimers. In tumors from two different patients (A-IV-5, A-IV-8), with a 5 bp frameshift mutation in exon 4, we observed a clear excess of mutant sequence versus wild-type sequence in the tumors' cartilage cores, as compared to the patients' peripheral blood and bone/marrow from the lesion (Figure 4A). We quantified the amount of mutant versus wild-type sequence, by extracting DNA from the cartilage core of patient A-IV-8, PCR amplifying exon 4, and subcloning amplimers to determine the percent that contained the mutant allele. Forty-four of 52 individual subclones contained the mutant allele, which is significantly higher (p<0.001) than expected for a heterozygous mutation. In contrast, 58% of subclones (34/59) from adjacent unaffected bone/bone marrow contained the mutant allele, which is not significantly different from the expected value of 50% (p = 0.24). These data are consistent with an MC exostosis arising from a second hit (loss of the wild-type allele) within a cell that ultimately contributes to the lesion's cartilage core.
Because the mutant allele is 5 bp shorter than wild-type PTPN11 in these two patients, we tested for loss of heterozygosity at a second polymorphic site in PTPN11 to control for potential PCR bias in amplifying the exon with the deletion. In their peripheral blood DNA, patient A-IV-5 and her unaffected mother are heterozygous for a benign polymorphism in intron 11 of PTPN11 (rs41279092). The abundance of this SNP, which is in the wild-type PTPN11 allele, was markedly reduced in the lesion's cartilage core, again consistent with LOH occurring in the cell that drives formation of the cartilage core (Figure 4B).
We finally asked whether mutations in PTPN11 are associated with other cartilaginous tumor syndromes. We sequenced the coding exons of PTPN11 in 38 lesions excised from patients with Ollier disease, 2 peripheral blood samples from patients with Ollier disease, 15 lesions excised from patients with Maffucci syndrome, 4 solitary enchondromas, 9 chondrosarcomas (1 polyostotic), and 3 osteochondromas without EXT1 or EXT2 mutations. We did not find PTPN11 coding sequence mutations in any patient sample. In 24 percent of the samples we observed heterozygosity for noncoding SNPs that are known common variants, suggesting that large PTPN11 gene deletions and other causes of LOH are not frequently associated with these other cartilaginous tumors.
We identified 17 unrelated families with MC. Clinical features were similar to previously published cases [5]–[10], [15]. The exostoses of MC had been assumed to be identical to the osteochondromas of MO; however, we demonstrate that they are histologically unique lesions with a large cartilaginous core (Figure 1J–1L). We combined linkage analysis in a single MC family with DNA capture and parallel sequencing of bar-coded DNAs from several MC families to identify mutations in PTPN11 as a cause of MC. In MC patients without PTPN11 coding sequence and splice site mutations, we generated and pooled barcoded DNAs, and performed a second targeted capture that included the entire PTPN11 gene and exons from 76 other genes. This led us to detect an ∼15 kb deletion in a patient by analyzing the depth of sequencing reads (Figure 3B). In total, we found likely disease-causing PTPN11 mutations in 11 of 17 families.
Concurrent with our studies of MC, Sobreira et al. (2010) reported PTPN11 mutations in 2 MC families [11]. They performed whole-genome sequencing (WGS) in a single affected individual who was a member of family in which MC was segregating. This approach also required these investigators to include linkage data to reduce the number of novel potentially disease-causing heterozygous changes that are identified by WGS [19], [20]. The investigators next identified an independently arising PTPN11 mutation in an unrelated patient to strengthen the evidence for causality. However, having only studied genomic DNA and finding frameshift mutations in the same exon (exon 4) in their two unrelated patients, Sobreira et al. (2010) could not definitively determine the mechanism by which PTPN11 mutations cause MC.
Missense mutations in PTPN11 have previously been identified in patients with Noonan, Noonan-like, and LEOPARD syndromes, as well as in juvenile myelomonocytic leukemia [21]. In these disorders, the mutations are gain-of-function and/or dominant negative for SHP2, which is the PTPN11 protein product [22], [23]. SHP2 is a protein tyrosine phosphatase and an important intracellular signaling molecule linking several growth factor receptors to the Ras/MAPK and other signaling pathways (Reviewed in [24]). Therefore, frameshift mutations in exon 4 might also create an abnormal protein product by altering PTPN11 mRNA splicing. Alternatively, the frameshift mutations might result in loss-of-function because of nonsense mediated mRNA decay or rapid degradation of a truncated SHP2 polypeptide. Our finding of nonsense, frameshift, and splice-site mutations in multiple exons, as well as a whole-exon deletion, suggests that MC-causing PTPN11 alleles are loss-of-function. We tested this hypothesis by performing Western blots on whole protein extracts from white blood cells and from an excised lesion containing affected and unaffected tissue, and detected only full-length, wild-type SHP2 protein (Figure S6), confirming that the mutant alleles are loss-of-function.
Exostoses in MO originate from the “second hit” mutations [18]. Mice with conditional alleles at the EXT1 locus demonstrate that only a few cells having two mutant alleles are sufficient to cause other cells to become misdirected and form an exostosis [25]. By performing microdissection, we found evidence for loss of the wild-type PTPN11 alleles in the majority of cells within the cartilage cores of exostoses from two MC patients (Figure 4), consistent with a “second hit.” Recently, Bauler et al. used a ubiquitously expressed Ert2-Cre driver in 6–8 week-old Ptpn11 floxed mice to generate mice that were Ptpn11-null in multiple tissues. Among the consequences of completely deleting SHP2 was the appearance of ectopic cartilage islands in the animals' metaphyseal trabecular bone and growth plates [26]. These findings are consistent with the distribution of endosteal tumors and exostoses seen in patients with MC. The findings in mice with homozygous deletion of Ptpn11 contrast with the absence of skeletal lesions in mice that have heterozygous loss-of-function mutations [27]. We suspect that mice with heterozygous mutations have a much lower incidence of noticeable “second hits” compared to humans because they have fewer skeletal cells and shorter lifespans. Homozygous inactivation of Ptpn11 solely in mouse chondrocytes may be required to enable a detailed understanding of how SHP2 deficiency leads to tumorigenesis.
We did not detect PTPN11 mutations in 6 of 17 patients with MC phenotypes, including 1 patient with a family history of MC and 5 patients who are the first affected members in their families. DNA is not available from other affected family members of the familial case to determine whether MC exhibits locus heterogeneity. Two patients with de novo disease did have DNA variants found in the 3′ UTR and/or in introns. None of these variants are in likely regulatory regions or in regions important for mRNA splicing; however, we cannot conclude they are benign. Furthermore, we cannot exclude the possibility that patients with de novo disease are somatic cell mosaics for PTPN11 mutations that are not present in white blood cell DNA, similar to the mildly affected mother in Family F who had two affected children. Despite these caveats, MC could be locus heterogeneous, similar to Noonan syndrome, which can be caused by mutations in PTPN11 or in other components of the Ras/MAPK pathway [28]; however, our targeted capture and sequencing of 74 genes that included most of the Ras/MAPK and PI3K/Akt signaling pathways did not find an obvious mutation in another gene in any of the 6 PTPN11 mutation-negative MC patients.
We found no evidence of PTPN11 coding mutations in other cartilage tumor syndromes, including Ollier disease and Mafucci syndrome. Although sequencing was performed on lesional tissue rather than whole blood, it is possible that we may have missed causative mutations that are present in only a subset of cells within the lesion. We may also have missed mutations in the 5′ and 3′ untranslated regions of PTPN11 contained within exons 1, 15, and 16, that we did not sequence in these patients. Based on our finding heterozygosity for noncoding SNPs in many of these samples, it is unlikely that large PTPN11 gene deletions or other causes of LOH are common in these syndromes. Despite the aforementioned limitations of our mutation detection method, our data are consistent with the separation of MC from the other cartilage tumor syndromes based on clinical and pathologic features.
In conclusion, we combined linkage analysis in a single family with DNA capture and parallel sequencing of bar-coded DNAs from several families to identify mutations in PTPN11 as a cause of MC. The advantages of this approach are its ability to identify a region of interest, then simultaneously sequence affected individuals from multiple unrelated families, and then focus on genes for which novel SNPs or other mutations are seen in more than one family, all at reasonable cost (∼$10,000 in consumables). In patients with MC and PTPN11 mutations, we conclude that the mutations are loss-of-function since the mutant protein is not expressed, and that the loss of the remaining wild-type allele via a “second hit” is responsible for the formation of the exostoses. Since we did not detect PTPN11 mutations in all MC families, MC may be locus heterogeneous, although we have not found evidence after sequencing more than 70 genes that function in related pathways. Finally, precisely how mutations in PTPN11 give rise to the exostoses and endosteal tumors in patients with MC is not yet known. However, this question can now be addressed since mice with alleles of Ptpn11 that can be conditionally inactivated in temporal and site-specific manner are available [26], [29].
Informed consent was obtained through a Children's Hospital Boston IRB approved protocol. Specimens and/or DNA received from external institutions were collected under IRB approved protocols at host institutions and received coded without identifying information.
Raw genotype data were generated for multiple members of family A using Affymetrix 6.0 SNP arrays, and genotypes were called using Affymetrix Genotyping Console with the Birdseed v2 algorithm and a confidence threshold of 0.02. SNPs with <100% sample call rate or pedigree minor allele frequency of 0 were removed, then multipoint parametric linkage analysis of the remaining 421,922 autosomal and 17,169 X-linked SNPs was performed using MERLIN and its derivative MINX, respectively, with Affymetrix Caucasian allele frequencies and deCODE Genetics genetic map positions. The disease allele frequency was estimated at 1E-7, and phenocopies and non-penetrance were not permitted (affectation probability 0/1/1). Because non-penetrance and non-ascertainment are potential confounding factors in MC, we analyzed only founders and affected individuals.
To generate genomic libraries for each individual, 2 µg of genomic DNA were first sheared to ∼100 bp–200 bp using Adaptive Focused Acoustics following the manufacturer's protocol (Covaris, Inc). Blunt-ended fragments were generated using an End-it DNA End-Repair kit (Epicenter), purified using Agencourt AMPure XP magnetic beads (Beckman Coulter), and eluted in 10 mM Tris Acetate, pH 8.0. The fragments were A-tailed using the Klenow fragment (NEB), purified, eluted in 1× Quick Ligase Buffer (Quick Ligation kit, NEB), and incubated with Quick T4 DNA ligase and 100 µM barcoded-adapters (Table S4) to create a library of adapter-ligated fragments. A different barcoded adapter was used for each genomic DNA library. Each library was again purified using Agencourt AMPure XP magnetic beads and eluted in 40 µl of 10 mM Tris Acetate, pH 8.0. Libraries used for hybridization to the first capture array were amplified according to two strategies: 3 µl amplified for 18 cycles in four 50 µl PCR reactions (Phusion High-Fidelity DNA polymerase, Finnzymes), or 2 µl amplified for 11 cycles in one 50 µl PCR reaction that was then purified and amplified for 17 cycles in ten 50 µl PCR reactions (FastStart Taq DNA polymerase, Roche). For the libraries used for hybridization to the second capture array, 13 µl was amplified for 15 cycles in ten 50 µl PCR reactions (Phusion High-Fidelity DNA polymerase, Finnzymes). Primers are provided in Table S5. Sizes of amplified libraries were confirmed to be between 200–300 bp necessary for Illumina GA II sequencing prior to hybridization (Figure S2).
To enrich regions of interest in the linked interval for sequencing, we used an Agilent Technologies 1M SureSelect DNA capture array. Target regions were defined using the UCSC Genome Browser and included: the union of exons from multiple GRCh37/hg19 gene, mRNA, and Alt Events tracks; 30 bp of proximal and distal intronic flanking sequence; and 1000 bp of upstream promoter sequence. Targets were padded with 60 bp of additional proximal and distal flanking sequence to promote uniform capture coverage, for a total size of 1,187,477 bp. Probes were designed against NCBI36/hg18 using the Agilent eArray software (https://earray.chem.agilent.com/earray/) and translated coordinates, with 60-nt length, 3-nt spacing, and repetitive elements masked. The resulting 243,488 probes spanned 844,339 bp (GRCh37/hg19), and included 71.1%, 72.2%, 88.4%, and 98.6% of the padded target, unpadded target, UCSC exons, and CCDS coding sequence, respectively. The probes and their reverse complements were each applied in duplicate to the capture array for a total of 973,952 probes.
For the second 1M SureSelect DNA capture array, Biomart (http://uswest.ensembl.org/biomart/) was used to obtain the Ensembl NCBI37/hg19 coordinates for the exons of 76 genes (Table S2). Exons were padded with 90 bp to define a 718,566 bp target region. eArray was used to design 568,634 probes to target the repeat masked sequences of this region (91%). The target region for PTPN11 was defined as 93,180 bp spanning 1 kb upstream to 2 kb downstream of the gene. Repeat masker (http://www.repeatmasker.org/) was used to mask only Alu repeats (37% of the region) resulting in a target region of 61,365 bp, for which 55,205 probes were designed using eArray. All probes were 60-nt in length and spaced every 1-nt. The capture array was designed to include all probes (623,839 total), as well as the reverse complement of every 2nd probe and every 17th probe, for a total of 972,455 probes.
For the first capture array, 1.4 µg of each of the 16 amplified libraries was pooled and hybridized to the array following Agilent's SureSelect DNA Capture Array protocol version 1.0. Different blocking oligonucleotides (Table S5) were added to the hybridization. After elution from the array, half of the captured library was amplified in five 50 µl PCR reactions for 18 cycles using Phusion High-Fidelity DNA polymerase (Finnzymes) and post-capture primer pair (Table S5), purified using an E-Gel CloneWell (Invitrogen) to remove primer dimers, and re-amplified using fifteen PCR cycles with the same primer pair. The amplified library was again purified to eliminate primer dimers using E-Gel. For the second capture array, 2 µg of each of the 12 amplified libraries was pooled and hybridized to the array. After elution, half of the captured library was amplified in five 50 µl PCR reactions for 15 cycles and purified using Agencourt AMPure XP magnetic beads. Further details of the array design and methods for sequence analysis are provided in the supporting information. Additional methods for Illumina data analysis, copy number analysis, Sanger sequence analysis of PTPN11 (Table S6), DNA extraction from lesional tissue, PCR product subcloning experiments, aCGH analysis, MLPA (Table S7), and immunodetection of SHP2 are also provided in the supporting information (Text S1).
|
10.1371/journal.ppat.1000797 | Fine-Tuning Translation Kinetics Selection as the Driving Force of Codon Usage Bias in the Hepatitis A Virus Capsid | Hepatitis A virus (HAV), the prototype of genus Hepatovirus, has several unique biological characteristics that distinguish it from other members of the Picornaviridae family. Among these, the need for an intact eIF4G factor for the initiation of translation results in an inability to shut down host protein synthesis by a mechanism similar to that of other picornaviruses. Consequently, HAV must inefficiently compete for the cellular translational machinery and this may explain its poor growth in cell culture. In this context of virus/cell competition, HAV has strategically adopted a naturally highly deoptimized codon usage with respect to that of its cellular host. With the aim to optimize its codon usage the virus was adapted to propagate in cells with impaired protein synthesis, in order to make tRNA pools more available for the virus. A significant loss of fitness was the immediate response to the adaptation process that was, however, later on recovered and more associated to a re-deoptimization rather than to an optimization of the codon usage specifically in the capsid coding region. These results exclude translation selection and instead suggest fine-tuning translation kinetics selection as the underlying mechanism of the codon usage bias in this specific genome region. Additionally, the results provide clear evidence of the Red Queen dynamics of evolution since the virus has very much evolved to re-adapt its codon usage to the environmental cellular changing conditions in order to recover the original fitness.
| Each organism has a specific codon usage signature. Translational selection i.e., selection for the codon adaptation to the tRNA pools, is one of the driving forces of codon bias. In the virus world, this implies an adjustment of the virus codon usage to that of the host cell. Hepatitis A virus appears as an exception to the rule, with a highly deoptimized codon usage, suggesting that translational selection is not the underlying mechanism of its codon bias. However, since the virus lacks a mechanism of cellular protein synthesis inhibition, the deoptimized codon usage may be envisaged as a hawk (cell) and dove (hepatitis A virus) competition strategy for tRNAs and translational selection as well. To confirm this possibility, we artificially induced cell protein synthesis shut-off, thus increasing the tRNA pool availability for the virus, and we took advantage of the quasispecies dynamics to elucidate changes in its codon usage. Virus adaptation to the drug results in a re-deoptimization of codon usage in the capsid region, suggesting a requirement of a slow translation rate, i.e., a translation kinetic selection, instead of a translational selection associated with an optimization of the codon usage. Translation kinetics control is based on the right combination of codons (common and rare) that allows a regulated ribosome traffic rate ensuring the proper protein folding. Capsid folding is critical for a virus transmitted through the fecal-oral route with long extracorporeal periods.
| Non-random usage of synonymous codons is a widespread phenomenon observed in genomes from many species in all domains of life and it has been proposed that each genome has a specific codon usage signature that reflects particular evolutionary forces acting within that genome [1],[2]. Such codon usage biases may result from mutational biases, from natural selection acting on silent changes in the genomes or both. Selection on codon usage due to codon adaptation to the tRNA pool was clearly demonstrated in Escherichia coli [3]. Later on, this premise was also confirmed in eukaryotic organisms including vertebrates [4]–[8], supporting the hypothesis of translational selection as a driving evolutionary force. However, this hypothesis is not extensive to all organisms and in humans there is no clear evidence of translation selection as the single driving force of codon bias [9]. In this latter case many different reasons have been identified to explain codon bias: isochoric structure of GC content [10], mRNA secondary structure selection [11], exonic splicing enhancer selection [12],[13], and last but not least translation selection [14],[15]. Additionally, there is also evidence of pressure on codon usage for the control of translation kinetics rather than for the efficiency and accuracy of translation [16],[17]. Translation kinetics control is exerted through the use of common and rare codons, which affect the rate of ribosome traffic on the mRNA due to the longer time required for incorporating those tRNAs pairing with rare codons into the ribosome A-site. The right combination of codons allows a regulated ribosome traffic rate that temporally separates protein folding events, ensuring “beneficial” and avoiding “unwanted” interactions within the growing peptide [18]. This kind of selection, known as fine-tuning translation selection, differs from translation selection in that preferred codons are not always advantageous if the optimal folding requires a slow translation.
A statistical model for translational selection measure was developed based on fully sequenced genomes from archaea, bacteria and eukarya [19]. This model shows that small genome size allows higher action of translation selection, whilst lack of tRNA gene redundancy accounts for the absence of translationally selected codons. Bearing in mind this model, it seems reasonable that translation selection might play a key role on eukaryotic viruses, which present tiny genomes and have huge numbers of tRNA genes available from their hosts. Certainly codon bias has been observed in several viruses, however, its driving force has only been studied in particular cases, such as Epstein-Barr Virus [20] and papillomavirus [21] among DNA viruses, and poliovirus (PV) [22] and hepatitis A virus (HAV) [23] among RNA viruses. While translation selection seems to be the underlying mechanism of the codon bias of those genes expressed during the productive phase in the DNA viruses, the mechanism in the RNA viruses is more variable. The codon usage bias in the capsid of PV, which presents a highly optimized codon usage with that of its host, has been proposed to be mostly the result of an additional step on translation selection, i.e., the effect of certain codon-pair combinations on the rate of translation [24] rather than only the tRNA availability. This may be due to the fact that translational step times are influenced by the compatibilities of adjacent tRNA isoacceptor molecules on the surface of a translating ribosome [25]. Alternatively, it has also been proposed the GC dinucleotide content as the cause of codon bias in the capsid of PV [22].
However, HAV is clearly the exception to the rule. HAV presents a highly biased codon usage and mostly opposed to that of the host cell [23]. Despite lacking mechanisms of inducing cellular shutoff [26],[27] and having a very inefficient IRES [28], HAV is able to synthesize its proteins by adapting their codon usage to those less commonly used cellular tRNAs, resulting in a low replication rate [29]. With that naturally deoptimized codon usage the role of translation selection in shaping the HAV codon bias does not seem very obvious.
HAV is the type species of the genus Hepatovirus within the family Picornaviridae, existing as a single serotype. The occurrence of highly conserved clusters of rare codons in the HAV capsid-coding region has been related to the low antigenic variability [30], since mutations in these clusters are negatively selected even in the presence of immune pressure. Thus, a certain beneficial role of these rare codons is envisaged. Altogether, it can be concluded that codon usage plays a key role in HAV replication and evolution.
To elucidate the underlying mechanisms of the naturally deoptimized codon usage of HAV, a system to cultivate the virus in a cellular environment with a modified tRNA pool was developed. This was achieved using actinomycin D (AMD), which specifically inhibits the DNA-dependent RNA polymerases with no effect on the RNA-dependent RNA polymerases [31]. In these conditions of cellular transcription inhibition and hence cellular protein synthesis shut-off, the tRNA pool available for the virus is expected to increase and consequently the virus codon usage may readapt to the new conditions. HAV, being an RNA virus, replicates as a complex and dynamic mutant spectrum or swarm of non identical but very closely related individuals, called viral quasispecies [32],[33]. The population landscape of these molecular spectra allows a more broad analysis of all ongoing mutations necessary for the study of codon usage adaptation. The response regarding the viral replicative fitness and codon usage assessed during the adaptation to AMD exclude translation selection and instead suggest fine-tuning translation kinetics selection as the underlying mechanism of the codon usage bias in the capsid coding region.
Actinomycin D (AMD) treatment induced a clear dose-dependent inhibition of gene expression of FRhk-4 cells with a total cytoplasmic RNA reduction of around 40% and over 80% at concentrations of 0.05 µg/ml and 0.2 µg/ml, respectively (Fig. 1). Additionally, the expression of the housekeeping genes HPRT-I and GAPDH was also significantly reduced by 62% and 80%, respectively, at 0.05 µg/ml of AMD and 97% and 94%, respectively, at 0.2 µg/ml of AMD. In contrast, neither the cellular RNA abundance nor the housekeeping gene expression was significantly affected as a consequence of HAV replication. The AMD-induced cellular shut-off resulted in a cell viability of 75% and 0.25% with 0.05 µg/ml and 0.2 µg/ml, respectively, at 7 days post-treatment; of around 100% and 10%, respectively, at 4 days post-treatment, and of 100% and 60%, respectively, at 2 days post-treatment.
The effect of AMD-induced cellular shut-off on viral fitness was analyzed in an attempt to reproduce what occurs in other picornaviruses such as poliovirus. In the presence of 0.05 µg/ml of AMD, and during the early passages, HAV showed a loss in fitness (figured as virus production per cell), and at 0.2 µg/ml of AMD concentration the virus population was completely extinct (Fig. 2). In contrast PV, which induces cellular shut-off by itself and that has an optimized codon usage, readily showed a fitness gain in the presence of 0.2 µg/ml of AMD (Fig. 3).
To find out whether HAV could recover its fitness through a change of its codon usage, further passages in the presence of the drug were performed (Fig. 4). Replication in the presence of 0.05 µg/ml of AMD (lineage 1) was examined for over 150 passages (Fig. 4B) using as baseline control the viral replication in the absence of AMD (lineage 0) (Fig. 4A). As aforementioned, after few passages in the presence of the drug, viral production was severely decreased, with peaks of less than 1 TCID50 per cell (Fig. 4B). However, thereafter the viral population progressively adapted to AMD, giving rise again to viral progenies quantitatively equivalent to those of the population growing in absence of the drug. Additionally, after 65 passages the AMD-pre-adapted population was submitted to an increased concentration of the drug (0.2 µg/ml) (lineage 2), and again a decrease followed by a clear recovery in the viral production was observed (Fig. 4C). To further test the fitness of the populations adapted to the different AMD concentrations, viral populations were again submitted to the original conditions. For instance, once adapted to 0.05 µg/ml (lineage 1), the virus population was brought back to grow in the absence of the drug (lineage 3) (Fig. 4D). The immediate response was again an abrupt decrease in the viral progeny that, however, was very rapidly recovered. Similarly, when the 0.2 µg/ml AMD-adapted population (lineage 2) was brought back to the original 0.05 µg/ml concentration (lineage 4), the same pattern of behavior of an initial loss in fitness followed by a fitness recovery was observed (Fig. 4E). Nevertheless, the analysis of the viral production of the first 20 passages of the different adaptive processes showed that the kinetics of adaptation was faster during the re-adaptation to the original conditions than during the first adaptation to the different AMD concentrations, with significantly (p<0.05) different slopes of the regression lines (Fig. S1).
The Relative Codon Deoptimization Index (RCDI) [34] measures the adaptation of a virus codon usage to that of its host. An RCDI value of 1 indicates the virus follows the cell host codon usage, while the higher the value the higher the deviation from the host. Thus, a value of around 1.70 denotes that HAV posses a highly deoptimized codon usage compared to other picornaviruses whose RCDI range from 1.14 to 1.39 (Table S1). Since HAV is not able to induce cellular shut-off [26],[27] (Fig. 1), the optimization of the viral codon usage to that of the host cell could lead to an unfair competition for tRNA resources. In contrast replication in the presence of AMD, which clearly inhibits cellular expression, may represent an environment with increased available tRNA pools. Hence, an analysis to test the adaptation to tRNA pools was performed by studying the mutant spectrum of each adapting lineage at different passages and determining the codon usage in each of these spectra. Particularly passages P4, P5, P20, P36, P38, P41, P44, P65 and P85 of lineage 1 and P20 and P38 of lineage 2 (P20' and P38' in Fig. 5) were analyzed. Additionally, the mutant spectrum of lineage 3 was also analyzed at P21. In each case the quasispecies distribution of lineage 0 was used as baseline control of the molecular evolution. Three genomic regions were analyzed: two fragments from the structural polyprotein coding region (a VP3 fragment and a VP1 fragment) and a fragment from the polymerase coding region.
Any mutation induces a codon change and the newly generated codons were classified as being similarly frequent (within a 10% range), less frequent (below 10%) or more frequent (above 10%) than the original ones with respect to the cell host codon usage, as an indication of the adaptation to the cellular tRNA pool. A different pattern of molecular evolution was observed depending on the genomic region analyzed. The structural polyprotein coding regions of the AMD-adapting populations showed a tendency to progressively accumulate mutations that induced the use of codons less common than the original ones (Fig. 5B), with an average of 75% of mutations in this direction in both the last passages of lineage 1 and all passages of lineage 2. On the contrary, the new generated codons in lineage 0 were mostly similar to the original ones (Fig. 5A). Particularly, 54% of the mutations gave rise to codons of the same frequency, 31% to less frequent codons and 15% to more common codons. Quite the opposite, in the polymerase region a complete dominance of mutations inducing changes in the level of frequency (almost equally in both directions less and more frequent) of the new codons was observed in all lineages (Fig. 5C and 5D).
Assuming a parallel behavior between the particular capsid coding regions analyzed and the complete capsid coding region, and applying the Poisson distribution model, it may be postulated that while at P5, during the adaptation to 0.05 µg/ml of AMD, around 50% of the genomes would present only 1 mutation inducing the generation of a less common codon, from P38 and further 50% of the genomes would harbor at least 5 of such mutations. In other words, the percentage of genomes with zero mutations of this kind would progress from 20% at P5 to 0.20% at P85.
Since competition is established for the tRNA resources, a refined analysis involving the study of anticodon usage was performed by inferring theoretical anticodon usage tables from the actual codon usage tables which were built from the analysis of 50 molecular clones representative of a mutant spectrum (Table S4, A and S5, A), but introducing the codon-anticodon multiple pairing effects corrected by the codon-anticodon coupling efficiency [19]. Two codon usage tables were made for each viral population at each analyzed passage, one using the mutant spectra in the VP1- and VP3-region analyzed (Table S4, A) and a second one using the mutant spectrum of the 3D-region analyzed (Table S5, A). Likewise, two anticodon usage tables were built: one for the capsid region (Table S4, B) and another for the polymerase region (Table S5, B). The relative anticodon usage for each amino acid family was also figured (Tables S4, C and S5, C). Additionally, an anticodon usage table for the host cell was also built and a relative anticodon usage determined by arbitrarily giving to the most abundant anticodon in each amino acid family a value of 100% and the remaining anticodons were percentualy referred to this most abundant one. Anticodons were then sorted in three groups following this cellular anticodon usage table: anticodons used by the cell at a proportion below 20% (<20%), anticodons used by the cell at a proportion between 20% and 60% (20%–60%) and anticodons used by the cell at a proportion above 60% (>60%). The viral relative anticodon usage variation of each population in each passage, with respect to the initial passage in the absence of the drug, was calculated (Table S4, D and Table S5, D). Increases or decreases (in percentage) of use of the anticodons belonging to the above mentioned groups were analyzed and the mean variation figured (Fig. 6). Lineage 0 showed no major statistical (p<0.05) variation of the anticodon use, with the exception of the anticodon group <20% in the capsid region at P41, P44 and P65 which decreased (Fig. 6). In contrast, lineage 1 showed a significant (p<0.05) and consistent tendency, in the capsid region, to decrease the anticodon group >60% and increase the anticodon group 20%–60% (Fig. 6). This tendency was further confirmed with lineage 2 at P20 and P38 (Fig. 7). Such a tendency was not observed in the polymerase region (Fig. 6). Five anticodons (Ile: uag, uai; Thr: ugg; Val: cau, cac) among the >60% group were responsible for the decrease of the whole group, being their decrease at P65 of lineage 1 of −5.6%±2.5%. Further on this group decreased from P20 to P38 of lineage 2 from −7.4%±3.2% to −9.5%±7.3% (Fig. 7). All these variations were significantly (p<0.05) different from the values shown by the same anticodons in lineage 0. On the other sense, the variation of the six anticodons (Arg: ucc; Ile: uaa, uau; Val: cag, cai, caa) of the 20%–60% group responsible of the increase of the whole group was also significantly (p<0.05) different from that of lineage 0 and of 5.4%±4.0% (P65 of lineage 1), 8.6%±5.5% (P20 of lineage 2) and 9.6%±4.5% (P38 of lineage 2) (Fig. 7). Interestingly, when lineage 1 was returned to the original growing conditions of absence of AMD (lineage 3) for 21 passages (P21R), the anticodon groups >60% and 20%–60% increased and decreased, respectively. However, although the five specific anticodons of the >60% group and the six specific anticodons of the 20%–60% group increased from −6.3%±3.5% to −1.1%±4.8% and decreased from 6.9%±3.9% to 3.5%±2.9%, respectively, these variations were not statistically significant. However, six additional anticodons belonging to the >60% group (Ala:cgg, cgi; Glu:cuc; Gly:ccc; His:gug; Pro:ggu) did significantly (p<0.05) increase an average of 7.8%±6.5% in comparison with the original lineage 1 (P0R) and four more from the 20%–60% group (Ala:cgc; Gly:cca; His:guc; Ser:ucg) significantly (p<0.05) decreased an average of −6.30%±1.7% (Fig. S2).
To confirm that the observations made with the molecular quasispecies analysis of the three specific regions may be inferred to the whole genome, consensus sequences at P127 of lineages 0 (absence of AMD) and 1 (presence of 0.05 µg/ml of AMD), and at P62 of lineage 2 (presence of 0.2 µg/ml of AMD) were obtained. Nine mutations characterized lineage 1. Three of them were located in the capsid region (µ = 1.3×10−3) and 6 at the non-structural proteins region (NSP) (µ = 1.4×10−3). Lineage 2 showed 16 mutations, one at the 5′ non coding region (µ = 1.4×10−3), 8 in the capsid region (µ = 3.4×10−3) and 7 at the NSP region (µ = 1.6×10−3). Most of the mutations occurring in the capsid region (63%) induced a change to a less frequent codon, as also occurred in the mutant spectra (75%), while at the NSP region they mainly induced the change to a more frequent or similar one (57% and 29%, respectively). Furthermore, the anticodon analysis was also performed and it showed the same trend. A decrease of the anticodon group of >60% was observed and the average variation of the five main anticodons (Ile: uag; Leu: gau; Phe: aag; Val: cau, cac) responsible for the change of the whole group in lineage 1 was −2.08% and the variation of the eight main anticodons (Ile: uag; Leu: gau; Phe: aag; Pro: ggg, ggi; Tyr: aug; Val: cau, cac) responsible for the change in lineage 2 was −3.03%. Also an increase in the 20%–60% group was detected, with variations of the six main responsible anticodons (Ile: uaa, uau; Leu: aac; Phe: aaa; Val: cag, cai) in lineage 1 of 4.92% and of the ten main responsible anticodons (Ile: uaa, uau; Leu: aac; Phe: aaa; Ser: aga, uca; Tyr: aua; Val: cag, cai, caa) in lineage 2 of 4.25%. Although the population landscapes obtained with consensus sequences are less representative than those obtained with the molecular spectra, this is in some way compensated by the analysis of a wider length and the results observed support the conclusions from the molecular quasispecies analysis.
Moreover, during the process of adaptation to AMD, the RCDI of the capsid coding region significantly (p<0.05) increased (Fig. S3 and Table S1), indicating a re-deoptimization of the virus codon usage under the new growing conditions.
To investigate whether codon usage plays a significant role on viral fitness in conditions of depleted and abundant tRNA pools, competition experiments between viral populations with codon usages adapted to replicate in the absence or presence of AMD were carried out. With this aim mixed populations at different quantitative ratios were grown under different conditions (Fig. 8). These experiments clearly demonstrated that lineage 1 (adapted to grow in 0.05 µg/ml of AMD) was the fittest in the presence of 0.05 µg/ml of AMD and as early as after 4 passages completely out-competed lineage 0 (adapted to grow in the absence of AMD) when mixed at equal concentrations (Fig. 8A). In the most quantitatively unfavorable condition, 12 passages were required to out-compete lineage 0 (Fig. 8B). On the contrary, this latter population was the fittest in the absence of the drug and clearly out-competed the drug-adapted population after 6 and 12 passages, when mixed at equal concentrations (Fig. 8A) and in quantitatively unfavorable condition (Fig. 8B), respectively.
In the presence of 0.2 µg/ml of AMD, lineage 2 (adapted to grow in 0.2 µg/ml of AMD) clearly showed a better fitness and rapidly out-competed lineage 1 (Fig. 8C) even in the most unfavorable condition that required only 9 passages (Fig. 8D). However, in the presence of 0.05 µg/ml of AMD, lineage 1 was never able to out-compete lineage 2 and this latter lineage, although showing a relative better fitness in this condition, was unable to totally out-compete lineage 1 when mixed at equal or unfavorable ratio (Fig. 8C and 8D). This relative better fitness is also evidenced by the incapacity of lineage 1 to affect lineage 2 in 0.05 µg/ml of AMD when the starting concentration of lineage 2 was highly favorable (Fig. 8D).
The effect of the Hsp90 chaperone inhibitor geldanamycin on HAV production was investigated. HAV titers were not affected by the presence of increasing concentrations of the drug ranging from 0 to 1 µM. The estimated geldanamycin concentration inducing a 50% virus titer reduction (IC50) was 5.370 µM (Fig. 9). In contrast, PV titers were severely affected by the presence of the drug, with an estimated IC50 of 0.275 µM. Since in the particular case of PV, the decrease in titer is thought to be the result of an impairment of capsid folding, which is dependent on the activity of the Hsp90 [35], it can be assumed that HAV capsid folding is not dependent on the activity of this specific chaperone and that it should depend on other factors as might be the codon usage.
Translation selection drives the optimal co-adaptation of the codon usage and tRNA concentration, i.e. the most abundant codons pair with the most abundant tRNAs, in order to get a quantitatively highly efficient and accurate rate of translation [18]. In contrast, fine-tuning translation kinetics selection also presses for a co-adaptation between codon usage and tRNA concentration but in a different sense, i.e. the use of many different rare codons pairing with non-abundant tRNAs, in order to get a locally slow ribosome traffic rate to allow the proper protein folding [18]. The codon usage of HAV is indeed highly biased and highly deoptimized [23] with the highest RCDI value among picornaviruses, and thus translation selection does not seem to be the evolutionary driving force of its codon bias. In such a situation the viral translation rate is expected to be very slow, in agreement with a highly inefficient IRES [28], since the rate of translation is proportional to the concentration of charged tRNAs. We attempted to adapt HAV to grow in an environment with a higher tRNA availability through the specific inhibition of the cellular protein synthesis with AMD, and to study the codon usage re-adaptation, if any, to these new conditions. Although our initial hypothesis was that an environment of increased tRNA availability would result in an improvement of HAV viral translation rate, what we found during the first passages of the virus in the presence of AMD was a significant decrease of the infectious viral production per cell (Fig. 2). Nevertheless, the most striking finding was that further on (over 40 passages) a fitness recovery was observed (Fig. 4), in both the population adapting from 0 to 0.05 µg/ml (lineage 1) and the population adapting from 0.05 to 0.2 µg/ml of AMD (lineage 2). In contrast, PV did not suffer a decrease in fitness during the process of adaptation to AMD and rather experienced a significant increase of virus production per cell when growing in the presence of 0.2 µg/ml of AMD (Fig. 3). However, PV follows a completely different strategy with a highly optimized codon usage to that of the cell and thus confirming that viral production is at its best when there is a good match between codon usage (demand) and tRNA availability (supply).
The analysis of the HAV codon usage adaptation to AMD revealed an interesting adjustment in the capsid region. If translation selection is the driving force of the codon usage, a tendency to optimize the codon usage should be expected. Instead what was detected was a re-deoptimization getting to an increased use of those uncommon tRNAs. This re-deoptimization was associated with a fitness recovery in terms of infectious virus production, suggesting that the loss of efficiency in translation would be compensated by a different capsid folding, affecting stability and/or exposure of the receptor binding site. Preliminary data point to a change in the antigenic structure and thermal stability of the viral capsid during the adaptation to AMD (data not shown). Further proteomic analyses to confirm these hypotheses are in progress, but the genomic studies provide evidence suggesting that fine-tuning translation selection is actually contributing to the codon usage bias of HAV. Additional selective pressures may derive from the decrease of some cellular factors required for HAV replication and translation in conditions of cellular shut-off. Virus adjustment to this new situation may be mediated by mutations inducing changes in the RNA structure. However, it is unlikely that these mutations result in a change in RNA structure concomitant with a change in codon usage. Other alternative interpretations include AMD-associated alterations of 3D, 3C or 3CD activities although fitness recovery was not associated with mutations in these regions.
Fine-tuning translation of the capsid is graphically evidenced during the back adaptation process of viral lineage 3, from 0.05 µg/ml to 0.0 µg/ml of AMD (Fig. S2), where the population seeks a kind of dynamic equilibrium regarding the variation in the anticodon usage. Most rare codons in the capsid coding region are rare codons pairing with abundant tRNAs, while only a few of them are rare codons pairing with rare tRNAs. Many of these rare codons pairing with abundant tRNAs (62%) were replaced during the process of adaptation to AMD, all of them to codons pairing with less abundant tRNAs, while those more common codons pairing with non-abundant tRNAs were replaced at a lower frequency (34%), and most of them to codons pairing with even less common tRNAs (Table S2). Most HAV capsid residues coded by rare codons are strategically located in the carboxy terminal regions of the putative highly structured elements [23]. A higher tendency of replacement of these strategically located rare codons in comparison with those located apart from these regions was observed (Table S3), indicating the potential relevance of the translation kinetics in providing locally slow ribosome traffic rates and thus contributing to the proper capsid protein folding. Actually, it is interesting that whereas the capsid folding of many picornaviruses is dependent on the activity of the heat-shock protein 90 (Hsp90) chaperone [35], that of HAV is not (Fig. 9). Additionally, the fact that substitutions detected in the 3D region during the adaptation to AMD do not tend to re-deoptimize the codon usage as occurs in the capsid region (Fig. 6) together with a significantly lower mutation rate in the 5′ NCR of the population adapting to the drug (data not shown), which are not under the pressure of the translational machinery, re-enforces the critical role of translation kinetics in the capsid region.
Selection for fine-tuning translation kinetics in the HAV capsid acts on the whole virus population and the flattest population (mutant spectrum) rather than the fittest individual is selected. In fact, a blend of mutations occurring around the swarm of genomes affecting the overall codon usage was associated with fitness recovery. The critical role of the mutant spectra is also observed in the faster adaptation of the populations during the back processes from higher to lower and from presence to absence of AMD than during the forward processes, pointing to the existence of molecular memory in the quasispecies as described elsewhere [36].
Codon usage adaptation to tRNA availability must find a critical balance between the rate of translation and the proper protein folding to reach the highest fitness. While, generally, the viral populations adapted to a given tRNA pool, out-competed the non-adapted populations under those specific conditions (Fig. 8), the exception to the rule was the particular case of lineage 2 (adapted to grow in 0.2 µg/ml of AMD) in competition experiments with lineage 1 (adapted to grow in 0.05 µg/ml of AMD) in the presence of 0.05 µg/ml of AMD. Under these conditions, the populations were unable to out-compete each other, suggesting some kind of cooperation rather than competition (Fig. 8 C and D). Although difficult to predict, it may be hypothesized that the expected slower translation rate of lineage 2 might be compensated with a higher quality capsid-folding and that the faster translation rate of lineage 1 might provide a higher level of the viral enzymes required for RNA replication and capsid maturation. Experiments are, presently, in progress to assess this point.
HAV behavior in terms of mutation-selection for a fine-tuning translation kinetics allowed for fitness recovery but not fitness gain during the different processes of adaptation to the cellular tRNA changing conditions and thus it may represent an additional view of the Red Queen dynamics of protein translation [37]. At least from the viral side it is clear that in a hostile environment “it takes all the running you can do to keep in the same place”. Evolutionary adaptive changes are required to maintain fitness and cessation of change may result in extinction.
The cytopathogenic pHM175 43c strain of HAV was used for the study of HAV replication and evolution in the presence of actinomycin D (AMD, Sigma). Serial passages in 0.0 µg/ml (lineage 0), 0.05 µg/ml (lineage 1) and 0.2 µg/ml (lineage 2) of AMD were carried out with a multiplicity of infection (m.o.i.) of 1, at a 7-day interval. Additionally, pre-adapted populations were returned to the original conditions from 0.05 µg/ml to 0.0 µg/ml of AMD (lineage 3) and from 0.2 µg/ml to 0.05 µg/ml of AMD (lineage 4). The LSc 2ab strain of poliovirus was also grown in the absence or presence of AMD and passaged every 2 days at a m.o.i. of 1. The infectious virus titer (TCID50) was obtained for both viruses in FRhK-4 cell monolayers. Virus yield per cell was figured taking in consideration the average cell viability under each condition (between days 4 and 7, and at day 2, for HAV and PV, respectively). HAV and PV production in the presence of geldadamycin concentrations from 0.062 to 1 µM was evaluated in the same way.
AMD-associated cytotoxicity was measured by counting viable cells using the trypan blue exclusion method. Total cytoplasmic RNA abundance from 106 cells treated with 0.2 µg/ml, 0.05 µg/ml or 0.0 µg/ml of AMD, and from untreated cells infected with HAV was quantified using the NanoDrop® ND-1000 spectrophotometer, as a measure of the cellular genome expression. Additionally the expression level of two cellular genes, HPRT-I (hypoxanthine phosphoribosyl-transferase I) and GAPDH (glyceraldehide-3-phosphate dehydrogenase) [38],[39], was also monitored by end-point dilution RT-PCR using previously described primers [40],[41] for all aforementioned conditions.
Two genomic regions of the capsid coding region were analyzed: a fragment of the VP3-coding region, corresponding to amino acids 1–123, and a fragment within the VP1-coding region, corresponding to amino acids 85–245. Besides, another fragment corresponding to amino acids 1–253 of the nonstructural protein 3D (polymerase), was also analyzed. RT-PCR amplification of the specified RNA fragments was performed as described elsewhere [33]. Previously described primers [33] were used for the amplification of the VP3 and VP1 fragments, while for the 3D- fragment primers 3D- (5′ATGATTCTACCTGCTTCTCT3′) and 3CD (5′ATTGGGATCCAAGAAAATTGAAAGTCA3′) were designed. PCR products were cloned and the sequence from 50 molecular clones obtained as previously described [33].
Viral codon usage tables were obtained for each viral population at several passages through the analysis of the sequences of 50 molecular clones. Two codon usage tables were made, one using the sequences of the VP1- and VP3-fragments as a model for the structural proteins coding region and another using the sequences of the 3D-region as a model for the non-structural proteins coding region. Anticodon usage tables were inferred from these codon usage tables by assuming a model based on the frequency of the codons, the anticodon degeneracy and the codon:anticodon match pairing preferences [19] (Tables S4 and S5). Additionally, an anticodon usage table was also drawn for the host cell (Tables S4 and S5) and anticodons sorted in those used less than 20%, those used between 20 and 60% and those used more than 60%. The relative variation of usage of each anticodon at each viral passage compared to the initial passage was calculated. The mean variation of the viral anticodons belonging to each of the previously defined groups was calculated and significant differences between the mean variations in the populations growing in the absence or presence of the drug analyzed by a T-student test.
Growth competition experiments between viral lineages 0 and 1 were performed after mixing the populations at ratios of 1∶1, 100∶1 and 1∶100 and grown in the absence or presence of 0.05 µg/ml of AMD. Additionally, competition experiments between lineages 1 and 2 at ratios 1∶1, 100∶1 and 1∶100, and grown in the presence of 0.05 µg/ml and 0.2 µg/ml of AMD were also performed. A m.o.i. of 1 was used, with the exception of those experiments in which the mixing ratios were 1∶1 in which the m.o.i. was 2. Viral progeny of each competition experiment was passaged several times and consensus sequences obtained.
To follow up the proportion of each population several genetic markers were used: two mutations that were present in the consensus sequence of lineage 1 and absent in lineage 0 (a→g at nucleotide 2459 and a→g at nucleotide 2643), and two additional mutations only present in lineage 2 (c→u at position 1282 and c→u at nucleotide 1393). These latter mutations occurred in the VP0 coding region which was sequenced using previously described primers [42]. These specific genetic markers allowed a semiquantitative monitoring of the populations through determination of the proportional height of the two peaks at each nucleotide position inferred from the chromatogram of the consensus sequences.
|
10.1371/journal.pcbi.1005464 | Functional asymmetry and plasticity of electrical synapses interconnecting neurons through a 36-state model of gap junction channel gating | We combined the Hodgkin–Huxley equations and a 36-state model of gap junction channel gating to simulate electrical signal transfer through electrical synapses. Differently from most previous studies, our model can account for dynamic modulation of junctional conductance during the spread of electrical signal between coupled neurons. The model of electrical synapse is based on electrical properties of the gap junction channel encompassing two fast and two slow gates triggered by the transjunctional voltage. We quantified the influence of a difference in input resistances of electrically coupled neurons and instantaneous conductance–voltage rectification of gap junctions on an asymmetry of cell-to-cell signaling. We demonstrated that such asymmetry strongly depends on junctional conductance and can lead to the unidirectional transfer of action potentials. The simulation results also revealed that voltage spikes, which develop between neighboring cells during the spread of action potentials, can induce a rapid decay of junctional conductance, thus demonstrating spiking activity-dependent short-term plasticity of electrical synapses. This conclusion was supported by experimental data obtained in HeLa cells transfected with connexin45, which is among connexin isoforms expressed in neurons. Moreover, the model allowed us to replicate the kinetics of junctional conductance under different levels of intracellular concentration of free magnesium ([Mg2+]i), which was experimentally recorded in cells expressing connexin36, a major neuronal connexin. We demonstrated that such [Mg2+]i-dependent long-term plasticity of the electrical synapse can be adequately reproduced through the changes of slow gate parameters of the 36-state model. This suggests that some types of chemical modulation of gap junctions can be executed through the underlying mechanisms of voltage gating. Overall, the developed model accounts for direction-dependent asymmetry, as well as for short- and long-term plasticity of electrical synapses. Our modeling results demonstrate that such complex behavior of the electrical synapse is important in shaping the response of coupled neurons.
| In most computational models of neuronal networks, it is assumed that electrical synapses have a constant and ohmic conductance. However, numerous experimental studies demonstrate that connexin-based channels expressed in neuronal gap junctions can change their conductance in response to a transjunctional voltage or various chemical reagents. In addition, electrical synapses may exhibit direction-dependent asymmetry of signal transfer. To account for all these phenomena, we combined a 36-state model of gap junction channel gating with Hodgkin–Huxley equations, which describes neuronal excitability. The combined model (HH-36SM) allowed us to evaluate the kinetics of junctional conductance during the spread of electrical signal or in response to chemical factors. Our modeling results, which were based on experimental data, demonstrated that electrical synapses exhibit a complex behavior that can strongly affect the response of coupled neurons. We suggest that the proposed modeling approach is also applicable to describe the behavior of cardiac or other excitable cell networks interconnected through gap junction channels.
| In most models of neuronal networks, it is assumed that electrical synapses exhibit constant conductance, and that electric synaptic transmission is bidirectional and symmetric. However, experimental studies show that these assumptions are not always satisfied. For instance, some synapses formed of gap junction channels exhibit an instantaneous conductance–voltage rectification, which promotes a direction-dependent asymmetry of electrical signaling [1–4]. In addition, all members of the connexin (Cx) family forming gap junction channels exhibit a sensitivity of junctional conductance to the transjunctional voltage [5]. Moreover, voltage sensitivity of gap junctions can be strongly affected by chemical factors, e.g. by intracellular concentrations of H+, Ca2+ or Mg2+ [6–8]. Thus, electrical synapses are not just passive pores, but can exhibit dynamic changes of junctional conductance. Presumably, these changes in electrical synaptic strength could affect the transfer of an electrical signal. The purpose of our study was to develop a computational model for evaluation of such an interaction between electrical synapses and signal transmission between coupled neurons.
The first quantitative models describing equilibrium [9, 10] and kinetic [11] properties of junctional conductance dependence on transjunctional voltage were based on the assumption that the channel can be in two states, open and closed. Later, single channel studies have shown that transjunctional voltage causes channels to close to a subconductance (residual) state [12, 13] with fast gating transitions, and to a fully closed state with slow gating transitions [14, 15]. Thereafter, it was proposed that gap junction channels comprise two types of gating mechanisms, fast and slow, each exhibiting rectification of their unitary conductances depending on the voltage across them. These properties were described in a stochastic 16-state model (16SM) of gap junction channel gating [16] in which fast and slow gates operate between open (o) and closed (c) states. However, experimental data from our and other groups [17, 18] allowed us to suggest that the slow gate operates between open (o), initial-closed (c1) and deep-closed (c2) states. Such a suggestion was implemented in a 36-state model (36SM) of voltage gating [19]. The 36SM allowed us to reproduce experimentally observed gating behavior of gap junction channels more adequately than 16SM, especially regarding the kinetics of conductance recovery, or a low fraction of functional channels clustered in junctional plaques.
Earlier [20], we combined a 16SM of gap junction channel gating and rectification with the Hodgkin–Huxley (HH) equations [21]. The developed model (HH-16SM) allowed us to evaluate the kinetics of junctional conductance during the spread of excitation in neuronal networks. In this study, we replaced 16SM with 36SM for a better evaluation of junctional conductance kinetics. We applied the combined model (HH-36SM) to investigate the signal transfer between electrically coupled neurons in response to different types of presynaptic inputs, such as electrotonic signals or action potentials (APs).
In this study, we analyse three main aspects of the 36SM with respect to the functional behavior of electrical synapses. Firstly, transjunctional voltage distribution across each channel gate can result to almost instantaneous asymmetric conductance-voltage rectification of gap junction channel. We showed that such rectification of gap junctions can affect the asymmetry of the electrical cell-to-cell signaling, especially the spread of a single AP.
Secondly, in the 36SM, the gap junction channel can transit between open and closed states, and probabilities of these transitions depend on voltage across each channel gate. We demonstrated that closing of gap junction channels could be induced by transjunctional voltage spikes, which develop during the spread of excitation. More precisely, our modeling results showed that voltage spikes induced by the trains of APs can cause an accumulation of gap junction conductance decay. As a result, the junctional conductance can significantly decrease in just a few seconds, and substantially modulate electrical signaling between neurons. This short-term plasticity of electrical synapses was supported by our electrophysiological experiments in HeLa cells expressing connexin45.
Thirdly, we suggested that some types of chemical modulation of electrical synapses could be explained by an assumption that the values of 36SM parameters depend on chemical factors. Under such a hypothesis, the chemical modulator would influence the junctional conductance by modifying voltage sensitivity properties of a gap junction channel. In this case, the chemically-induced variation of junctional conductance would be explained by the changed equilibrium of open and closed voltage sensitive channels, and not by a separate chemical gate. To illustrate the feasibility of this idea, we fitted the 36SM to explain the kinetics of connexin36 gap junctional conductance under different concentrations of free magnesium ions ([Mg2+]i). We demonstrated that a long-term (a few minutes) plasticity, which is induced by variation in [Mg2+]i, can be adequately reproduced through the changes of 36SM parameters.
Thus, the presented model accounts for the complex behavior of electrical synapses under a wide variety of voltage and temporal conditions. Moreover, all these phenomena can be explained by the underlying mechanisms of gap junction channel voltage gating. Such a modeling approach allows one to evaluate the response of neuronal networks, which would be very difficult to measure experimentally.
Junctional conductance of electrical synapses was evaluated using a Markov chain 36-state model of voltage gating, which is detailed in [19]. The model describes the probabilistic behavior of gap junction channels in response to the transjunctional voltage. In the 36SM, the gap junction channel consists of two hemichannels, each enclosing one fast and one slow gates (Fig 1). Thus, the channel comprises four gates (fast left, slow left, slow right and fast right), all arranged in series (Fig 1B). The fast and the slow gates operate according to a linear kinetic schemes, o↔c and o↔c1↔c2, respectively (Fig 1A). Thus, gap junction channel can be in 36 (2∙3∙3∙2) different states, and overall junctional conductance is estimated as an averaged value of each state conductance weighted to their probabilities. Transition probabilities between system states depend on transjunctional voltage distribution across gates, which must be evaluated first. In general, the voltage distribution can be nonlinear due to rectification of unitary conductances of channel gates.
The developed HH-36SM combines Hodgkin–Huxley equations that describe excitability of neurons and a 36-state model (36SM) of gap junction channel gating that evaluates conductance of the electrical synapse. More precisely, membrane voltages of the neurons are estimated using the Hodgkin-Huxley model. The resulting transjunctional voltage can affect the junctional conductance, which is evaluated using the 36SM. Thus, the HH-36SM allowed us to simulate electrical signal transfer between neurons connected through modulatable gap junctions.
Asymmetry of electrical synaptic transmission has been observed in numerous studies [41–44]. Such asymmetry might arise due to differences in input resistances (Rins) of coupled neurons [29], even when gap junctions themselves are symmetric. Rin depends on the conductivity of the plasma membrane and its surface area, as well as on the number of neighboring neurons connected through electrical synapses. Another source of electrical synaptic transmission asymmetry is related to instantaneous conductance–voltage rectification of gap junction channel, which results from the inhomogeneous distribution of charged amino acids lining the pore [45]. Such rectification of gap junction channels typically arises in heterotypic junctions under normal conditions [4, 17, 46], but it can also develop in homotypic gap junctions under an asymmetry of intracellular milieu, e.g. gradients of [Mg2+]i [47]. In addition, electrical signaling asymmetry across heterotypic channels can arise with repeated stimulation due to voltage gating (see Fig 7). This type of asymmetry in electric synaptic transmission is not instantaneous and depends on past history. The other factors that contribute to asymmetry of signaling do not have this property.
Our data show that asymmetry in electrotonic cell-to-cell communication is more affected by the difference in Rins of coupled cells (see Figs 3D and 5D), while gap junctional rectification primarily influences an asymmetry of AP transfer between neurons (Fig 5A–5C). This can be explained by the conductance–voltage curves in Fig 2, which show that conductance changes are small at low voltages (±10 mV), which typically arise during measurements of coupling coefficients. Significant changes of junctional conductance can only be expressed at high voltages (±100 mV), which develop during the spread of excitation. We believe that these observations might have practical applications in electrophysiological experiments when studying the strength and rectification properties of electrical synapses.
The aforementioned sources of functional asymmetry are independent by nature, e.g. Rin of a neuron directly depends on plasma membrane area, while synaptic rectification is determined by properties of gap junction channels [48]. Thus, they can act antagonistically promoting bidirectionality of electrical synapses, as was demonstrated in the teleost auditory system [4]. Alternatively, if rectification of gap junctions and differences in Rins acted synergistically, it could facilitate unidirectional AP transfer. Thus, unidirectionality, which is a genuine property of chemical synapses, could be executed through electrical synapses alone. Because electrical synaptic transmission is faster than chemical, unidirectional spread of AP through gap junctions might be useful in rapid response warranting behavior such as escape reflex [49, 50].
Asymmetry of electrical synaptic transmission plays an important role in spike-timing regulation, as was demonstrated in neurons of the thalamic reticular nucleus [44]. In larger networks, even a small asymmetry would add up during the spread of excitation and could significantly affect the latency of AP transfer along neural pathways. This process could be crucial in temporal coding activities, such as coincidence detection, in which gap junctions are reported to play an important role [50]. Presumably, the effect of asymmetry of electrical signaling would be difficult to measure and observe experimentally in highly complex neuronal networks, and a simulation-based approach could provide valuable insights on the role of rectification in network dynamics [51].
It is well established that gap junctional conductance depends on voltage [10]. Our previous [20] and current modeling studies show that decay of junctional conductance can be induced by voltage gating of gap junction channels during bursting activity of neurons. To our knowledge, at least one study reported such spiking activity-dependent reduction of electrical synaptic strength in brain slices [52]. Our data showed that even in gap junctions formed of low-voltage-sensitive Cx36, this decay exceeds 10% while in more voltage-sensitive Cx isoforms it could reach ~50% over several seconds (Figs 6 and 7). The magnitude of junctional conductance decrease and duration of its recovery depends not only on Cx properties but also on the firing rates of neurons (Fig 6). Because the transfer of electrical signal and its asymmetry depends on junctional conductance [53], an activity-induced inhibition of electrical synapses can significantly diminish (Fig 6A-b) or even abolish AP transfer between neurons (Fig 8C-c). Such a role of electrical synaptic plasticity was acknowledged in [54] and was demonstrated by an activity-dependent decrease of junctional conductance together with enhanced asymmetry of electrical synaptic transmission in TRN slices [52].
Heterotypic gap junctions exhibit structure-determined voltage-gating asymmetry, which could result in even more diverse functional behavior with respect to plasticity and directionality than homotypic gap junction. As we showed in Fig 8, changes in junctional conductance and the response rate of neurons depends on the direction of AP spread with respect to the orientation of heterotypic gap junctions. Thus, heterotypic synapses could promote direction-dependent asymmetry of electrical signal transfer not only by its rectification properties but by asymmetric voltage gating as well. We presume that such processes might have an important functional role in sensory systems where heterotypic electrical synapses are detected [55, 56].
Regulation of the strength of electrical synapses by a variety of chemical reagents is well established. Others and our data showed that junctional conductance decay caused by chemical uncouplers can be reversed by voltage, while some chemical factors can change voltage sensitivity of Cxs [8, 38, 39, 57]. These observations, as well as the fact that all known chemical uncouplers close gap junction channels fully but not to residual conductance, suggest that some chemical factors act through the slow gate. We implemented this idea by simulating Mg2+-mediated changes in junctional conductance of Cx36 gap junctions using the 36SM. The obtained data revealed that an effect of [Mg2+]i can be relatively well reproduced (Fig 9) assuming variation in values of 36SM parameters, mainly V0 and probabilities of c1↔c2 transitions of slow gates. Moreover, because the voltage sensitivity of the gap junction channels is defined by the same parameters (see Fig 4 in [20] and Fig 6 in this paper), chemically modulated gating would also affect spiking activity-dependent short-term plasticity of the electrical synapse. Our modeling results showed that even a moderate change (±20%) in [Mg2+]i could result in very significant differences in the spread of APs between two neurons (see Fig 10). Thus, chemically modulated gating of Cx36 can expand the time window of electrical synaptic plasticity for as long as chemical factors are present, which could last for minutes or even hours. Therefore, even Cx36, which exhibits relatively low voltage sensitivity, could act as a highly modulatable constituent of neuronal networks due to chemically mediated gating.
Our modeling results show that a persistent spiking activity or chemical factors could keep a significant proportion of gap junction channels in a closed state. We assume that this process could offer at least a partial explanation to a well-documented ‘low functionality’ of gap junctions, especially those expressed in excitable cells, such as neurons or cardiomyocytes. Low functionality refers to a small fraction of channels residing in the open (or high conductance) state. This applies to all connexins, such as Cx36 [58], Cx43 [26], Cx45 [57] and Cx57 [59], examined on this issue, and likely applies to other Cx isoforms.
The strength of electrical synapses directly affects the level of synchronization in neuronal networks, which can underlie various physiological processes and pathological brain conditions. For example, increased cortical synchronization correlates with reduced information processing capability in the primary auditory cortex [60]. The rise in junctional conductance can lead to over-synchronization, which is associated with episodes of epileptic seizures. Interestingly, an activity-induced decrease in the coupling of electrical synapses through an intracellular Ca2+ mechanism was observed in the thalamic reticular nucleus of epileptic rats and was proposed to act as a compensatory mechanism to reduce excessive synchronization [61]. Thus, both voltage- and chemically induced gating of gap junction channels can play an important role in shaping activity of neuronal networks through modulation of neuronal synchrony. In addition, short-term plasticity induced through voltage gating of electrical synapses could contribute to lateral inhibition and resulting center-surround effect, which is important in sensory systems of the CNS. This hypothesis is supported by studies showing that more voltage-sensitive Cx isoforms are expressed in the structures associated with sensory functions. For example, one of the most voltage-sensitive Cxs, mouse Cx57 and its human homolog Cx62 are expressed in horizontal cells of the retina [62], while Cx45, which is significantly more voltage-sensitive than Cx36, predominates in the olfactory bulb [34].
The chemically mediated gating could play an important role in regulating longer term changes, especially in less-voltage-sensitive Cx36. For example, it was reported that Cx36 plays an important role in shifting between sleep and wake states [63]. We believe that the unique sensitivity of Cx36 to Mg2+ could contribute to this process. This view is supported by accompanying changes in ATP levels, which effectively influence [Mg2+]i. It was reported that ATP levels increase during the initial hours of sleep in wake-active regions of rat brain [64]. This should decrease [Mg2+]i and, consequently, increase conductance of Cx36 gap junctions. As a result, an increased synchronization could suppress activities in brain regions associated with the waking state, thus maintaining sleep.
In this study, we used the Hodgkin–Huxley equations to describe excitability of neurons. The developed model can be adapted to various brain regions and circuits by choosing an appropriate set of ionic currents. For example, the inclusion of Ca2+ currents, which underlie bursting trains of APs in thalamic relay neurons [65], might be relevant for short-term plasticity as well as for chemical modulation of electrical synapses. Furthermore, major principles used to develop an HH-36SM can be applied in cardiac tissue modeling, provided that the Hodgkin–Huxley equations are replaced by those specific for cardiomyocytes [66, 67]. Cardiomyocytes are predominantly connected through Cx43, Cx40 and Cx45, which are more voltage sensitive than Cx36; therefore, it might exhibit more expressed activity-dependent conductance decrease, especially during tachyarrhythmias. Furthermore, chemically mediated gating of cardiac gap junction channels, e.g. by acidification [68], could be important in describing enhanced arrhythmogenicity of the ischemic myocardium [69].
Obviously, the 36SM of gap junction channel voltage gating is a simplification of complex processes underlying changes of electrical synaptic strength. However, we believe that rectification and voltage gating properties of gap junction channel can be reasonably well reproduced using the 36SM. On the other hand, an inclusion of chemical modulation into 36SM is far less explored. So far we made only the first steps in this direction to explain Cx36 mediation by [Mg2+]i, and presented modeling results (Fig 10) are obtained from just a few data points. Moreover, cytosolic conditions are rarely defined by a single chemical factor, and various different reagents might affect electrical synapses synergistically or antagonistically. For example, our preliminary data suggest that [Mg2+]i effect on Cx36 gap junctions might depend on the pH level. In addition, modulation of electrical synapses by other chemical reagents, such as Ca2+ ions, might be more relevant for the spread of excitation than that of [Mg2+]i.
In this study, we simulated electrical synaptic transmission between two cells connected through a soma-somatic gap junction. For a more realistic neuronal network simulation, it would be beneficial to include dendro-dendritic connections, which are far more prevalent in mammalian brain. Another important extension of our model would be an inclusion of chemical synapses. Presumably, this would allow one to study an interaction between chemical and electrical synapses, which was observed in numerous experimental studies [70].
However, all physiologically relevant extensions, and especially an increased number of cells and synapses, might require a large amount of computational recourses. To our knowledge, Hodgkin-Huxley type models are rarely applied for large neuronal network simulation due to computation time constraints. This problem would be enhanced by our modeling approach, because evaluation of junctional conductance using the 36SM consumes ~95 percent of overall computation time. We presume that simulation time could be decreased by two different approaches: 1) Creation of a more simplistic model of gap junction voltage gating, which would roughly describe relative changes of junctional conductance in response to a single AP. Somewhat similar approach is applied in mathematical models of chemical synapses [71]. This would allow one to combine a model of electrical synapse with integrate-and-fire type models, which are often used for simulation of large neuronal networks. 2) Application of advanced computation techniques, such as an extensive parallelization together with graphic processing unit computation.
|
10.1371/journal.ppat.1003802 | The Genetic Basis of Escherichia coli Pathoadaptation to Macrophages | Antagonistic interactions are likely important driving forces of the evolutionary process underlying bacterial genome complexity and diversity. We hypothesized that the ability of evolved bacteria to escape specific components of host innate immunity, such as phagocytosis and killing by macrophages (MΦ), is a critical trait relevant in the acquisition of bacterial virulence. Here, we used a combination of experimental evolution, phenotypic characterization, genome sequencing and mathematical modeling to address how fast, and through how many adaptive steps, a commensal Escherichia coli (E. coli) acquire this virulence trait. We show that when maintained in vitro under the selective pressure of host MΦ commensal E. coli can evolve, in less than 500 generations, virulent clones that escape phagocytosis and MΦ killing in vitro, while increasing their pathogenicity in vivo, as assessed in mice. This pathoadaptive process is driven by a mechanism involving the insertion of a single transposable element into the promoter region of the E. coli yrfF gene. Moreover, transposition of the IS186 element into the promoter of Lon gene, encoding an ATP-dependent serine protease, is likely to accelerate this pathoadaptive process. Competition between clones carrying distinct beneficial mutations dominates the dynamics of the pathoadaptive process, as suggested from a mathematical model, which reproduces the observed experimental dynamics of E. coli evolution towards virulence. In conclusion, we reveal a molecular mechanism explaining how a specific component of host innate immunity can modulate microbial evolution towards pathogenicity.
| The selective pressure imposed by the host immune system is an important component of microbial adaptation from commensalism to pathogenicity. We used experimental evolution to study the initial steps of the adaptation of Escherichia coli to cells of the innate immune system, i.e., macrophages. Our results demonstrate that bacteria can evolve remarkably fast, and acquire adaptations increasing survival inside macrophages and/or ability to escape engulfment. The mechanism underlying this pathoadaptive process involves the accumulation of mutations caused by transposon insertions, increasing pathogenicity in vivo. These findings reveal the remarkable fast pace at which bacteria can evolve to escape a central component of the host innate immunity, namely macrophages.
| Bacteria can be used to study evolution in real time in controlled environments, i.e. experimental evolution [1]. Different studies have demonstrated that bacterial populations have an enormous potential to adapt to relatively simple abiotic challenges under laboratory environments [2], [3]. On the other hand, far less is known on how biotic interactions shape bacterial adaptive evolution. Antagonistic interactions (predation, parasitism) are likely to be important determinants of the rate of adaptive change observed in bacteria, their trait diversity and genome complexity [4], [5], [6]. The best-studied antagonistic interaction in an evolutionary laboratory setting is the one involving bacteria and their phages, which increases rates of bacterial adaptation and diversification [7], [8], demonstrating that biotic interactions can have an important role in bacterial evolution [9]. Another common antagonistic interaction faced by bacteria occurs when these infect mammals and are directly exposed to cells of the host immune system. To our knowledge this interaction has never been addressed in an experimental evolution context. Here, we determined the mechanisms via which E. coli evolve to overcome the antagonistic interaction imposed by one of the central components of host innate immunity, namely monocyte/macrophages (MΦ).
E. coli is both a commensal and a versatile pathogen, acting as a major cause of morbidity and mortality worldwide [10]. Moreover, there is evidence that some pathogenic E. coli evolved from commensal strains [11], [12], making E. coli an ideal organism to study the transition from commensalism to pathogenicity. E. coli colonizes the infant gastrointestinal tract within hours after birth, and typically builds a mutualistic relation. However, non-pathogenic strains of E. coli can become pathogenic, when the gastrointestinal barrier is disrupted as well as in immunosuppressed hosts [13], [14], [15].
MΦ are a key component of host defense mechanisms against pathogens [16]. They can provide direct bactericidal response through phagocytosis, a process by which bacteria are killed inside endocytic phagosomes, through the generation of reactive oxygen and nitrogen species among other effector mechanisms. Yet many bacterial species are capable to escape and resist eukaryotic cells [17], [18], suggesting that several bacterial defense mechanisms evolve upon encounter with MΦ. Adaptive microbial mechanisms to escape MΦ include surface masking and capsule formation (to avoid engulfment and phagocytosis), increased motility, filamentation and biofilm formation. Mechanisms acting after engulfment by MΦ include toxin release. Within the species of E. coli alone, there are examples of several different mechanisms [19].
In the present study, we established an in vitro system in which E. coli is allowed to evolve under continuous selective pressure of MΦ, and ask how quickly and by which mechanisms commensal E. coli evolve resistance to one of the sentinels of the innate immune system, the MΦ.
We followed the evolution of six E. coli populations (all founded from the same ancestral clone), when adapting to the antagonistic interaction imposed by the murine monocytic cell line (RAW 264.7), referred throughout the text as MΦ. The bacterial populations (M1 to M6) evolved, by serial passage, in complete culture medium with MΦ and were propagated at a multiplicity of infection (MOI) of 1∶1 (106 E. coli to 106 MΦ, see Fig. S1). After 24 hours bacterial numbers reach around 4×108 and are subsequently bottlenecked to start the next passage with 106 bacteria again. In parallel, we also evolved E. coli under identical experimental conditions in the absence of MΦ (the resulting evolved clones are named CON). In this case the population is propagated by daily passages involving a bottleneck of 104 cells at each passage. This results in ∼15 generations per day, given the increase in bacteria numbers observed during 24 hours. All populations evolved for a period of 30 days, which corresponds to approximately 450 generations. We note that this is an approximate value because as adaptation proceeds the population dynamics will change and differences in the number of generations per day will occur.
Adaptation of the bacterial lines in the presence of MΦ was characterized by the emergence of phenotypic variation within populations. After 4 days of evolution, i.e. approximately 60 generations, distinct colony morphologies emerged in all populations, detected when plating on LB plates (Fig. 1A). Such morphological diversity was never observed in control populations evolved for 30 days under the same experimental conditions in the absence of MΦ (n = 6). Two distinct heritable morphs were identified and scored, i.e. small colony variants (SCV) and large translucid mucoid (MUC) colonies and their frequencies were quantified over time (Fig. 1B). SCVs were observed in five out of six populations, but this morph remained at low frequency and was only detected transiently. The parallel emergence of SCVs in independent evolving populations, suggests that this phenotype constitutes an initial adaptation of E. coli to the antagonistic interaction imposed in vitro by MΦ. In contrast, MUC clones which rose in frequency in all populations, reached fixation in five out of six populations by day 30. The changes in frequency of SCVs and MUCs showed complex dynamics (Fig. 1B). In some populations, once SCVs decreased in frequency MUCs tended to increase, e.g. populations M2 and M3. This suggests that MUCs can outcompete SCVs, presumably due to a larger fitness advantage. These observations suggest that E. coli morphological diversity can emerge very rapidly as a result of their adaptation to MΦ.
Competitive fitness of E. coli populations was measured at two time points during the process of evolution (day 19∼285 generations and day 30∼450 generations), revealing that all populations exhibit a significant fitness increase (Fig. 2A). On average, fitness increase was of 0.10 (2SE = 0.07) and 0.27 (2SE = 0.10) after 19 and 30 days, respectively. Fitness increased between generations 285 and 450 across populations (Students' paired t-Test, P = 0.02).
The observation that SCVs emerged in at least 80% of the independent evolving populations but with low frequency strongly suggests that SCVs have a transient selective advantage that is outcompeted over time. To better understand this selective advantage we performed two assays: 1) exposure of MΦ in vitro to SCVs to test for possible intracellular versus extracellular growth differences relative to that of the ancestral strain; 2) a fitness assay to determine the ability of SCV to outcompete the ancestral strain, in the presence of MΦ. We did not observe any difference in SCV growth either intracellularly (Rr = 0.99+0.16 (2SE)) or extracellularly (Rr = 1.01+0.13 (2SE)) relative to the ancestral non-evolved clone, while there was an advantage in the competitive fitness assay (Fig. 2B). SCVs (clones SCV_M1_D8 and SCV_M3_D5) exhibited a fitness advantage relative to the ancestral strain, inside MΦ, as assayed 2 hours after infection. However, this advantage was restricted to the early phase of infection, given that SCVs showed a disadvantage outside MΦ 24 hours after infection (Fig. 2B). These results probably explain why SCVs increased in frequency but failed to reach fixation (see Fig. 1B).
We tested the in vitro evolved SCVs for traits common to those of clinical SCV isolates from different bacterial species [20] [21]. The evolved E. coli SCVs showed an increased resistance to aminoglycosides, but not to other antibiotics (see Supplemental Table S1), were catalase negative and showed a remarkable instability. In rich medium SCVs reverted to a large colony phenotype at a frequency of 9×10−4 (2SE = 4×10−4) and supplementation with hemin enhanced their growth relative to the ancestral (SCV_M1_D8: 2.9±1 (2SE) and SCV_M3_D5: 2.5±0.7(2SE)). These results imply that the selective pressure of MΦ led to the emergence of phenotypes typical of pathogenic bacteria.
Mucoidy, the trait evolved in the MUC clones, is also a trait observed in certain infections, for example in Pseudomonas aeruginosa or E. coli [22], [23]. The in vitro evolved MUCs produce high levels of exopolysaccharides when plated on LB. Since colanic acid is present in most natural E. coli isolates [24], and this capsule is made in mutants of E. coli that emerge under stress conditions [25], we tested mucoid clones for overproduction of this exopolysaccharide. Mucoid clones showed overproduction of colanic acid (Fig. 2C, Fig. S2), suggesting that rapid evolution to change this trait can occur under the specific selection pressure imposed by MΦ. We tested whether MUCs escaped MΦ engulfment, by quantifying the relative abundance of intracellular and extracellular of MUC after 3 hours co-incubation with MΦ. Relative abundance of intracellular bacteria in MΦ was lower for MUC versus the ancestral strain in 6 out of 6 MUC clones tested (Fig. 2D). Moreover, the extracellular abundance of MUC clones relative to ancestral was higher in 4 out of 6 MUCs tested. We then asked whether MUCs would trigger MΦ cytotoxicity, a process that would contribute to reduce the negative impact exerted by MΦ on MUC versus ancestral clones. MΦ cytotoxicity was similar in the presence of MUC versus ancestral clones (Fig. S3A and Fig. S3B). Furthermore MUCs did not cause any significant changes in MΦ ability to engulf the ANC clone (Fig. S3C). Taken together, these results strongly suggest that MUCs are better adapted to escape MΦ but do not diminish the ability of MΦ to internalize ancestral E. coli.
We tested whether adaptation of evolved MUC clones to escape MΦ is associated with increased virulence. We compared the survival of mice infected systemically via the intra-peritoneal route, with increasing amounts of MUC versus ANC bacteria or bacteria that evolved in the absence of MΦ (CON) (Fig. 3A). The lethal dose 50 (LD50) of MUC infection (LD50: 2.8×107, with 95% CI 1.4×107–5.8×107) was 5–10 times lower than that of ANC (LD50: 1.6×108, with 95% CI 8.5×107–2.8×108) or CON (LD50: above 1×108), as inferred from the confidence intervals (Fig. 3A and 3B), suggesting that MUC clones have increased virulence. In agreement with these observations, infection with ancestral or with bacteria evolved in the absence of MΦ at a dosage corresponding to the MUC LD50 was not lethal, i.e., 100% survival of mice occurred (log-rank test: χ22 = 9.9, p = 0.007; Fig. 3C). Higher lethality of MUC infection was associated with significant reduction in temperature (but not weight), as compared to infection with ANC bacteria at the dosage corresponding to the MUC LD50 (χ22 = 0.61, p = 0.0004; Fig. 3D).
We then asked whether MUC bacteria elicited a MΦ response in vitro that would be somehow altered, as compared to the response elicited under the same conditions by the ANC or CON clones. When co-cultured with MUC, primary mouse peritoneal MΦ produced similar levels of the pro-inflammatory cytokine TNF, as compared to MΦ co-culture with ANC or CON clones (see Text S1 and Fig. S4). This suggest that although MUC clones have evolved to escape MΦ in vitro and increasing their pathogenicity in vivo, these clones are still readily detected by MΦ, as revealed by TNF secretion. This read out was used hereby as out-put of pattern recognition receptor triggered signaling leading to the activation of a core pro-inflammatory signal transduction pathway, which appears to be equally responsive to the different bacterial clones tested.
Overall our results show that the MUC clones, which overproduce colanic acid and dominated the bacteria populations during the interaction with MΦ, exhibit increased virulence.
Given the phenotypes of the MUCs and their dynamics, we sought to determine the molecular basis of the mutations responsible for their increase in frequency along the evolutionary process. Whole genome sequencing of a clone sampled from M3 population at day 19 of the evolution process (MUC_M3_D19) revealed that it carries two transposon insertions, i.e. a IS186 insertion into the promoter region of lon and one IS1 insertion upstream of the yrfF gene (see Table 1). The IS1 insertion event occurred in all sequenced clones sampled at day 30 (Table 1, Fig. 4A), revealing parallelism at the genetic level across all independently evolved lines. The function of the yrfF gene is unknown in E. coli, but its homologue in Salmonella, i.e. igaA, prevents over-activation of the Rcs regulatory system, which regulates colanic acid capsule synthesis [26]. It is therefore likely that the insertion upstream of yrfF alters E. coli ability to produce colanic acid, in keeping with the observation that MUC clones produce high levels of colanic acid, as compared to ANC bacteria (Fig. 2C).
Other important parallelisms (observed in 3 out of 6 populations analyzed) include two transposition events, namely, one in yiaW coding region and the other in the coding region of the pot operon. potD is one of the four genes of the potABCD operon, a spermidine-preferential uptake system [27]. All four genes are essential for spermidine uptake, indicating that the insertions in potD detected in clones MUC_M4_D30 and MUC_M6_D30, or the insertion in potA observed in clone MUC_M3_D30, are likely to impair uptake of spermidine. We tested the effect of polyamines in the evolved MUC clones and observed that while all exhibit a growth advantage in the presence of spermidine, the clones with insertions in potD (MUC_M4_D30 and MUC_M6_D30) exhibit an increased growth advantage compared to the other MUC and the ancestral, in the presence of spermine (Fig. S5). During adaptation to MΦ, insertion in yiaW (whose function is unknown) was followed rapidly by insertion in potA or potD genes (see Fig. 4B, M3, M4 and M6 populations), indicating a potential interaction between these two events. This parallelism was observed in populations exposed to MΦ and not in bacteria that evolved in the absence of MΦ, suggesting that insertions in yiaW contribute functionally to adaptation of E. coli to MΦ. Given that many of the adaptive mutations observed under the different forms of stress imposed by MΦ were caused by IS insertions, we tested if the frequency of spontaneous mutations (including IS insertions) is higher in the presence versus absence of MΦ in the ancestral strain. No significant differences were found, suggesting that selection was the main force driving the increase in frequency of IS elements (see Text S1 and Fig. S6).
Other parallelisms were observed at the level of point mutations in two clones with the same non-synonymous SNP in fusA, a gene coding for elongation factor G, which catalyzes the elongation and recycling phases of translation [28]. Mutations in fusA reduce the rate of protein synthesis, a hallmark of stress responses, with pleiotropic effects on bacterial physiology [29]. Mutations in fusA have also been related with the development of SCVs in S. aureus [30]. We sequenced fusA in our in vitro evolved E. coli SCVs (11 clones sampled from M1 population at day 8 and 10 clones sampled at day 4) but did not find any substitutions in this gene.
To further understand the dynamics of adaptation in each independent evolved bacterial population, we sought to determine the frequency of the mutations found (see Table 2), in clones sampled along the evolution experiment. Adaptation involved the competition between distinct haplotypes and the successive accumulation of beneficial mutations, mainly caused by IS insertions (Fig. 4B). Such haplotype dynamics is characteristic of clonal interference [3], where clones carrying distinct beneficial mutations compete for fixation. We modeled this process, within the basic ecological scenario of our experiment (see Fig. 5 and Text S1, Fig. S7 to S11), fully reproducing the complex dynamics of the mucoid and non-mucoid phenotypes observed in Figure 1B.
An IS186 insertion into the promoter region of lon, was observed in clones sampled from populations M3 and M4 (Fig. 4). Lon (Long Form Filament) is a heat shock protease responsible for degradation of defective proteins in the cell [31]. The promoter region of lon is a hotspot for IS186 insertions [31], which may contribute to the occurrence of this mutation in independently evolved clones. We tested if the proportion of spontaneous lon::IS186 mutants is higher in the presence versus absence of MΦ, however no difference was observed (see Text S1). As lon mutants tend to overproduce colanic acid [32], a trait that appears to be strongly selected for in our experimental system, it is possible that this was the main beneficial effect caused by the insertion. However, the IS186 insertion could only be detected at intermediate time points in the experiment and not at day 30 (see Fig. 4B). Interestingly, lon has been reported to be a mutator gene in mutants that bear an IS186 insertion in its promoter, thus increasing the rate of IS transpositions 10- to 100-fold [33]. This happens because the stability of several transposases is dependent on the Lon protease [34], [35], which seems to regulate their transposition activity. We tested MUC_M3_D19 for increased mutagenesis. This clone carries an IS186 inserted in −10 promoter region of lon and since mutations in this position were shown to significantly decrease level of lon transcription [36], it is likely that it could be a mutator. If so this could contribute to the burst of transposition events that occurred during adaptation. We found a significant increase in the frequency of D-cycloserine resistant clones in MUC_M3_D19 relative to the ancestral non-evolved clone (median frequency 2.6×10−6 vs. 1×10−7, for the ancestral background, P = 5.5×10−13, W = 203.5, Mann-Whitney U test, Fig. S12). Consistent with this increased mutagenesis being driven by IS insertions, no significant differences in the frequency of rifampicin resistant clones, which are caused by point mutations, were observed (median frequency 3.3×10−7 vs 6.9×10−7 for the ancestral background, P = 0.1, W = 21, Mann-Whitney U test). The presence of IS186 in the lon promoter region was also found to be highly unstable. A spontaneously derived non-mucoid clone from MUC_M3_D19 (MUC_M3_D19_REV) shows a precise excision of this element, while maintaining the IS1 insertion in regulatory region of yrfF (see Table 1). These results indicate that this IS186 insertion enhances mucoidy levels, increases mutagenesis and is also very unstable in this genetic background. The latter may explain why it did not fix in the populations. The dynamics of the IS186 insertion in populations M3 and M4 suggest that this mutation was beneficial in the background with an IS1 insertion upstream of the homologue of igaA. Support for a selective advantage of this mutation is suggested by the observation that, in Salmonella, the transcription of igaA (yrfF in E. coli) is regulated by lon [37].
Bacterial evolution towards pathogenicity may occur through the acquisition of new genes – a gain of function mechanism- or modification of their current genomes, including loss of genes - change-of-function mechanism [38]. The later constitutes a pathoadaptation, in which mutations enhance bacterial virulence without horizontal transfer of specific genes. For example, the deletion of hemB in Staphylococcus aureus increases its ability to persist intracellularly [39] while the loss of mucA increases Pseudomonas aeruginosa ability to evade phagocytosis and resist to pulmonary clearance [40].
We followed the evolution of a commensal strain of E. coli under the selective pressure imposed by MΦ phagocytosis, to determine the rate of adaptive evolution and to uncover the nature of possible E. coli pathoadaptive mutations. From the infection dynamics and the fitness assays (Fig. 1B and 2), we conclude that at least two different adaptations, detected by the emergence different colony morphologies, occurred, namely, i) an intracellular advantage evolved by SCV clones early in the process; ii) an extracellular advantage evolved by MUC clones emerging later. The intracellular adaptation is characterized by increased bacterial resistance, plasticity and survival in the early phase of interaction with MΦ, and was accompanied by a reduced extracellular growth. The extracellular adaptation is associated with overproduction of colanic acid and characterized by increased resistance to MΦ phagocytosis. The functional link between overproduction of colanic acid and escape from phagocytosis is likely but remains to be formally established. Overtime this phenotype dominated all populations.
The mutations acquired by commensal E. coli adapting to MΦ, occurred within a few hundred generations and were characterized by traits reminiscent of those found in pathogenic bacteria. Clinical isolates sampled from patients suffering from recurrent and persistent infections in the blood [41] or urinary tract [21], [42], are SCVs. The distinctive traits of this phenotype are: i) ability to form small colonies, to revert to larger colony forming bacteria at high frequencies and ii) increased resistance to certain antibiotics. In S. aureus SCVs have been implicated as an intermediate form before mutations in gyrA occur to produce ciprofloxacin resistance [43]. In addition, Besier et al. have reported thyA mutant S. aureus SCVs show hypermutator status [44]. These findings suggest that SCVs could potentiate the emergence of mucoid clones, which latter go on to dominate the populations. However, we did not detect in SCVs the mutations found in the mucoid clones, indicating a distinct molecular basis for the SCV phenotype, an issue that we will investigate in future work. Given that mucoidy is also frequently observed in certain infections [22], [23], our finding that this trait can rapidly emerge under the selective pressure of MΦ, may have implications not only for the understanding of host-microbe interactions but also for the treatment of bacterial infections. Interestingly mucoidy can also be selected by the pressure imposed by phages in different bacterial species [45], [46]. Whether mucoid strains evolved to resist to phages also exhibit increased virulence remains to be established.
Translocation of commensal E. coli from the gut can be associated with severe health complications (e.g. sepsis), particularly in immunosuppressed hosts or after surgery [47], [48]. Bacteria that reach the mesenteric lymph nodes or the peritoneal cavity (extensively populated by MΦ) and that are able to escape MΦ should have a fitness advantage and potentially cause more severe disease. Indeed, we found that increased ability to escape MΦ of in vitro evolved clones lead to increased pathogenesis in vivo, when tested in a mouse model. We also found that this pathoadaptative process was characterized by three main paths. Although distinct in the number and type of mutations, these share an initial mutation: an IS insertion upstream of yrfF, a gene which shares 84% sequence similarity at the protein level to IgaA of Salmonella enterica serovar Typhimurium. In S. Typhimurium it was shown that the stability and responsiveness of the RcsCDB system depends on IgaA [49]. The RcsCDB system controls the production of colanic acid, virulence in diverse pathogens [24], [50], [51], [52], [53], modulates responses to environmental changes and is activated upon exposure to antimicrobial peptides [54], [55], [56], [57]. IgaA represses the RcsCDB system [58] and mutations causing instability of IgaA activate the RcsCDB system, leading to overproduction of colanic acid capsule (mucoid phenotype) [58]. Given the repressive function of IgaA on RcsCDB, which controls many traits likely to be important for bacterial fitness, it is likely that the observed IS insertion upstream of yrfF is an adaptive mutation with pleiotropic effects. If so the adaptive path may proceed through the occurrence of new mutations, which may compensate for the pleiotropic effects of that first adaptive step. Interestingly, the same amino-acid substitution in fusA occurred in two independent lines. FusA is an elongation factor and is part of the str operon of E. coli, which has 3 other genes: rpsL, rpsG and tufA. Since the strain that we studied carries a mutation in rpsL that confers streptomycin resistance, which is costly in RPMI yet increases survival inside MΦ [59], it is possible that the SNP in fusA could be compensatory to cost of the rpsL mutation in the milleu outside MΦ.
One of the adaptive paths taken by E. coli included insertions into the coding regions of yiaW and potA or potD. While the function of yiaW is unknown, the later genes are involved in spermidine transport, which may affect E. coli interaction with MΦ. Spermidines are polyamines, polycationic molecules, which interact with nucleic acids and have been described as important in escape from phagolysosomes, biofilm formation and protection from oxidative and acidic stress amongst other traits important in bacterial pathogenesis [60].
The adaptive process was also marked by the occurrence of an IS186 insertion into the promoter region of the Lon protease. Such insertion was not only likely adaptive (it was observed in two independent lineages and it increases mucoidy), but also likely leads to increased rates of transposition. Given that many of the adaptive mutations observed under the stresses imposed by MΦ were caused by ISs, these may constitute an example of Barbara McClintock proposal that transposable element movement under stress could aid organisms to adapt to new environments [61].
The mechanisms via which different mutations underlying E. coli pathoadaptation increase its virulence remain to be established. It is likely however, that such mechanisms would interfere with one or two host defense strategies against infections [62]. Presumably, by escaping MΦ killing pathoadaptation should provide MUC clones with a proliferative advantage, ultimately compromising host survival. This should be revealed by increased bacterial burden in the MUC infected hosts, as compared to hosts infected with non-evolved E. coli clones, revealing a compromise in host resistance [62]. An alternative, but not mutually exclusive, interpretation would be that pathoadaptation is associated with the induction of a immunopathologic response compromising host survival, irrespectively of pathogen burden. This should be revealed by similar bacterial burdens in the MUC infected host, as compared to hosts infected with non-evolved E. coli clones, revealing a compromise in host disease tolerance [62]. While critical to further understanding of the mechanism via which E. coli pathoadaptation increases its virulence, this is beyond the scope of the current study.
In conclusion, we demonstrate that E. coli can adapt to better resist to MΦ within a few hundreds of generations and that clones with different morphologies and traits similar to those of pathogenic bacteria rapidly emerge. This pathoadaptive process and the complex dynamics of the evolved phenotypes can be reasonably described by a model of clonal interference, where distinct haplotypes, carrying new transposon insertions and other mutations, increase in frequency and compete for fixation.
All experiments involving animals were approved by the Institutional Ethics Committee at the Instituto Gulbenkian de Ciência (project nr. A009/2010 with approval date 2010/10/15), following the Portuguese legislation (PORT 1005/92) which complies with the European Directive 86/609/EEC of the European Council.
The RAW 264.7 murine macrophage cell line was maintained in an atmosphere containing 5% CO2 at 37°C in RPMI 1640 (Gibco) supplemented with 2 mM L-glutamine (Invitrogen), 1 mM sodium pyruvate (Invitrogen), 10 mM hepes (Invitrogen), 100 U/ml penicillin/streptomycin (Gibco), 50 µM 2-mercaptoethanol solution (Gibco), 50 µg/ml gentamicin (Sigma), with 10% heat-inactivated FCS (standard RPMI complete medium). Before infection assays, MΦ were cultivated for 24 h in the same medium as before except for the three antibiotics which were now replaced by 100 µg/ml streptomycin antibiotic (RPMI-Strep medium). All bacterial cultures were also done in RPMI-Strep medium, except if stated otherwise.
The Escherichia coli strains used were MC4100-YFP and MC4100-CFP (MC4100, galK::CFP/YFP, AmpRStrepR) which contain the yellow (yfp) and cyan (cfp) alleles of GFP integrated at the galK locus in MC4100 (E. coli Genetic Stock Center #6152) and differ only by YFP/CFP locus that is constitutively expressed [63]. MC4100-CFP strain was used for the evolution experiment and MC4100-YFP as a reference strain for the fitness assays.
Twelve populations were founded from a single MC4100-CFP clone and were therefore genetically uniform in the beginning of the experiments. All populations evolved in RPMI, 6 populations in the presence of the MΦ (M1–M6) and the other 6 (C1–C6) in the absence of MΦ. Before each infection cycle, MΦ (0.7×106 to 1.3×106/ml) were centrifuged at 1200 rpm for 5 min, re-suspended in RPMI-Strep medium and activated with 2 µg/ml CpG-ODN 1826 (5′TCCATGACGTTCCTGACGTT 3′ - Sigma) for 24 h [64]. Cells were then centrifuged (1000 rpm for 5 min), re-suspended in 3 ml of fresh RPMI-Strep medium and seeded in 12-well microtiter plates (0.8×106 to 1.6×106/ml). Subsequently, they were incubated at 37°C for 2 h, washed in RPMI-Strep and infected with a MOI of 1∶1 (106 bacteria to 106 MΦ). After 24 hours of infection, MΦ were detached with cell scraper and the whole culture was centrifuged at 4000 rpm for 10 min to pellet cells. This procedure lyses MΦ releasing intracellular bacteria. Then these were washed twice with phosphate-buffered saline (PBS) and counted by flow cytometry using a FACscan cytometer (Becton Dickinson). Approximately 106 of recovered bacteria were used to infect new activated MΦ in the same manner as before. The same procedure was applied to control populations, except that 104 bacteria were transferred daily. This is because after 4 hours of infection with the MΦ bacteria numbers drop to 104. This adjustment results in similar number of generations in both environments. In both treatments (with and without MΦ), bacteria were allowed to propagate for approximately 15 generations per day. Generation time is estimated as: G = log2(Nf/Ni), where Ni is the initial number of bacteria and Nf is the final number of bacteria. Nf was approximately 6×108 in both treatments. Evolution occurred during approximately 450 generations in both environments. We note that in the context of a real infection repeated contact with macrophages will not likely occur with a similar period as the one in this experimental setup.
To estimate competitive fitness of M1–M6 populations, after 285 and 450 generations of evolution, each evolved population was competed against MC4100-YFP reference strain in the same conditions as used in the evolution experiment. Both evolved and ancestral strains were grown separately in RPMI-Strep, 106 cells of each type were used to inoculate the competition plate. The initial and final ratios of both strains were determined by Flow cytometry. The fitness of each population was measured 3 times and the fitness of the ancestral strain 10 times to confirm the neutrality of the marker. A measure of relative fitness increase, expressed as selection coefficient, was estimated as:[65] where Scoeff is the selective advantage of the evolved strain e over the ancestral strain a, Nfe and Nfa are the numbers of evolved (e) and ancestral (a) bacteria after competition and Nia and Nie are the initial numbers, before the competition.
Bacterial uptake was measured by the gentamicin protection assay as previously described [66], with modifications, as follows. MΦ were infected at MOI 1∶1 as described above to determine the number of intracellular and extracellular bacteria after 3 h of incubation. The number of extracellular bacteria at 3 h of incubation was estimated by taking a sample of the culture medium (without detaching the MΦ), centrifuging (4000 rpm for 10 min) to pellet the cells and finally washing these in PBS prior to plating on LB agar plates. The number of intracellular bacteria was estimated by washing infected MΦ twice with PBS and adding fresh medium containing 100 µg of gentamicin/ml to kill extracellular bacteria. After incubation for an additional hour, the medium was removed, the monolayer of macrophages was washed 3 times with PBS, detached using a cell scraper and centrifuged (4000 rpm for 10 min) to pellet the cells. These were further resuspended in PBS and the appropriate dilution was plated on LB agar plates to determine the number of intracellular bacteria. Relative abundance (Rr) of evolved clones to that of the ancestral in intracellular or extracellular environment of MΦ was estimated as:where N3he and N3ha are the numbers of evolved (e) and ancestral (a) bacteria at 3 hours post infection (in the intracellular or extracellular niche of MΦ) and Nia and Nie are the initial numbers of evolved (e) and ancestral (a) bacteria used for inoculation.
To measure numbers of MΦ that are alive, the same infection protocol was performed. However, after 3 h of infection, MΦ were washed from extracellular bacteria twice with RPMI, detached and counted by Trypan blue exclusion test [67] (see Fig. S3).
The method used to extract colanic acid was based on a procedure described previously [68]. Briefly, 50 ml of a bacterial cell culture was heated for 15 min at 100°C to denature EPS-degrading enzymes, cooled down and centrifuged at 13200 rpm at 4°C for 30 min. Then 40 ml of the supernatant was precipitated by addition of three volumes of ethanol. The mixture was maintained at 4°C overnight and centrifuged again at 13200 rpm at 4°C for 30 min. The resulting pellet was dissolved in 5 ml of distilled water, dialyzed for 48 h against distilled water (membrane MWCO, 3500 Da) and dried in SpeedVac. Residual polypeptides were removed by precipitation with 5 ml of 10% (v/v) trichloroacetic acid and centrifuged at 13200 rpm at 4°C for 30 min. The supernatant was dialyzed for five days against distilled water and dried. The resulting preparation was resuspended in 1 ml of distilled water. Quantification of colanic acid was carried out by measuring non-dialyzable methylpentose (6-deoxy-hexose), namely fucose, which is a specific component of this exopolysaccharide. 10 to 100 µl of the colanic acid preparation were diluted to 1 ml with distilled water, and mixed with 4.5 ml of H2SO4/H2O (6∶1; v/v). The mixture was prepared at room temperature, then heated at 100°C for 20 min, and finally cooled down to room temperature. For each sample, absorbance at 396 nm and 427 nm was measured either directly (control sample (A-co)) or after addition of 100 µl of 0.3% freshly prepared cysteine hydrochloride (cysteine sample (A-cy)). The absorption due to the unspecific reaction with H2SO4 was subtracted from the total absorption of the sample: A396-co and A427-co were subtracted from A396-cy and A427-cy, respectively, to obtain ΔA396 and ΔA427. Values of (ΔA396–ΔA427) were directly correlated to methylpentose concentration by using a standard curve obtained with a fucose concentration ranging from 2 µg/ml to 100 µg/ml (Fig. S2).
To determine the reversion frequency of SCV to the ancestral phenotype, single colonies grown on LB agar plates were resuspended in PBS, the appropriate dilution was plated on LB agar plates and incubated at 37°C. After 48 h small and large colonies were counted [21]. To test for the auxotrophy to hemin, individual SCV colonies were isolated, resuspended in PBS and plated on M9 minimal medium agar plates containing 2% glucose with and without hemin 50 µg/ml (Sigma-Aldrich). After incubation at 37°C for 48 h, CFUs were counted to estimate the ratio between the number of cells able to grow in presence and in absence of hemin.
Both the ancestral and 7 isolated MUC clones (MUC3_d19 sampled from population M3 after 19 days of evolution with macrophages and MUC1 to MUC6d30 sampled from M1 to M6 pops after 30 days of evolution) were grown overnight in 10 ml of RPMI-Strep at 37°C. DNA isolation from these cultures was done following a previously described protocol [69]. The DNA library construction as well as the sequencing procedure was carried out by BGI. Each sample was pair-end sequenced on an Illumina HiSeq 2000. Standard procedures produced data sets of Illumina paired-end 90 bp read pairs with insert size (including read length) of ∼470 bp.
Mutations in the two genomes were identified using the BRESEQ pipeline [70]. To detect potential duplication events we used SSAHA2 [71] and the paired-end information to map reads only to their best-match on the genome. Sequence coverage along the genome was assessed with a 250 bp window and corrected for GC% composition by normalizing by the mean coverage of regions with the same GC%. We then looked for regions with high differences (>1.4) in coverage. We did not find any such difference between the ancestral and evolved clones. See Table 2 for the identity and precise location of mutations identified in the sequenced clones. All mutations were confirmed by direct target sequencing.
In order to determine the frequency of the mutations in clones sampled along the experiment, DNA was amplified by PCR (to identify IS insertions) and sequencing PCR was performed (to identify SNPs). DNA was amplified by PCR in a total volume of 50 µl containing 1 µl bacterial culture, 10 µM of each primer, 200 µM dNTPs, 0.5 U Taq polymerase and 1× Taq polymerase buffer. The amplification profile was 15 min at 95°C, followed by 35 cycles at 94°C for 30 s, 60°C for 90 s, 72°C for 2 min with a final extension at 72°C for 10 min. All gene fragments were amplified using these conditions and oligonucleotide primers (Table S2). The same primers were used for sequencing straight from the PCR product.
We maintained male C57/BL6 mice, aged 8–10 weeks (in house supplier, Instituto Gulbenkian de Ciência), on ad libitum food (RM3A(P); Special Diet Services, UK) and water, with a 12 hour light cycle, at 21°C. We initiated infections by intra-peritoneal inoculation of bacteria in 100 µl saline. Several groups of mice were injected with different bacterial strains at doses ranging from 2×105 to 3×108 (sample sizes: ancestral – n = 46; control – n = 41, mucoid – n = 50). At doses 1×107, 5×107 and 1×108, we injected a minimum of 10 mice, in at least two independent experiments (data from the same animals was used for the Kaplan-Meier curves in Fig. 3C). The inocula consisted of the following: a single clone for ancestral bacteria (ANC), consisted of a mixture of equal numbers of the 6 sequenced clones from day 30 (MUC1-MUC6; see Fig. 4) for the mucoids (MUC) and mixture of 6 independent clones evolved in the absence of macrophages (CON). Furthermore, as a control, in each experimental block we injected a group of 2–3 mice with 100 µl of saline (these animals did not display any signs of disease). We monitored mice for a period of 6–10 days (twice a day for the first two days and daily for the remaining 8 days) and measured weight and temperature.
To estimate the LD50 values (Fig. 3A–B,E), we fitted a binomial generalized linear model (GLM) for each morphotype, using survival as a response variable and log10 bacterial dose as explanatory variable (following [72]). To analyze the temporal dynamics of mortality in mice infected with MUC or ANC at the MUC LD50 (Fig. 3C), we used Kaplan-Meier curves followed by a log-rank test. Finally, we used GLMs to test whether the variation maximum reduction in temperature or weight could be explained by the infecting strain.
The statistical analysis was performed using the R software: http://www.r-project.org/.
|
10.1371/journal.pntd.0001720 | Administration of Triclabendazole Is Safe and Effective in Controlling Fascioliasis in an Endemic Community of the Bolivian Altiplano | The Bolivian northern Altiplano is characterized by a high prevalence of Fasciola hepatica infection. In order to assess the feasibility, safety and efficacy of large-scale administration of triclabendazole as an appropriate public health measure to control morbidity associated with fascioliasis, a pilot intervention was implemented in 2008.
Schoolchildren from an endemic community were screened for fascioliasis and treated with a single administration of triclabendazole (10 mg/kg). Interviews to assess the occurrence of adverse events were conducted on treatment day, one week later, and one month after treatment. Further parasitological screenings were performed three months after treatment and again two months later (following a further treatment) in order to evaluate the efficacy of the intervention.
Ninety infected children were administered triclabendazole. Adverse events were infrequent and mild. No serious adverse events were reported. Observed cure rates were 77.8% after one treatment and 97.8% after two treatments, while egg reduction rates ranged between 74% and 90.3% after one treatment, and between 84.2% and 99.9% after two treatments. The proportion of high-intensity infections (≥400 epg) decreased from 7.8% to 1.1% after one treatment and to 0% after two treatments.
Administration of triclabendazole is a feasible, safe and efficacious public health intervention in an endemic community in the Bolivian Altiplano, suggesting that preventive chemotherapy can be applied to control of fascioliasis. Further investigations are needed to define the most appropriate frequency of treatment.
| Fascioliasis is highly prevalent in the northern Altiplano of Bolivia. We wanted to ascertain whether a preventive chemotherapy approach, involving the large-scale distribution of triclabendazole within endemic communities, would be feasible for controlling morbidity associated with this disease. Consequently, we implemented a pilot intervention among schoolchildren in a community near Lake Titicaca and assessed this intervention's safety (by evaluating the occurrence of adverse events following treatment) and its efficacy (by measuring changes in prevalence and intensity of infection). Adverse events on treatment day, and one week and one month later were infrequent and mild, and no serious adverse events were reported. We observed cure rates of 77.8% after one treatment and 97.8% after two treatments, egg reduction rates of 74–90.3% after one treatment and 84.2–99.9% after two treatments, and a decrease in the proportion of high-intensity infections (≥400 epg) from 7.8% to 1.1% after one treatment and to 0% after two treatments. We conclude that administration of triclabendazole is a safe and efficacious public health intervention for control of fascioliasis in an endemic community in the Bolivian Altiplano. Preventive chemotherapy with triclabendazole, without individual-level diagnosis and treatment, appears therefore as a feasible option. However, further investigation is needed to define the most appropriate frequency of treatment.
| Preventive chemotherapy, the large-scale administration of anthelminthic drugs to population groups at risk, is recommended by WHO for control and elimination of lymphatic filariasis, onchocerciasis, schistosomiasis and soil-transmitted helminth infections [1]. The aim of preventive chemotherapy is to regularly reduce worm load in infected individuals, thus controlling the associated morbidity and decreasing transmission rates. Biological and epidemiological similarities between Fasciola spp. and the helminths responsible for the diseases mentioned above, suggest that morbidity associated with fascioliasis could also be controlled through preventive chemotherapy by keeping intensity of infection at low levels among populations at risk [2].
Fascioliasis is a snail-borne zoonosis that can be transmitted to humans through the consumption of raw aquatic vegetables or fresh water contaminated with the cystic larval stages of the worms (metacercariae) [3]. Recent, conservative estimates on the burden of fascioliasis indicate that the number of individuals infected worldwide is at least 2.65 million, and more than 50% of them live in Latin America [4].
F. hepatica is the only liver fluke species transmitted in Bolivia [5], where endemic communities face among the highest prevalence and intensity of F. hepatica infection in the world [6]–[9]. The area endemic for talp'a laqu, as fascioliasis is known in the local Aymara language, is limited to a relatively small region (60×60 km) of the northern Altiplano (i.e. the plain between Lake Titicaca and the capital city La Paz) [8], where transmission is linked to the presence of rivers and subsoil effluences inhabited by the intermediate snail host, Galba truncatula. In this region, the main reservoirs of infection are domestic animals, including ovines, bovines, porcines and equines [10].
In humans, fascioliasis is associated with an acute clinical phase resulting from the migration of the immature worms through the liver. Symptoms include fever, abdominal pain, respiratory disturbances and skin rashes. The chronic phase starts when the worms reach the bile ducts: progressive inflammation leads to fibrosis and thickening of the walls of the biliary system and of the surrounding hepatic tissue. Biliary colic pain due to blockage of the bile ducts and jaundice are possible complications. Severe infections may result in biliary cirrhosis with scarring and fibrosis of the liver [3]. Anaemia is a common finding in both acute and chronic fascioliasis [11]–[14].
Triclabendazole is the WHO-recommended essential medicine for treatment of fascioliasis [15]. The range of the cure rate produced by a single 10 mg/kg administration is 78–100% [16]–[21], while information on egg reduction rate (ERR) is less abundant: three studies conducted in Egypt reported ERR of 73% and 100% based on arithmetic means [18], [19], and 63% based on geometric means [21]. Triclabendazole is generally regarded as a safe drug, although adverse events (AEs) can occur following treatment [16], [17]. Such events are directly proportionate to intensity of infection and can be classified as systemic or mechanical. Systemic AEs are caused by biological substances released by the dying worms and include mild/transient dizziness, headache, nausea, and urticaria. Mechanical events are generally linked to the expulsion of dead worms from the biliary system towards the intestinal lumen, and include biliary colic pain, possibly associated with jaundice.
Treatment with triclabendazole has usually been implemented in a clinical setting while its use in public health interventions is limited. In Egypt, however, a triclabendazole treatment programme based on selective chemotherapy (test-and-treat) of school-age children has been implemented in six districts in the Nile Delta area since 1998 [22], [23], while in Vietnam treatment has been decentralized since 2006 and is administered in peripheral hospitals and health posts based on a simplified diagnostic protocol [AF Gabrielli, personal communication]. In Bolivia, where prevalence of infection is higher than in Egypt or Vietnam, a more inclusive strategy offering treatment to entire population sectors without individual diagnosis might be appropriate in order to reduce costs and logistics related to the implementation of screening exercises. This approach would mirror the one currently recommended for schistosomiasis and soil-transmitted helminthiasis in areas of moderate and high risk [1]; the procurement of larger quantities of medicines needed for its implementation has been made possible via the donation of triclabendazole (Egaten) by Novartis Pharma AG through the WHO.
Consequently, in 2008, following suggestions by a panel of experts convened by the WHO [24], the Ministerio de Salud y Deportes of Bolivia decided to opt for large-scale distribution of triclabendazole in endemic areas without individual diagnosis. Before this approach was widely implemented, a pilot study was conducted to test the safety and efficacy of such intervention. Safety assessments consisted of monitoring and recording any AEs occurring after treatment, and efficacy was assessed by monitoring prevalence and intensity of infection and by calculating cure and egg reduction rates. Safety, in particular, was considered as a key component of the protocol as AEs are known to limit the feasibility of preventive chemotherapy interventions for helminth infections, as their occurrence confines treatment to a clinical setting where proper management of cases is ensured by health personnel. AEs also have the potential to jeopardize compliance to the intervention as effects of treatment can be perceived as a greater health risk than the disease itself [25].
The study protocol was approved by the Comisión de Etica de la Investigación (CEI) of the National Bioethics Committee of Bolivia on September 10, 2007. A written informed consent form explaining the purpose and the modalities of the study was developed, translated into Aymara and obtained from the parents/guardians of each participating child. The initiative was agreed by the civil (sub-alcalde, head of the health post, director of the educational unit), and traditional (jilakatas and malkus) authorities of the community where the study was implemented.
The study was conducted in Huacullani, a community in the Bolivian northern Altiplano, where prevalence of F. hepatica infection among school-aged children ranged between 31.2% and 38.2% in the 1990s [9]. Huacullani (16°26′0″S, 68°44′0″W) is located at an altitude of 3,850 metres, approximately 500 meters from the shores of Lake Titicaca, in the municipality of Tihuanaku (province of Ingavi, department of La Paz). At the time of the survey, the population of Huacullani was 2,472.
School-aged children (5–14 years) were selected as the target group of the intervention, as they are at higher risk of infection and morbidity. Children are more likely than adults to become infected, as exemplified by their higher levels of prevalence of infection, and to develop mechanical AEs following treatment, because of the smaller size of their bile ducts and thus higher likelihood of blockage. Consequently, they are considered both the group at highest risk and the one most sensitive for detection of AEs following treatment. All children attending the primary school and the junior high school of Huacullani were considered eligible for enrolment in the study.
A Scientific Committee formed by the Ministerio de Salud y Deportes, the Servicio Departamental de Salud of La Paz, the Universidad Mayor de San Andrés and the PAHO/WHO was established with the aim of developing a protocol and supervising the implementation of the pilot intervention. The protocol consisted of five consecutive study phases: baseline data collection; treatment; monitoring of AEs at day 0, day 7 and day 30; first parasitological follow-up 3 months after treatment, with further treatment of any cases still positive; and second parasitological follow-up 2 months after the first follow-up (Figure 1). After a few preparatory meetings, field activities started in April 2008, and were completed in November 2008.
At the time of the baseline survey (April 2008), the school population of Huacullani consisted of 459 children aged 5 to 14 years, who were all considered eligible for treatment. 447 children returned the plastic container. In total, 437 faecal samples from an equivalent number of children were examined by the Kato-Katz thick smear technique – 4 children returned an empty plastic container, and 6 other children provided insufficient stool quantities to prepare a Kato-Katz slide. Stool samples were transported to the Faculty of Medicine of the Universidad Mayor de San Andrés in La Paz and processed. Slides were read within 24 hours of preparation.
Overall, 95 children had positive and 342 had negative Kato-Katz smears. The parasitological prevalence of F. hepatica infection in this population was therefore 21.7%. Among the 95 children with positive Kato-Katz smears, 15 had an intensity of infection ≥300 epg (15.8%), and 11 a high-intensity infection (≥400 epg, 11.6%). The mean intensity of infection among all surveyed children (including the ones with negative smears) was 72.9 epg.
Triclabendazole was administered in June 2008 to each child testing positive to the Kato-Katz test. Among the 15 children with an intensity of infection ≥300 epg, 10 were hospitalized before treatment, while 5 could not be treated as their parents refused hospitalization and/or treatment. By contrast, all the 80 Kato-Katz positive children with an intensity of infection <300 epg were treated as outpatients at school premises. In total, 90 children were administered triclabendazole: among them, the mean intensity of infection was 264.3 epg, and 7 had a high-intensity infection (≥400 epg, 7.8%).
Among the 90 treated children, the number reporting one or more AEs on treatment day and one week after treatment (June 2008) was 11 and 10, respectively. One month after treatment (July 2008), only 82 children were interviewed, as 8 were neither at school nor could be traced in Huacullani; among them, only three children reported any AE. Details are provided in Table 2.
The number of reported AEs on treatment day, one week after treatment and one month after treatment was 15, 13 and 3, respectively. Headache was the most frequent event reported on treatment day, and abdominal pain was the most frequent one week later. All fevers were below 38°C. Only 3 of the children experiencing AEs on treatment day also reported an AE one week after treatment. Only 1 of the children with a high-intensity infection (≥400 epg) reported an AE one week after treatment (abdominal pain).
Among children treated at school, only one girl requested to be taken to the local health post on treatment day, but after a medical examination, she did not require any specific medical attention, and all the signs and symptoms resolved spontaneously. None of the other children contacted the health post for medical assistance during the follow-up period.
Overall, no medications were administered to treat AEs with the exception of antipyretics in case of fever. No SAEs occurred.
Overall, 21.7% of the children surveyed were found to be infected with F. hepatica at baseline. This was less than the prevalence of infection previously detected in Huacullani [9], but was nevertheless high when compared with the usually low levels of F. hepatica in most endemic countries across the world, such as Egypt, Iran, Vietnam or Yemen for example, where prevalence of infection by faecal examination rarely exceeds 5% [10], [22], [23], [30]–[32]. It is also likely that the true prevalence of infection is higher due to the low sensitivity of a single Kato-Katz smear.
Treatment with a single administration of triclabendazole (10 mg/kg) did not elicit frequent or considerable AEs, neither among children with a high intensity of infection, nor among the others. No significant medical attention was required in any case, as all symptoms resolved spontaneously without any appreciable consequence on the health status of treated individuals. The occurrence of AEs documented by our study contrasts with the absence of any event reported in Egypt [23] even though both the treatment regimen and the manufacturer of triclabendazole were the same. While comparison might not be fully appropriate, as in Egypt no active search of events was carried out, such discrepancy might be attributable to the lower mean intensity of infection observed in this country (12.2 epg at baseline). Frequency and severity of AEs are however expected to be less important at subsequent rounds of treatment in reason of the progressively decreasing intensity of infection, as shown by experiences from different helminth control interventions implemented across the world [1], [21], [33].
The parasitological cure rate achieved after a single administration of triclabendazole at 10 mg/kg was high (77.8%) and consistent with previous reports in the scientific literature for this treatment course [27]–[29]. ERRs were also considerable, even though lower rates were observed among individuals with a higher intensity of infection at baseline (Table 4). The negative relationship between ERR and baseline intensity of infection has been described in the case of other helminth infections, such as those by Trichuris trichiura: both density-dependent fecundity and reduced bioavailability of triclabendazole per adult worm have been proposed as possible explanatory hypotheses [34], [35]. Finally, only 1.1% of the 90 treated children still had high-intensity infections (≥400 epg) at the first parasitological follow-up, compared to 7.8% at baseline. If we apply to fascioliasis the model described in other helminth infections, that intensity of infection is proportionate to morbidity [1], [2], it can be inferred that morbidity was under control in a very high proportion of children three months after a single administration of triclabendazole 10 mg/kg.
Based on the results of the pilot intervention, we conclude that triclabendazole is a safe and efficacious drug when administered to a paediatric population living in a fascioliasis endemic area. These considerations suggest that a population-based drug distribution approach, without individual diagnosis and without direct medical supervision, in a manner comparable with the preventive chemotherapy interventions implemented worldwide against the four major helminth infections, is appropriate and feasible.
Notably, triclabendazole was well tolerated across the population examined, including individuals with a high intensity of infection: AEs elicited were self-limiting, did not require any specialist medical attention and could be managed by the local health staff. In terms of efficacy, a single administration of triclabendazole was effective in reducing considerably the number of infected individuals, the mean intensity of infection and the proportion of high-intensity infections, and in keeping these indicators at low levels for a few months after treatment.
Surveys with a longer follow-up are recommended in order to ascertain for how long a single administration of triclabendazole can sustain low prevalence and intensity of infection in endemic areas. Such a study would allow the most appropriate interval of re-treatment to be determined.
Following the successful implementation of the pilot intervention, the health authorities of Bolivia decided to implement distribution of triclabendazole on a large scale.
|
10.1371/journal.pgen.1007062 | A trehalose biosynthetic enzyme doubles as an osmotic stress sensor to regulate bacterial morphogenesis | The dissacharide trehalose is an important intracellular osmoprotectant and the OtsA/B pathway is the principal pathway for trehalose biosynthesis in a wide range of bacterial species. Scaffolding proteins and other cytoskeletal elements play an essential role in morphogenetic processes in bacteria. Here we describe how OtsA, in addition to its role in trehalose biosynthesis, functions as an osmotic stress sensor to regulate cell morphology in Arthrobacter strain A3. In response to osmotic stress, this and other Arthrobacter species undergo a transition from bacillary to myceloid growth. An otsA null mutant exhibits constitutive myceloid growth. Osmotic stress leads to a depletion of trehalose-6-phosphate, the product of the OtsA enzyme, and experimental depletion of this metabolite also leads to constitutive myceloid growth independent of OtsA function. In vitro analyses indicate that OtsA can self-assemble into protein networks, promoted by trehalose-6-phosphate, a property that is not shared by the equivalent enzyme from E. coli, despite the latter’s enzymatic activity when expressed in Arthrobacter. This, and the localization of the protein in non-stressed cells at the mid-cell and poles, indicates that OtsA from Arthrobacter likely functions as a cytoskeletal element regulating cell morphology. Recruiting a biosynthetic enzyme for this morphogenetic function represents an intriguing adaptation in bacteria that can survive in extreme environments.
| For free living bacteria, little is known about how environmental cues are perceived and translated into changes in cell morphology. Here we describe how a biosynthetic enzyme involved in synthesis of an important intracellular osmoprotectant doubles as an osmotic stress sensing morphogenetic protein. This protein is involved in an adaptive response involving a growth transition in stress-tolerant bacteria, from growing as individual cells to forming non-separating branched cell aggregates. We demonstrate that the protein can self-assemble into large networks, consistent with its role as a morphogenetic protein, this assembly process being promoted by a metabolic product of the enzyme. Depletion of either this metabolite or the morphogenetic protein results in the inability of the bacteria to grow as individual cells in conditions of low osmolarity.
| Trehalose (a-D-glucopyranosyl(1,1)-a-D-glucopyranoside) is a non-reducing disaccharide that functions as an important intracellular protectant against a variety of stress conditions including desiccation, dehydration, heat, cold, and oxidation [1]. At least four different pathways for trehalose biosynthesis have been reported, described as OtsA/B, TreY/Z, TreS, and TreT [2]. The OtsA/B pathway is the principal pathway for trehalose biosynthesis and is widely distributed in bacteria, fungi and plants (trehalose-6-phosphate synthase and trehalose-6-phosphate phosphatase in Arabidopsis thaliana). OtsA utilises UDP-glucose and glucose-6-phosphate to synthesize trehalose-6-phosphate (T6P) and subsequently OtsB converts T6P into Pi and trehalose. As a signaling molecule, T6P is important as an ‘energy checkpoint’ during development in eukaryotes. For example, in Saccharomyces cerevisiae, T6P controls the decision to proceed through cell division [3] and in plants it regulates flowering both in the leaf and in the shoot apical meristem [4].
Osmotic stress can inhibit the growth rate and affect the morphology of bacteria belonging to several genera, including Arthrobacter species, Rhodococcus species [5] and Aeromonas hydrophila [6]. However, the relationship between osmotic stress, trehalose biosynthesis and the regulation of cell division and morphogenesis is unclear. Cell division and dynamic reorganisation of cell morphology depends on the internal cytoskeleton or scaffolding elements. In most bacteria, the tubulin homolog FtsZ is critical for driving binary fission [7]. The FtsZ protein assembles into protofilaments that are bundled together to form the Z-ring at the site of cell division, usually at the mid-cell. Other components of the division machinery are then recruited to form a multi-protein divisome complex responsible for mid-cell peptidoglycan synthesis. Contraction of the Z-ring also drives constriction of the cell envelope to form the septum [8]. Other cytoskeletal elements that have a role in determining cell morphology in typical rod-shaped eubacteria include MreB, an actin-like ATPase cytoskeletal proteins, that in vitro can polymerize into filaments in the presence of ATP or GTP [9, 10] and guide lateral wall peptidoglycan synthesis [11, 12].
The actinobacteria include species of contrasting morphologies, including coccoid-shaped Rhodococci, rod-shaped Mycobacteria and Corynebacteria, filamentous spore-forming Streptomyces and pleomorphic Arthrobacter. Actinobacteria studied thus far grow by apical extension, with new peptidoglycan synthesized and added at the cell poles. This contrasts with other eubacteria that insert new peptidogylcan in their lateral walls, with the poles being inert [13–15]. Apical growth in actinobacteria is independent of MreB. In fact, the genomes of actinobacteria that adopt bacillary-type growth lack mreB homologs, and filamentous Streptomyces use an MreB protein only during sporulation [16–18]. Apical growth is guided by the protein DivIVA [13–15]; DivIVA assembles to form an internal cytoskeletal element at the cell poles that appears to function to recruit proteins for apical cell-wall synthesis.
Actinobacterial Arthrobacter species typically inhabit soil ecosytems and are fascinating for their pleomorphism. During exponential growth, rod-shape cells elongate and undergo cell division at the midcell region; the two daughter cells remain joined forming a V shape and subsequently separate by snapping apart [19]. A reversible transition from rod-shaped cells to non-separating multi-cellular, branching myceloids is induced in some species by osmotic stress and this is documented to be an adaptive response to promote bacterial survival through altered metabolism and increased resistance to environmental stress [20, 21]. Indeed, Arthrobacter typically exhibit high resistance to, among other stresses, cold, heat and dessication [22, 23]. Stress resistance is likely related to their pleomorphic behaviour, making them an interesting model for analysis of environmentally triggered developmental switches although, to date, there has been a paucity of molecular characterization of these bacteria. Arthrobacter sp. strain A3 (hereafter referred to as Arthrobacter A3), a psychrotrophic bacterium, was isolated from the alpine permafrost of the Tianshan Mountains in China [24]. It has an optimal growth temperature of 20 0C, but can survive and grow at near-freezing temperatures as low as -4 0C. Its stress tolerance is in part due to synthesis of trehalose catalyzed by OtsA/B [25].
As in Escherichia coli [26], the otsAB genes in Arthrobacter A3 are arranged as an operon [25]. This organisation allows for efficient co-regulation of both genes [27]. OtsA of E. coli contains an N-terminal loop, located between Arg9 to Gly22, based on the crystal structure. This N-loop is located in the catalytic centre of the OtsA enzyme and interacts with the phosphate moiety of glucose-6-phosphate and the distal phosphate of UDP-glucose, respectively [28]. Furthermore, both ends of the amino acid sequence of the N-loop are conserved in many microorganisms. During the catalytic reaction, the N-loop undergoes significant conformational changes [29], suggesting that the N-loop is directly related to the catalytic efficiency of OtsA. The enzymatic activity of OtsA of Arthrobacter A3 at low temperatures is due to a very flexible N-loop containing the active site [25], a key feature that distinguishes the protein from its E. coli counterpart.
Here we demonstrate that depletion of OtsA or T6P results in constitutive myceloid growth. Further analyses indicate that OtsA doubles as a novel self-assembling morphogenetic protein. OtsA, acting as an osmotic stress sensor together with T6P, mediates the switch to myceloid growth during osmotic stress. Recruiting a biosynthetic enzyme for this morphogenetic function represents an intriguing adaptation in bacteria that can survive in extreme environments.
An otsA deletion mutant (Ar0002) has significantly reduced intracellular trehalose levels compared to the wild-type strain, Ar0001 (S1A Fig) [24]. We also observed that the mutant exhibits an apparent markedly slower growth rate in low osmolarity medium as determined by optical density (OD600; Fig 1A). Whereas the doubling time for the wild-type was 2.5 h, for the mutant it was 3.3 h, approximating the 3.2 h doubling time of the wild-type grown in salt-amended Luria broth (LB). When early log-phase cells were examined by phase-contrast microscopy, we observed extensive aggregate formation by the mutant, similar to previously reported myceloids formed by other Arthrobacter species after osmotic stress [30]. Hence the mutant fails to grow by snap division, but instead adopts non-separating myceloid growth, characteristic of the morphological switch of the wild-type when subjected to salt stress. Consequently, OD600 measurements do not necessarily reflect slower growth of the mutant or wild-type subject to salt stress, but simply the growth of cell aggregates. We subsequently employed OD600 measurements as a proxy for measuring the formation of myceloids, verifying the presence of cell aggregates in early log-phase cultures by phase-contrast microscopy. Quantification and analysis of aggregate dimensions of early log-phase cells of the otsA mutant grown in chemically-defined minimal medium revealed 58% of colony-forming units existing as myceloids with a maximum dimension between extremities of the aggregates of >4 μm (as viewed in two dimensions under the microscope), with the average value being 4.34 μm (n = 385; Fig 1B). In contrast, the proportion of wild-type cell aggregates of greater than 4 μm maximum dimension formed was 21% of the total, with 79% of colony forming units being single cells or small multiples of 2 to 4 cells still joined prior to snap division (Fig 1B). The aggregates of the mutant resembled those formed after 16 h growth by osmotically stressed cultures of the wild-type which have an average maximum dimension of 6.4 μm (n = 200), with over 80% of aggregates being > 4 μm in maximum dimension (Fig 1B). Similar proportions of cell aggregates were observed during growth in LB medium with or without addition of salt (S2 Fig). Cells of the wild-type, salt-induced wild-type myceloids and constitutive myceloids of the mutant were examined by scanning electron microscopy (Fig 1C–1E). Although many single or small chains of non-separated cells of the wild-type grown in LB appeared normal (Fig 1C, bottom left), we also detected a small proportion of cells in chains with branches emerging from their lateral walls (e.g. Fig 1C, top panel, arrowed). The much larger constitutive or salt-induced myceloids consisted of networks of extensively branched chains of cells (Fig 1D and 1E). Genetic complementation of the mutant with a single copy of otsA under control of its native promoter sequence restored both the normal growth rate and the largely bacillary morphology of log-phase bacterial cells grown in LB (S3 Fig), whereas complementation with the flanking genes, otsB or dsbA, or E. coli otsA (otsAEc) did not restore bacillary growth (S3 Fig), although the latter gene was biochemically functional (see below).
To examine a possible link between intracellular trehalose and growth morphology, cultures of the ΔotsA mutant were subjected to a biochemical complementation test. Addition of between 0.5 mM and 4.0 mM trehalose, which is effectively taken up by the bacteria ([24], S1A Fig) had no effect on growth rate (Fig 1A) or myceloid formation. In addition, when a trehalase enzyme encoded by treF was over-expressed in the wild-type (strain Ar0008), resulting in a more than 5-fold reduction in intracellular trehalose to less than that detected in the ΔotsA mutant (S1 Fig), there was only a modest reduction in growth rate (Fig 1A). The growth rate was reflected in a low proportion (24%) of cell aggregates with maximum dimension > 4 μm (Fig 1B). We also determined that trehalose biosynthesis in the ΔotsA mutant was restored due to complementation by otsAEc (S1 Fig). Consequently, we concluded that the constitutive myceloid formation of the ΔotsA mutant did not reflect a reduction of intracellular trehalose. We also observed that the ΔotsA mutant is osmotic stress- sensitive and whereas this phenotype could be rescued by genetic complementation, it could not be by biochemical complementation with trehalose (S4 Fig).
We constructed a strain, wild-type + up-otsA, containing the gene fused with a strong promoter on a multi-copy plasmid. Overexpression was verified by western blot, indicating an approximate 10-fold greater intracellular abundance of the protein relative to FtsZ (Fig 2A). We used a fluorescent derivative of vancomycin (fluo-vancomycin) that binds to nascent peptidoglycan to establish firstly if, as in other studied actinobacteria, growth is at the cell poles, and secondly to visualise how overexpression of OtsA affects cell wall biosynthesis. The antibiotic bound to nascent peptidoglycan at the poles, confirming apical growth, and, with more intense fluorescence, at the newly forming septum in the midcell region of rod-shaped wild-type cells (strain Ar0003) growing in LB (Fig 2B). In cells with evidence of septum formation, the distance between the stained poles and midcell was on average 0.7 μm. There was evidence for some ‘mini-chains’ of cells, for example the 4 joined cells in the top panel of Fig 2B, due to inefficient snap division. The result of OtsA over-expression (wild-type + up-otsA; strain Ar0004) was a pattern of peptidoglycan synthesis consistent with the formation of multiple septa in very long, enlarged cells with bulbous poles and limited branching (Fig 2C). Staining these cells with both DAPI and fluo-vancomycin revealed that many of the newly-formed compartments possessed less intensely staining nucleoids, with a small proportion lacking detectable DNA (arrowed in the overlay image, Fig 2C). This indicates that overexpression of OtsA affects the coordination of DNA synthesis and chromosome segregation with septum formation. The doubling time of this OtsA overexpression strain was 3.1 h (Fig 1A), indicating that these observed abnormalities in cell division can retard growth.
These results, indicating that OtsA function has a role on growth and division, prompted us to test the effects of overexpression of OtsAEc. Over-expression of C-terminal His-tagged OtsAEc, verified by western blotting, had no effect on septum formation or the rod-shaped morphology of the strain (wild-type + up-otsAEc; Ar0010) grown in LB (Fig 2D and 2E). We tested the activity of the overexpressed OtsAEc in Arthrobacter A3 by analysis of the trehalose concentration in Ar0010; this strain had almost 5 times greater intracellular trehalose than the wild type strain (S1C Fig). Moreover, whereas the strain overexpressing OtsA was sensitive to osmotic stress, the strain overexpressing OtsAEc was not (S4 Fig). Consequently, we inferred that specific features of the Arthrobacter protein confer its function as a morphogenetic determinant.
To investigate the relationship between OtsA enzyme activity and its role as a morphogenetic protein and effector of the osmotic stress response, an amino acid substitution was introduced in the active site, as determined by crystallography of the corresponding E. coli enzyme [29], replacing the conserved arginine residue (R36) with alanine (the arginine residue of the E. coli protein is involved in glucose-6-phosphate binding). The mutant protein, OtsAR36A, was overexpressed in strain Ar0011 (wild-type + up-otsAR36A) at similar levels to OtsA in strain Ar0004 (Fig 3A). However, the Ar0011 strain had a much shorter doubling time compared to strain Ar0004, and similar to that of the wild-type (Fig 1A). Staining with fluo-vancomycin revealed single septa located at the midcell of dividing bacillary-form cells (Fig 3B), together with evidence of occasional foci located in the lateral walls (indicated by white arrows in the NCW image). The mutant protein was also expressed under control of the native promoter in the ΔotsA mutant (ΔotsA + otsAR36A; strain Ar0115). It could not restore normal bacillary growth to the mutant (S1 Fig). Moreover, when the mutant protein was overexpressed this did not affect bacillary growth or sensitivity to osmotic stress (S4 Fig).
We hypothesised that the loss of a morphogenetic function of OtsAR36A in vivo could reflect an inability to synthesise a threshold concentration of trehalose-6-phosphate (T6P) that may be necessary for the morphogenetic function of OtsA. To examine this we monitored intracellular levels of T6P in wild-type cells before and after salt stress. Based on a cell volume of 10−15 l, the T6P concentration can be estimated as ranging from approximately 2 mM in the wild-type (no salt stress) to 0.2 mM in the otsA mutant. An approximate 40% decrease in the intracellular concentration of the metabolite was noted in wild-type cells 3 h after salt stress (Fig 3C, bars 2 and 3, compared to bar 1), coincident with the time when we noted changes in cell morphology (see below). To examine this further, we overexpressed the E. coli treC gene [31] encoding T6P hydrolase in the wild-type (wild-type + treCEC; strain Ar0012). The recombinant strain grew slowly and formed constitutive myceloids in the absence of salt-stress (Fig 1A and see below, Fig 4C). These cell aggregates had an average maximum dimension between extremities of 5.4 μm (n = 361), similar to those formed by the ΔotsA null mutant strain. Measurements of intracellular T6P revealed a significant depletion (17% of the level in the wild-type) of this metabolite in Ar0012 compared with the wild-type (Fig 3C, bar 4), indicative of functional activity of TreCEc in Arthrobacter. The reduced level of T6P in Ar0012 was comparable to the amount detectable in the ΔotsA mutant strain (Fig 3C, bar 7). We also compared the levels of this metabolite in the strains overexpressing OtsA and OtsAR36A. Whereas the former strain, Ar0004, contained 137% of the amount in the wild-type (bar 6), reflecting increased synthesis due to amplification of the enzyme, the latter had levels similar to wild-type, reflecting expression of the single-copy wild-type gene in this strain and an absence of metabolic activity due to the active site mutation in the overexpressed enzyme (Fig 3C bar 5; relative to OtsA, purified OtsAR36A exhibited 7.53% +/- 0.22 enzyme activity). Quantification of T6P in strain Ar0010 overexpressing OtsAEc revealed 118% of the levels found in the wild-type (Fig 3C, bar 8), implying that an increase in the metabolite in the absence of increased amounts of OtsA protein is insufficient to promote any change in cell morphology. Moreover, expression of otsAEc in the ΔotsA mutant strain restored T6P synthesis (bar 9) but, as described above, did not change the constitutive myceloid phenotype of the mutant.
We used fluo-vancomycin to stain myceloids. Due to the extensive three-dimensional structure of myceloids of the ΔotsA mutant (strain Ar0002), fluorescence microscopy was more challenging and better resolution images were obtained with smaller cell aggregates. In these myceloids we observed irregular peptidoglycan synthesis, with most staining associated with adjacent poles of contiguous non-separated cells, the sites of joined cells being evident as cell envelope constrictions in the corresponding differential interference contrast images (as indicated by black arrows, S5A Fig). In many long cells (of average 1.1 μm length, examples indicated by white arrows in the corresponding overlay image), there was no observable nascent peptidoglycan in the midcell region. A similar picture emerged when salt-induced myceloids of the wild-type were examined. Cultures were grown to early log phase (20 h) in LB amended to a final concentration of 0.57 M NaCl. Fluo-vancomycin staining revealed irregular patterns of nascent peptidoglycan, with a reduced frequency of midcell peptidoglycan synthesis evident in longer cells (indicated by white arrows in the overlay image, S5B Fig). The myceloids of the strain expressing TreCEc (strain Ar0012) also showed evidence of reduced synthesis of peptidoglycan at the midcell with evidence of longer cells (indicated by white arrows in S5C Fig).
In addition, to examine how salt stress affected the transition from bacillary to myceloid growth of the wild-type in a time-course, cultures were grown to early log-phase in LB and subsequently in LB amended to a final concentration of 0.57 M NaCl, sampled at successive time-points and stained with fluo-vancomycin. In non-amended medium, after 3h, the proportion of cells scored with midcell peptidoglycan synthesis was 62% (n = 480), whereas with salt-stress, the percentage was reduced to 49% (n = 280). In addition, in cells from non-amended medium, we also observed one or two foci of peptidoglycan synthesis in the lateral walls in 39.4% of cells (n = 513); these foci are likely sites for growth of branches, consistent with the tendency for the wild-type to form occasional emerging branches as observed in SEM images (Fig 1C). After 3 h salt-stress, there was an increase to 66.5% (n = 524) of the proportion of cells with foci of nascent peptidoglycan in the lateral walls.
Using dynamic light scattering (DLS), we examined the native state of Arthrobacter OtsA expressed and purified from E. coli. This revealed two populations of the protein: smaller assemblies with an average diameter of 68.86 nm and much larger assemblies with an average diameter of 1738.42 nm (Fig 4A). The protein was reanalyzed after denaturing the multimeric forms in 4 M urea followed by dialysis. If all urea was removed prior to DLS, during the dialysis process (16 h) all the protein self-assembled into two populations of multimeric forms with similar average diameters to those of assemblies prior to denaturation (S6 Fig). After denaturing in 4M urea and subsequent dilution to a final concentration of 1M urea, a much slower self-assembly process could be monitored by DLS in real-time (Fig 4B). We used these conditions to then ask if T6P can promote OtsA self-assembly. We observed that increasing concentrations of this metabolite had a dramatic effect on promoting the rate of OtsA polymerization (Fig 4B). Prior to addition of T6P, the average diameter of OtsA was 64.86 nm. After polymerization of OtsA promoted by 500 μM T6P (a physiologically relevant concentration–see above), there was only one population of the protein detected consisting of large assemblies with an average diameter of 2745.58 nm. The implication is that T6P can promote the assembly of OtsA into large protein networks.
As 1mM MgCl2 was used in these assays and magnesium ions are required for OtsA enzyme activity [28], we then examined whether magnesium ions have a role in the self-assembly of OtsA. Purified OtsA was denatured as described above and then dialysed against a phosphate buffer containing 1M urea and no magnesium ions. The protein was then incubated with between 0 and 1mM MgCl2 and assembly monitored in real time using DLS. Little or no assembly was observed in the absence of the ion, whereas addition of 1mM or greater MgCl2 promoted assembly (Fig 4C), although the maximum assembly was much less than that observed in the presence of T6P (Fig 4B).
We also used DLS to analyse the native state of OtsAR36A, revealing protein structures of 26.5 nm average diameter (S7A Fig). Moreover, addition of T6P failed to promote self-assembly of OtsAR36A (Fig 4B). Consequently, we inferred that the lack of any morphogenetic activity of OtsAR36A is not simply due to its loss of enzyme activity but presumably because the amino substitution also affects the protein’s tertiary structure and its ability to both interact with T6P and form large assemblies. DLS analysis of OtsAEc indicated protein structures with a range of sizes, and a modal diameter of approximately 7 nm (S7A Fig). No increase in diameter was observed after addition of T6P (Fig 4B).
A feature of the Arthrobacter OtsA is its enzymatic activity at low temperatures due to a very flexible ‘N-loop’ containing the active site [25], a characteristic that distinguishes the protein from its E. coli counterpart. To test if this flexible N-loop affects assembly formation, we purified and tested the assembly of two more OtsA proteins, OtsAA3mu and OtsAEcmu, which have, respectively, the E. coli N-loop replacing that of the Arthrobacter protein and vice versa. Whereas OtsAA3mu retained the ability to polymerize, albeit less efficiently, OtsAEcmu behaved like OtsAEc with no evidence for self-assembly (S8 Fig). Consequently, we inferred that the N-loop alone is insufficient to promote self-assembly.
We used transmission electron microscopy (TEM) to examine different states of assembly of OtsA. As urea will interfere with negative staining, we chose to dialyse purified OtsA in phosphate buffer lacking magnesium ions. This resulted in depolymerization as evident in the sizes of imaged protein structures which had an average diameter of approximately 60 nm (Fig 4D), consistent with the dimensions of OtsA depolymerized after urea treatment as determined by DLS. Negative-stained OtsAEc, purified the same way, was most abundant as structures of approximately 10 nm diameter (S7B Fig), again consistent with the size of protein structures determined by DLS analysis (see above). TEM of OtsAR36A revealed structures of approximately 25 nm diameter (S7B Fig), consistent with the DLS data for this protein.
We then imaged OtsA after addition of magnesium ions alone or combined with T6P. After incubation for 30 min with 0.1mM MgCl2 we observed the appearance of branched protein filaments of varying lengths, up to approximately 200 nm in length (Fig 4D). After 30 min incubation with 1 mM MgCl2 and 500 μM T6P, very large assemblies could be observed but only with low resolution using negative staining. Consequently, we used positive staining to obtain images of better resolution, as exemplified in Fig 5D, indicating assembly of the protein into large networks of greater than 2000 nm diameter.
To examine localization of OtsA in vivo, a C-terminal translational fusion with mCherry was expressed in Arthrobacter A3 using the native otsA promoter sequence and a single-copy gene fusion integrated at the chromosomal otsA locus (strain Ar0007). Cells expressing the fusion protein grew normally. In addition, morphogenetic functionality was indicated both by the ability of the fusion protein expressed under control of the native promoter to restore normal snap-division growth to the ΔotsA mutant (strain Ar0116; S3 Fig) and by the promotion of the characteristic multiple-septation phenotype in long, enlarged cells when the fusion protein was overexpressed (strain Ar0006, Fig 5B). In Ar0007 cells grown in LB, the majority of the protein assembled at the midcell region (indicated as ‘m’ in the overlay image, Fig 5A) and some at the cell poles (indicated as ‘p’ in the overlay image, Fig 5A), co-localizing at sites of peptidoglycan synthesis as revealed by fluo-vancomycin staining of the same cells (Fig 5A). To examine if osmotic stress affects protein localisation, strain Ar0007 expressing OtsA::mCherry was grown to early log phase in LB, which was then amended to a final concentration of 0.57 M NaCl. After 3 h salt stress, we observed a diffuse distribution of OtsA throughout the cells (Fig 5A), some colocalising with sites of peptidoglycan synthesis at the cell poles, but no longer localized at the midcell region. When the fusion protein was overexpressed it clearly localized to the sites of multiple septum formation, again indicative of a morphogenetic function (Fig 5B).
Coordinating trehalose concentration and morphology requires that pleomorphic Arthrobacter cells can detect osmotic stress and communicate this information to the cell division apparatus. Here we describe an unexpected role for the trehalose synthase protein OtsA, which doubles as a morphogenetic protein, acting as a direct link between trehalose synthesis and cell morphology, and effecting the transition from bacillary growth to the development of myceloids. A product of the OtsA enzyme, T6P, can function as a signaling proxy for osmotic stress but is insufficient itself to direct changes in cell morphology as evidenced from the lack of any morphogenetic function of the otherwise enzymatically active OtsAEc.
In vitro, OtsA can self-assemble to form elaborate protein networks, this assembly being promoted by T6P. In vivo, when cells are growing in low medium osmolarity, we hypothesise that these protein networks have a morphogenetic cytoskeletal function in promoting normal cytokinesis leading to a bacillary growth-style. Indeed, in non-stressed cells the protein assembles at the midcell and poles, consistent with this hypothesis. We have analyzed the morphological outcomes of various permutations of the genetic background of Arthrobacter that affect either or both intracellular T6P and OtsA concentrations (Table 1). Overexpression of OtsA, resulting in increased T6P, leads to increased formation of septa and loss of coordination of cytokinesis. Moreover, the overexpressed protein localizes at the multiple sites of septum formation in filamentous cells. But an increase in intracellular T6P, due to overexpression of OtsAEc, is insufficient itself to cause aberrant cytokinesis. Salt stress leads to a depletion of T6P and this likely affects the dynamics of OtsA self-assembly in vivo, resulting in the observed diffuse cytoplasmic distribution of the protein in salt-stressed cells and the reduction of peptidoglycan assembly at the midcell, but increasing the likelihood of the emergence of branches from lateral cell walls. These effects, coupled with a reduction in snap-division frequency, presumably lead to the growth of myceloids. Additional evidence that T6P can act as an intracellular proxy for salt stress comes from experimentally depleting this metabolite by expression of E. coli T6P hydrolase. A consequence of this depletion is that the strain can only grow with a myceloid morphology, despite expressing normal levels of OtsA.
The psychrotrophic nature Arthrobacter A3 is to an extent due to its ability to accumulate trehalose as a cryoprotectant, which in turn is due to the enzymatic activity of OtsA at low temperatures [25]. This activity is due to a very flexible ‘N-loop’ containing the active site [25], a key feature that distinguishes the protein from its E. coli counterpart. However, exchanging the N-loops of the Arthrobacter and E. coli proteins indicated that this sequence alone is not responsible for the self-assembly characteristic of the former, although the Arthrobacter protein with an E. coli N-loop is less proficient at self-assembly compared to its wild-type counterpart. In the genetic backgrounds in which either OtsAR36A or OtsAEc are overexpressed and T6P levels are normal or increased, these proteins have no effect on cytokinesis. The lack of in vivo morphogenetic and in vitro self-assembly activities of either OtsAR36A or OtsAEc can be rationalized in part by the active site mutation of the former and reduced N-loop flexibility of the latter, both of which likely affect T6P binding. We are currently comparing the structural properties of these variant proteins to identify other features that contribute to the ability of Arthrobacter OtsA to self-assemble.
Our analysis of published Arthrobacter genome sequences indicates that these bacteria lack typical actin-like or intermediate filament cytoskeletal proteins found in other bacteria, including some other actinobacteria. The evolutionary recruitment of the principle biosynthetic enzyme involved in synthesis of the osmoprotectant trehalose to a function that regulates morphology in response to osmotic stress is an intriguing adaptation to coping with extreme environments. This adds to a few other known examples of biosynthetic enzymes co-opted for morphogenetic roles in bacteria. The primary enzyme involved in CTP synthesis, CtpS, from Caulobacter, E. coli and several eukaryotic species can self-assemble into linear filaments [32–34] and in Caulobacter this protein has a role in determining cell shape [32]. In the same bacterium, a NAD(H)-binding oxidoreductase, KidO, can inhibit Z-ring formation [35], linking cell division with metabolic status. In Bacillus subtilis the membrane-associated glucosyltransferase UgtP, involved in glycolipid biosynthesis, acts as a metabolic sensor governing cell size. During growth in rich media, when ugtP expression is upregulated, the enzyme localises to the midcell division site and inhibits Z-ring formation at the midcell [36]. In E. coli, a non-homologous glucosyltransferase OpgH, an integral inner membrane protein that is functionally analogous to UgtP of B. subtilis in linking cell size with central metabolism, is believed to inhibit Z-ring formation by a different mechanism involving sequestering FtsZ [37]. In these latter two examples, the bacteria utilise UDP-glucose as an intracellular signal and proxy for nutrient availability. In contrast to these examples, in Arthrobacter the OtsA glucosyltransferase doubles as a morphogenetic protein and determine cell morphology in response to an environmental signal.
Although several other bacterial species are known to exhibit morphological plasticity as a stress survival strategy, switching from a rod-shaped to a filamentous morphology [38], the mechanism for this transition, when known, is quite different. In E. coli, the product of an SOS-induced gene, SulA, binds to FtsZ monomers, inhibiting polymerization [39]. In older Caulobacter cells, depletion of FtsZ leads to filamentation [40]. It will be of interest to determine if OtsA in Arthrobacter interacts with FtsZ and to investigate the properties of OtsA from other actinobacteria, including M. tuberculosis which undergoes filamentation in macrophages [41].
The growth conditions for Arthrobacter strains and E. coli cultures are given in detail in the Supplemental protocols. The list of all relevant strains and plasmids is provided in S1 Materials and Methods. Details of how each plasmid and strain were constructed, and the primers used for amplification and cloning of DNA sequences are also provided in the Supplemental protocols.
All strains were analyzed in exponential growth phase unless otherwise stated. Nascent cell walls were stained using fluo-vancomycin as described previously [42]. Bacterial cells were stained by DAPI (0.1ug/ml, PBS), vancomycin (1ug/ml, PBS) and fluo-vancomycin (1ug/ml, PBS) for 20min. Cultures were then washed in PBS, and suspended in 1.6% formaldehyde (in PBS) and left on ice for 1 hr. Treated cells were distributed on microscope slides that had been treated with 0.1% (wt/vol) poly-L-lysine (Sigma). Images were acquired on a confocal laser scanning microscope (Olympus FV1000).
The purification of proteins, immunoblotting and coimmunoprecipitation analysis were carried out as described previously [24, 25].
The assembly of OtsA was monitored in real-time with a dynamic light scattering assay using a Brookhaven Instruments BI-200SM system (USA). The wavelength of the stable argon ion laser was 532 nm. The assay was performed at 20°C. OtsA was denaturated in 4 M urea for 5 min and then diluted in polymerization buffer O (20mM Tris HCl, pH 7.5, 1 mM MgCl2), with different concentrations of T6P, to 10 uM final concentration of OtsA and 1M of urea. To measure the effect of magnesium ions, MgCl2 was excluded from buffer O. The intensity of scattered light was measured at an angle of 90°.
Negative staining electron microscopy was used to visualize OtsA. Carbon-coated copper grids (400 mesh, Electron Microscopy Sciences) were glow discharged for 5 s before use. Before applying OtsA, a drop of 0.2 mg/ml cytochrome c was pipetted onto the carbon, incubated for 30 s, and then blotted with filter paper. A drop of OtsA solution was then applied to the carbon and incubated for 10 s before the excess was blotted. The grid was immediately rinsed with 3–4 drops 2% uranyl acetate, blotted, and air-dried. For positive staining, 5 ul of protein (10 μM) were placed on carbon coated grids, incubated for 2 min, washed in buffer A (20 mM sodium phosphate, pH 7.5 and 20 mM NaCl), and stained with 2% uranylacetate for 30 s. Protein were visualized and photographed using a Tecnai-G2-F30 electron microscope. To image cells, exponentially growing Arthrobacter strains were fixed in 4% paraformaldehyde, 2.5% glutaraldehyde (PBS, PH 7.4), at 25°C for 2 h. SEM of whole cells were carried out as described previously [43].
Trehalose-6-phophate phosphatase was obtained from E. coli as described [44]. To measure trehalose-6-phosphate, cells were broken using ultrasonication at 4 0C and extracts prepared and assayed as described [45]. For each strain, four replicate assays were conducted. Trehalose was assayed as previously described [24].
|
10.1371/journal.pcbi.1003781 | Logarithmic and Power Law Input-Output Relations in Sensory Systems with Fold-Change Detection | Two central biophysical laws describe sensory responses to input signals. One is a logarithmic relationship between input and output, and the other is a power law relationship. These laws are sometimes called the Weber-Fechner law and the Stevens power law, respectively. The two laws are found in a wide variety of human sensory systems including hearing, vision, taste, and weight perception; they also occur in the responses of cells to stimuli. However the mechanistic origin of these laws is not fully understood. To address this, we consider a class of biological circuits exhibiting a property called fold-change detection (FCD). In these circuits the response dynamics depend only on the relative change in input signal and not its absolute level, a property which applies to many physiological and cellular sensory systems. We show analytically that by changing a single parameter in the FCD circuits, both logarithmic and power-law relationships emerge; these laws are modified versions of the Weber-Fechner and Stevens laws. The parameter that determines which law is found is the steepness (effective Hill coefficient) of the effect of the internal variable on the output. This finding applies to major circuit architectures found in biological systems, including the incoherent feed-forward loop and nonlinear integral feedback loops. Therefore, if one measures the response to different fold changes in input signal and observes a logarithmic or power law, the present theory can be used to rule out certain FCD mechanisms, and to predict their cooperativity parameter. We demonstrate this approach using data from eukaryotic chemotaxis signaling.
| One of the first measurements an experimentalist makes to understand a sensory system is to explore the relation between input signal and the systems response amplitude. Here, we show using mathematical models that this measurement can give important clues about the possible mechanism of sensing. We use models that incorporate the nearly-universal features of sensory systems, including hearing and vision, and the sensing pathways of individual cells. These nearly-universal features include exact adaptation-the ability to ignore prolonged input stimuli and return to basal activity, and fold-change detection- response to relative changes in input, not absolute changes. Together with information on the input-output relationship-e.g. is it a logarithmic or a power law relationship-we show that these conditions provide enough constraints to allow the researcher to reject certain circuit designs; it also predicts, if one assumes a given design, one of its key parameters. This study can thus help unify our understanding of sensory systems, and help pinpoint the possible biological circuits based on physiological measurements.
| Biological sensory systems have been quantitatively studied for over 150 years. In many sensory systems, the response to a step increase in signal rises, reaches a peak response, and then falls, adapting back to a baseline level, (Fig. 1a upper panel). Consider a step increase in input signal from to , such that the relative change is . There are two commonly observed forms for the input-output relationship in sensory systems: logarithmic and power law. In the logarithmic case, the relative peak response of the system is proportional not to the input level but to its logarithm . A logarithmic scale of z versus I, namely , is often called the Weber-Fechner law [1], and is related but distinct from the present definition . In the case of a power-law relationship, the maximal response is proportional to a power of the input (Fig. 1a lower panel) [2]. In physiology this is known as the Stevens power law; the power law exponent varies between sensory systems, and ranges between [2]. For example the human perception of brightness, apparent length and electrical shock display exponents respectively.
Both logarithmic and power-law descriptions are empirical; when valid, they are typically found to be quite accurate over a range of a few decades of input signal. For example, both laws emerge in visual threshold estimation experiments [3]. In that study, the logarithmic law was found to describe the response to strong signals and the power-law to weak ones. However the mechanistic origins of these laws, and the mechanistic parameters that lead to one law or the other, are currently unclear. Theoretical studies have suggested that these laws can be derived from optimization criteria for information processing [4], [5], such as accounting for scale invariance of input signals [6]. Both laws can be found in models that describe sensory systems as excitable media [7]. Other studies attempt to relate these laws to properties of specific neuronal circuits [8], [9]. Here we seek a simple and general model of sensory systems which can clarify which mechanistic parameters might explain the origin of the two laws in sensory systems.
To address the input-output dependence of biological sensory systems, we use a recently proposed class of circuit models that show a property known as fold-change detection [10], [11]. Fold change detection (FCD) means that, for a wide range of input signals, the output depends only on the relative changes in input; identical relative changes in input result in identical output dynamics, including response amplitude and timing (Fig. 1b). Thus, a step in input from level 1 to level 2 yields exactly the same temporal output curve as a step from 2 to 4, because both steps show a 2-fold change. FCD has been shown to occur in bacterial chemotaxis, first theoretically [10], [11] and then by means of dynamical experiments [12], [13]. FCD is thought to also occur in human sensory systems including vision and hearing [11], as well as in cellular sensory pathways [14]–[17].
FCD can be implemented by commonly occurring gene regulation circuits, such as the network motif known as the incoherent feed-forward loop (I1-FFL) [10], as well as certain types of nonlinear integral feedback loops (NLIFL) [11]. Recently, the response of an FCD circuit to multiple simultaneous inputs was theoretically studied [18]. Mechanistically, FCD is based on an internal variable that stores information about the past signals, and normalizes the output signal accordingly. We find here, using analytical solutions, that simple fold-change detection circuits can show either logarithmic or power law behavior. The type of law, and the power-law exponent , depend primarily on a single parameter: the steepness (effective Hill coefficient) of the effect of the internal variable on the output.
We begin with a common gene regulation circuit [19] that can show FCD, the incoherent type 1 feed-forward loop (I1-FFL) [10]. In transcription networks, this circuit is made of an activator that regulates a gene and also the repressor of that gene. More generally, we can consider an input X that activates the output Z, and also activates an internal variable Y that represses Z (Fig. 2). We study a model (Eq. 1, 2) for the I1-FFL with AND logic (that is, where X and Y act multiplicatively to regulate Z), which includes ordinary differential equations for the dynamics of the internal variable Y and the output Z [20]–[22]. We use standard biochemical functions to describe this system [23].(1)(2)The production rate of Y is governed by the input X according to a general input function (in cases where X is a transcription factor, X denotes the active state). The maximal production rate of Y is . The repressor Y is removed (dilution+degradation) at rate (Eq. 1). We assume here that saturating signal of Y is present, so that all of Y is in its active form. The product Z which is repressed by Y and activated by X is produced at a rate that is a function of both X and Y, denoted . An experimental survey of E. coli input functions suggested that many are well described by separation of variables: the two-dimensional input function separates to a product of one dimensional functions, [24], where and are Hill functions (for more explanation see the Methods section). We therefore use a general form for the X dependence, , and multiply it by a repressive Hill function of Y (Eq. 2), with a maximal production rate . The removal rate of Z is . Here we consider step input functions in which X changes rapidly from one value to another. The values of and is determined by the step size in input.
For clarity, upper case letters relate to the elements in the circuit and lower case letters describe normalized model variables. The two-equation model (Eq. 1, 2) has 6 parameters. Dimensional analysis (fully described in Methods) reduces this to three dimensionless parameters (Eq. 3, 4).
The first parameter, , is the normalized halfway repression point of the output, defined by , where is the pre-step steady state level and is the level of Y needed to half-way repress Z. The second parameter is the cooperativity or steepness of the input function, . The final parameter is the ratio of decay rates of Z and Y, . The normalized variables, and , are the new dimensionless variables in the model. Table 1 summarizes the parameters in the model for the I1-FFL.
This model for the I1-FFL describes the response to a step increase in input, starting from fully adapted conditions. We consider a change between an input level of , to a new level . The step is thus characterized by the fold change F equal to the ratio between the initial and final input levels, .
In order for FCD to hold, the production rate of Z must be proportional to (), where the power law exponent is the same as the Hill coefficient that describes the steepness of the input function. In this way, the internal variable, Y, can precisely normalize out the fold change in input (see Methods). The model thus reads:(3)(4)The higher , the more Y is needed to repress Z. The parameter - the Hill coefficient of the input function - is important for this study, and determines the steepness of the regulation of the output Z by the internal variable Y (Fig. 2). The higher the more steep the repression of Z by Y. The limit resembles step-like regulation. Biochemical systems often have Hill coefficients in the range [23]. The ratio between the removal rates, , describes the relative time scale between Y and Z. For , Y and Z have the same removal rates, and for , the output Z is much faster than Y.
Goentoro et al. [15] showed, using a numerical parameter scan, that this circuit can perform FCD provided that threshold of the Z repression, , is small: that is . We therefore further analyze the limit of , meaning strong repression of Z, where the equation for the product Z (Eq. 4) becomes:(5)In this limit, the system exhibits fold change detection since it obeys the sufficient conditions for FCD in Shoval et al (2010) (see Methods). We analytically solved the model (Eqs. 3, 5), in the limit of small , for all values of , with initial conditions corresponding to steady state at the previous signal level, (in the limit ). The solution (derived in Methods) is a decaying exponential multiplied by a term that contains a Beta function (Fig. 3a):(6)where the Beta function is . The dynamics of the output z shows a rise, reaches a peak , and then falls to the pre-signal steady state (Fig. 3a). At the solution is approximately linear with a slope that depends on F, and :(7)At the solution decays exponentially:(8)As in all FCD systems, exact adaptation is found. The error of exact adaptation, goes as and vanishes at .
We explored how three main dynamical features depend on the input fold change F and the dimensionless parameters and . The first feature is the amplitude of the response, defined as the maximal point in the output z dynamics, . The second dynamical feature is the timing of the peak, . The third feature is the adaptation time, [25], [26] which we define as the time it takes z to reach halfway between and its steady state (Fig. 3a). We denote as the relative change in the input signal, and as the relative maximal amplitude of the response. Since has only mild effects, we discuss it in the last section, and begin with , namely equal timescales for the two model variables.
We tested the effects of cooperativity in the input function, , on the dynamics of the response. Cooperativity seems to have a weak effect on the timescales of the response: The adaptation time and the peak time decrease mildly with the fold F. For , the analytical solution of the time of the peak for all values of is: (see derivation in Methods). Substituting the corresponding relative response, , we receive a mildly decreasing function (Fig. 3b).
In contrast to the mild effect of cooperativity on timescales, cooperativity has a dramatic effect on the response amplitude. The maximal amplitude of the output z relative to its basal level, , increases with the fold and behaves differently for each . For low steepness, , increases in an approximately logarithmic manner with (for ), (normalized root-mean-square deviation, for fitting to compared to for fitting to - see Methods). More precisely the analytical solution is (see Methods) (Fig. 4a). The function is defined as the solution to the equation . The productlog function is approximately linear at , and approximately at .
For , the peak response increases linearly . For , the increase is approximately quadratic, (Fig. 4b). We find that for any , the increase is approximately a power law with exponent in the limit of large : (see Methods) ( for fitting to compared to for fitting to for respectively). Note that the pre-factor in the power law is also predicted to depend simply on the Hill coefficient for , namely to be equal to (for ). Indeed in fitting the numerical solution the best fit parameter is approximately : for respectively. The dependence of output amplitude on input fold-change is thus a power law, similar to Stevens power law, except for where the output dependence is logarithmic.
One point to consider regarding step input functions is that realistic inputs are not infinitely fast steps; however, a gradual change in input behaves almost exactly like an infinitely rapid step, as long as the timescale of the change in input is fast compared to the timescale of the Y and Z components. To demonstrate this, we computed the response to changes in input that have a timescale parameter that can be tuned to go from very slow to very fast: (Fig. 5a). When , the behavior of the relative maximal amplitude of the response, , as a function of the relative change in the input signal, , is very similar to the infinitely fast step solution (less than 5% difference for and , Fig. 5b). When the change in input is much slower than the typical timescales of the circuit, the response is very small, since the signal is perceived almost as a steady-state constant. For slow changes in input, the I1-FFL response can be shown to be approximately proportional to the logarithmic temporal derivative of the signal [27]–[30].
In addition to the I1-FFL mechanism, a non-linear integral feedback based mechanism (NLIFL) for FCD at small values of has been proposed by Shoval et al [11] (see Methods section) (Fig. 6a). This mechanism is found in models for bacterial chemotaxis [28]. The full model is described by:(9)(10)Its dimensionless equations following dimensional analysis (fully described in Methods) are:(11)(12)Where the new variables are: , and the dimensionless parameters are defined as: and (Methods). Table 2 summarizes the parameters in the model for the NLIFL.
We solved the NLIFL model (Eqs. 11, 12) numerically for the limit and find that the maximal response increases with the relative change in the signal in a power-law manner, (Fig. 6b). The best-fit power law exponents increase with , namely at for . A dependence does not fit the data at ( for fitting to compared to for fitting to for respectively). To a good approximation, the power law is linearly related to the steepness parameter , by (Fig. 6c).
The time scales in this circuit seem to decrease faster with the fold F for than in the I1-FFL case, where and at (Fig. 6d, all the fits of have ).
Given the results so far, one can use the present approach to rule out certain mechanisms. If one observes a logarithmic dependence, one can draw at least two conclusions: (i) the NLIFL model addressed here can be rejected, (ii) if the I1-FFL model addressed here is at play, its steepness coefficient is .
If one observes a linear dependence of input on output, the I1-FFL and NLIFL mechanisms cannot be distinguished. The steepness can be inferred to be about for both circuits.
We applied the present approach to data from Takeda et al [17] on Dictyostelium discoideum chemotaxis. In these experiments, the input is cAMP steps applied to cells within a micro-fluidic system, and the output is a fluorescent reporter for Ras-GTP kinetics. The output showed nearly perfect adaptation and FCD-like response to a wide range of input cAMP steps. We re-drew the peak amplitude (Fig. 7a) and the time of peak (Fig. 7b) as a function of the added cAMP concentrations and find that it is well described by the analytical solution of the maximal response and time of peak for an I1-FFL circuit with . The peak amplitude () as a function of the relative input is well described by a logarithmic relationship (mean-square weighted deviation, for fitting the data to considering the error-bars – see Methods). Fitting it to a power law results in a small exponent () (Fig. 7c). Such a small power law exponent can only be obtained with a negative cooperativity in the NLIFL model considered here. Such negative cooperativity is rare in biological systems [31], [32]. If we consider only positive cooperativity (), as found in most biological systems, the NLIFL model considered here provides a poor fit to the data () (Fig. 7c).
Thus, the present analysis is most consistent with an I1-FFL mechanism considered here with . The same is found when plotting the observed time-to-peak () versus the analytical solution of the I1-FFL model () with ( for fitting to ) (Fig. 7d). This agrees with the numerical model fitting performed by Takeda et al, who conclude that an I1-FFL mechanism is likely to be at play (they used in their I1-FFL model, which is based on degradation of component Z by Y, rather than inhibition of production of Z by Y as in the present model).
In this analysis we assumed that the experimentally measured fluorescent reporter is in linear relation to the biological sensory output, Ras activity. If this relation turns out to be nonlinear, the conclusions of this analysis must be accordingly modified.
In the eukaryotic chemotaxis system, the two model variables Y and Z have similar timescales (). We also studied the effect of different timescales (), and find qualitatively similar results. A logarithmic dependence of amplitude on F is found when , and a power law when . The power law increases weakly with (Fig. 8a). In the limit of very fast Z (), the solution approaches an instantaneous approximation (obtained by setting ) in which the power law is instead of (Fig. 8b). There is a cross over from the Stevens power law when , to the instantaneous model power law when (Fig. 8c). An analytical solution that exemplifies this crossover can be obtained at , where (Methods). Because of the limit behavior of the productlog function mentioned above, at small fold values , and at large values . In summary, the instantaneous approximation, commonly used in biological modeling, must be done with care in the case of FCD systems.
This study explored how two common biophysical laws, logarithmic and power-law, can stem from mechanistic models of sensing. We consider two of the best studied fold-change detection mechanisms, and find that a single model parameter controls which law is found: the steepness of the effect of the internal variable on the output. We solved the dynamics analytically for the I1-FFL mechanism, finding that logarithmic-like input-output relations occurs when , and power-law occurs when , with power law , and prefactor at . The nonlinear integral feedback loop (NLIFL) mechanism - a second class of mechanisms to achieve FCD - can only produce a power law. Thus, if one observes logarithmic behavior, one can rule out the specific NLIFL mechanism considered here. This appears to be the case in experimental data on eukaryotic chemotaxis [17], in which good agreement is found to the present results in the I1-FFL mechanism with in both peak response and timing.
This theory gives a prediction about the internal mechanism for sensory systems based on the observed laws that connect input and output signals. Thus, by measuring the system response to different folds in the input signal one may infer the cooperativity of the input function and potentially rule out certain classes of mechanism. For example, if a linear dependence of amplitude on fold change is observed (power law with exponent ), one can infer that the steepness coefficient is about for both the specific I1-FFL and NLIFL circuits considered here, with slight modification if the timescales of variables are unequal. Such a linear detection of fold changes may occur in drosophila development of the wing imaginal disk [33]–[35].
The problem of finding the FCD response amplitude shows a feature of technical interest for modeling biological circuits. In many modeling studies, a quasi-steady-state approximation, also called an instantaneous approximation, is used when a separation of timescales exists between processes. In this approximation, one replaces the differential equation for the fast variables by an algebraic equation, by setting the temporal derivative of the fast variable to zero. This approximation results in simpler formulae, and is often very accurate, for example in estimating Michaelis-Menten enzyme steady states [36]. However, as noted by Segel et al [36], this approximation is invalid to describe transients on the fast time scale. In the present study, we are interested in the maximal amplitude of the FCD circuits. In some input regimes, namely , the instantaneous approximation predicts an incorrect power law. To obtain accurate estimates, the full set of equations must be solved without setting derivatives to zero.
It would be interesting to use the present approach to analyze experiments on other FCD systems, and to gain mechanistic understanding of sensory computations.
Consider a general partition function for an input function with an activator and a repressor: . The regime in which FCD applies is that of strong repression, and non-saturated activation [10]. In this limit, , and is thus well approximated by a product.
More generally, G(X,Y) is a product of two functions whenever binding is independent, , which occurs when the relation holds. The biological meaning of the relation is that X and Y bind the Z promoter independently so that the probability of X to bind the promoter and the probability of Y to unbind equals the multiplication of the probabilities:
In the NLIFL case, one can show from the MWC model chemotaxis by Yu Berg et al [28] that in the FCD regime it is simply a power law.
We performed dimensional analysis of the full model of the I1-FFL (Eq. 1, 2) by rescaling as many variables as possible. The rescaled variables:(M1)Where is the pre-signal steady state of Y, derived by taking : , and is the steady state of Z derived by taking . Substituting these rescaled variables we receive:(M2)Since we assume that is determined by the step size in input, we can consider merely the fold change F in input, . For FCD to hold we consider . Defining the rescaled repression threshold we receive in the new rescaled variables (lower case letters y and z):(M3)Rescaling the time to and defining yields to Eq. 3, 4 in the main text.
We also performed dimensional analysis of the full model of the NLIFL (Eqs. 9, 10) by rescaling as many variables as possible. The rescaled variables:(M4)Where is the pre-signal steady state of Y, derived by taking and assuming : , and . Substituting these rescaled variables we receive:(M5)After algebraic manipulation and in the new rescaled variables (lower case letters y and z):(M6)We consider here also .
Rescaling the time to and defining yields to Eqs. 11, 12 in the main text.
Given a set of ordinary differential equations with internal variable y, input F and output z:(M7)(M8)According to Shoval et. al. (2010), FCD holds if the system is stable, shows exact adaptation and g and f satisfy the following homogeneity conditions for any :(M9)(M10)In the model for I1-FFL (Eq. 3, 4) at the limit of strong repression :Exact adaptation also holds at , . This holds also for the NLIFL (Eqs. 9, 10).
The solution for y is an exponent:(M11)The general solution for the ODE with the initial condition is:(M12)For our model Eq. M12 reads:(M13)By changing the variable in the integral in Eq. M13: we get:(M14)Which is by definition the solution in Eq. 6.
At the time of peak , therefore from Eq. 5 in the main text we get:(M15)From our definition of the relative response we have:(M16)Substituting the solution of y (Eq. M11) and by algebraic manipulation we receive the analytical solution for :(M17)
The analytical results were derived by taking the derivative of the solution for (Eq. 6 in the main text) and substituting time of the peak (Eq. M17), . This provides an equation for the amplitude of the maximal response, , yielding an intractable equation:(M18)Where we used the identity: . This identity can be easily proven by using the change of variable, , in the integral of the Beta function.
For Eq. M18 becomes:(M19)Using the Series function of Mathematica to expand Eq. M19 in the limit of large and keeping high orders in yields:(M20)Using in the limit of large x we receive:(M21)Taking the exponent of this Eq. M21 yields:(M22)The solution for Eq. M22 is by definition the productlog function: .
For Eq. M18 becomes:(M23)Since , Eq. M23 yields:(M24)By algebraic manipulation Eq. M24 becomes . Taking the exponent of this equation yields:(M25)The solution for Eq. M25 is by definition the productlog function: .
For we define , substituting this new variable into Eq. M18 we have:(M26)Using the Series function of Mathematica for large and yields:(M27)Keeping the highest order in and we receive: . Recall that for large and , and therefore .
For the instantaneous approximation to be true at large , the error, (Fig. 8a), between the maximal amplitude in the instantaneous approximation and the full model should vanish at .(M28)Where decrease with F slower than , therefore with f(F) a monotonic increasing function of F. This proves that even at large , the error increases with F (Fig. 8b) and can be very large.
All the numeric simulations and fits were made in Mathematica 9.0.
The root-mean-square deviation (RMSD) [37] calculated for comparing the goodness of fit between the two models is defined as: .
The data points from Takeda et al were extracted by using the ‘ginput’ function of MATLAB. The fits for the data were made using the NonlinearModelFit function considering the error-bars, , as weights, .
The goodness of fit was tested using the mean-square weighted deviation (MSWD) [37] which sums the residuals (r) - sum of squares of errors with weights of : .
We define logarithmic response as . In contrast, traditional definition of the Weber-Fechner law (also called the Fechner law) in biophysics is (e.g. ref. [3]) as . Thus the present definition concerns relative change in input and output, whereas the Weber-Fechner law concerns absolute input and output. Note also that the Weber-Fechner law is distinct from Weber's law, on the just noticeable difference in sensory systems, whose relation to FCD was discussed in Ref [11].
|
10.1371/journal.pgen.1004511 | Knock-In Reporter Mice Demonstrate that DNA Repair by Non-homologous End Joining Declines with Age | Accumulation of genome rearrangements is a characteristic of aged tissues. Since genome rearrangements result from faulty repair of DNA double strand breaks (DSBs), we hypothesized that DNA DSB repair becomes less efficient with age. The Non-Homologous End Joining (NHEJ) pathway repairs a majority of DSBs in vertebrates. To examine age-associated changes in NHEJ, we have generated an R26NHEJ mouse model in which a GFP-based NHEJ reporter cassette is knocked-in to the ROSA26 locus. In this model, NHEJ repair of DSBs generated by the site-specific endonuclease, I-SceI, reconstitutes a functional GFP gene. In this system NHEJ efficiency can be compared across tissues of the same mouse and in mice of different age. Using R26NHEJ mice, we found that NHEJ efficiency was higher in the skin, lung, and kidney fibroblasts, and lower in the heart fibroblasts and brain astrocytes. Furthermore, we observed that NHEJ efficiency declined with age. In the 24-month old animals compared to the 5-month old animals, NHEJ efficiency declined 1.8 to 3.8-fold, depending on the tissue, with the strongest decline observed in the skin fibroblasts. The sequence analysis of 300 independent NHEJ repair events showed that, regardless of age, mice utilize microhomology sequences at a significantly higher frequency than expected by chance. Furthermore, the frequency of microhomology-mediated end joining (MMEJ) events increased in the heart and lung fibroblasts of old mice, suggesting that NHEJ becomes more mutagenic with age. In summary, our study provides a versatile mouse model for the analysis of NHEJ in a wide range of tissues and demonstrates that DNA repair by NHEJ declines with age in mice, which could provide a mechanism for age-related genomic instability and increased cancer incidence with age.
| DNA damage disrupting both DNA strands, termed double strand breaks (DSBs), poses a threat to cell survival. If repaired inappropriately, such DNA breaks lead to genomic rearrangements, mutations, and ultimately cancer. Nonhomologous end joining (NHEJ) is the major pathway for repairing double-stranded breaks in mammals. Errors associated with NHEJ have been implicated in the aging process because mice with mutations in NHEJ genes exhibit premature aging. It remains unknown, however, whether NHEJ becomes impaired during normal aging. Studies of age-related changes in NHEJ have been hampered by the lack of a mouse model that would allow detection and quantification of NHEJ events. Here we report generation of NHEJ reporter mice containing a GFP-based NHEJ cassette knocked-into the ROSA26 locus. Using this mouse model, we were able to compare NHEJ across different tissues and demonstrate that NHEJ becomes less efficient and more error-prone with age. Our results provide a mechanism for age-related genomic instability and increased cancer incidence with age. The NHEJ reporter mice will be useful for a broad range of studies in the fields of aging and DNA repair.
| The somatic mutation theory of aging posits that the accumulation of unrepaired somatic mutations over time leads to the ‘functional failure’ frequently observed during the course of aging [1]. Cellular DNA is a target of various endogenous and environmental insults, leading to DNA damage, of which double-stranded DNA breaks (DSBs) are the most damaging since they can lead to loss of genetic information via deletions or insertions and chromosomal rearrangements via translocations. Indeed, accumulation of such genome rearrangements have been observed in aged human and mouse tissues [2]–[8]. The appearance of genome rearrangements with age suggests that the process of DSB repair becomes compromised with age.
DSBs are repaired by two major pathways: non-homologous end joining (NHEJ) and homology-directed repair (HDR). NHEJ is faster than HDR [9] and does not require a homologous template, allowing it to take place in both dividing and non-dividing cells. As such, NHEJ is the primary DSB repair pathway in mammals. NHEJ is categorized as canonical NHEJ (c-NHEJ) and alternative NHEJ (alt-NHEJ), also called microhomology-mediated end joining (MMEJ) [10], [11]. c-NHEJ involves core component proteins, including Ku70, Ku80, DNA-PKcs, Artemis, and the Ligase IV complex (Lig IV, XRCC4, and XLF) [12]. NHEJ is also regulated by SIRT6 [13] and Werner syndrome proteins [14]. MMEJ is a DNA-PK-independent repair mechanism, which utilizes a distinct set of proteins, including but not limited to PARP-1, CtIP, XRCC1, and Ligase III [15]–[19]. MMEJ typically uses 5–25 bp microhomology for end-joining. MMEJ is inherently more mutagenic, leading to deletion of the sequences between microhomology regions [20].
The most compelling link between impaired NHEJ and aging is demonstrated by the appearance of accelerated aging in human patients and in mouse models with mutations in genes involved in NHEJ. For example, mice with loss-of-function mutations in Ku70, Ku80, and SIRT6 each exhibit symptoms of premature aging [21]–[25]. Werner syndrome patients display alopecia, atrophy, osteoporosis and die of cardiovascular disease or cancer at age 50 [26]. These mutant studies, however, do not tell us whether NHEJ is affected during normal aging.
Analysis of human lymphocytes and kidney, liver and skin cells from wild type mice showed an increase in the frequency of genome aberrations and mutations with age [2]–[8]. This accumulation of mutations and genomic rearrangements indicates a potential age-associated decline in the ability of somatic cells to repair DNA. Age-related decline in NHEJ in the brain has been suggested by in vitro plasmid rejoining assays in nuclear extracts [27], [28]. Furthermore, studies using human peripheral blood mononuclear cells have demonstrated reduced levels of Ku70 and Ku80 proteins [29]–[31], and a reduced capacity to repair radiation-induced breaks with age, as determined by comet assay [32]. Together, these studies suggest that the capacity to repair DSBs declines with age. However, studies analyzing the efficiency of the complete NHEJ reaction across tissues are missing. We previously measured NHEJ in replicatively senescent human fibroblasts and showed that NHEJ efficiency declines with replicative age [33], [34]. However, since chronological aging is more complex and heterogeneous than in vitro senescence, it remained to be determined whether NHEJ declines during organismal aging.
To study age-related changes in NHEJ, we generated the R26NHEJ mouse model, where a GFP-based reporter cassette is knocked-in to the ROSA26 locus. DSBs in the reporter cassette are generated by the expression of the I-SceI endonuclease. Repair of the breaks by NHEJ restores the functional GFP gene, which is then expressed from a constitutive ROSA26 promoter. The efficiency of NHEJ is quantified by the number of GFP+ cells. We then used this model to compare NHEJ across tissues and to conduct the analyses of age-associated changes in NHEJ repair in brain astrocytes and fibroblasts from heart, kidney, lung, and skin. We show that the efficiency of NHEJ declines 1.8 to 3.8-fold with age, depending on the tissue examined. In addition, heart and lung fibroblasts were found to utilize MMEJ more frequently with age. Our results provide evidence that NHEJ efficiency declines with age and reports a novel mouse model for in vivo studies of NHEJ.
To study NHEJ, we generated the R26NHEJ mouse model where a GFP-based NHEJ reporter cassette was knocked-in to the ROSA26 locus. The NHEJ reporter cassette [35], [36] consists of the GFP gene interrupted by the Pem1 intron and an Ad exon flanked by I-SceI recognition sites (Figure 1A). The GFP gene in the NHEJ cassette is inactivated by the presence of the Ad exon. Upon induction of DSBs by I-SceI, the Ad exon is excised and rejoining of the intron sequences restores the GFP activity. The two I-SceI recognition sequences are in an inverted orientation such that the digestion of both sites generates non-compatible DNA ends (Table S1). NHEJ events can be quantified by the appearance of GFP+ cells. To generate the knock-in mice expressing the NHEJ reporter cassette under ROSA26 promoter, the NHEJ reporter was knocked into the ROSA26 locus following Exon 1. Gene targeting was performed in C57BL/6 mouse-derived embryonic stem (ES) cells (Figure 1B). We chose the C57BL/6 mouse strain because of its well-characterized aging pattern and relative longevity [37]. G418-resistant ES colonies were screened by Southern blot using BamHI digestion (Figure 1C) and injected into mouse blastocysts. Chimeric males were obtained and mated with C57BL/6 females and the resulting offspring were genotyped using the GC-Rich PCR system (Figure 1D). Founder mice, confirmed positive for the targeted integration, were used to establish aging colonies of R26NHEJ mice.
To test whether NHEJ efficiency changes with age across multiple tissues, five young (5 month old) and five old (25 month old) heterozygous R26NHEJ mice were sacrificed to obtain brain astrocytes and fibroblasts from heart, kidneys, lungs, and skin. Primary cell isolates were characterized using astrocyte-specific GFAP (Glial Fibrillary Acidic Protein) and fibroblast-specific ER-TR7 markers to confirm the identity of cells (Figure S1). The number of cell passages prior to the NHEJ assay was kept at a minimum of 2–3 to avoid clonal selection. To measure NHEJ efficiency, we co-transfected the cells with I-SceI and DsRed plasmids and quantified the number of fluorescent cells by flow cytometry. NHEJ efficiency was calculated as the ratio of GFP+/DsRed+ cells to normalize for any differences in the transfection efficiency. The absence of GFP+ cells in mock-transfections indicated that the reporter construct was not leaky. We found that NHEJ efficiency varied across tissues (Figure 2A). In young mice, the NHEJ efficiency was higher in the kidney, lung, and skin fibroblasts and lower in the astrocytes and heart fibroblasts. Remarkably, we observed a significant, 1.8 to 3.8-fold, decline of NHEJ efficiency with age across all the tissues tested (Figure 2A). The age-related reduction of NHEJ efficiency was highest in the skin (3.8-fold).
We chose the ROSA26 locus for integration of the NHEJ reporter cassette due to its ubiquitous expression and resistance to silencing [38]–[40]. However, whether the ROSA26 promoter is silenced with age had not been reported. To verify that the observed decrease in GFP repair events was not due to reduced ROSA26 transcription, we performed qRT-PCR on total RNA extracted from young and old cells, using primers annealing to Exon 1 of the ROSA26 locus and the first exon of the GFP ORF, to amplify the transcripts. ROSA26 transcription was found to remain unchanged with age in all the cell types tested (Figure 2B), indicating that the observed decrease in the number of GFP+ cells was not due to age-related changes in ROSA26 promoter expression. Although ROSA26 transcription was lower in the lung than in the other tissues, this did not correlate with the NHEJ efficiency because the FACS protocol used for the detection of NHEJ events scored the number of GFP+ cells and did not take into account the GFP intensity.
We next tested whether I-SceI expression changes with age. Western blot analysis of I-SceI levels 24 h after transfection did not show any appreciable differences in the I-SceI expression between young and old tissues or between cell types (Figure 2C), indicating that the observed reduction in NHEJ efficiency is not due to changes in I-SceI expression with age.
To test whether the fidelity of NHEJ changes with age, we cloned and sequenced the NHEJ repair junctions. Genomic DNA was extracted from the cells and digested with PstI to minimize the background of I-SceI uncut constructs. The NHEJ products were then amplified using primers that anneal within the Pem1 intron, cloned and sequenced (Figure S3). The primers we used allowed for the detection of deletions of up to 886 bp on the 5′ side and 750 bp on the 3′ side of the break. Multiple PCR reactions were set up in parallel, and only a single product from each reaction was included in the analysis to ensure that every original repair product was represented only once. A total of 300 sequences, 60 for each cell type were identified and are shown in Table S1. We observed NHEJ repair-associated deletions (ranging from 1 to 990 bp) and insertions (ranging from 1 to 138 bp) in different cell types.
Lung fibroblasts showed a significant increase in the average deletion size with age (p<0.05), and skin fibroblasts demonstrated a trend towards bigger deletions with age (p<0.1) (Figure 3A). Conversely, the average size of deletions decreased significantly (p<0.05) in old heart fibroblasts. In addition, for each cell type, deletion sizes were categorized into small (1–500 bp) and large (>500 bp). The frequency of NHEJ clones with large deletions was significantly increased in old lung fibroblasts and decreased in old heart fibroblasts (p<0.05) (Figure 3B). Astrocytes and kidney fibroblasts did not exhibit significant age-related changes with respect to deletion frequency or size.
A large number of junctions (33–67%) contained insertions, often combined with deletions. When we analyzed the average insertion sizes (Figure 3C), we found a significant increase in the insertion size with age in astrocytes (p<0.05). On the other hand, heart, kidney, and lung fibroblasts showed substantially smaller insertions with age. In addition, the frequency of clones with insertions was significantly greater in young kidney and lung fibroblasts compared to their older counterparts (p<0.05) (Figure 3D). No change with age in the insertion size and frequency was seen in skin fibroblasts. These data suggests that DNA synthesis leading to the generation of insertions is inhibited in old cells. In summary, NHEJ in aged cells exhibits a higher propensity to generate deletions and a reduced frequency of insertions.
We next tested whether aging affects the choice of the NHEJ sub-pathway. The intron within the GFP gene contains multiple microhomology sequences ranging from 1–16 bp. The theoretical probability of obtaining a junction that contains a microhomology if the ends are joined at random within the Pem1 intron is 44% [35]. The percentage of sequenced repair junctions with microhomology was significantly greater than expected by chance (Figure 4A). Since the MMEJ pathway is distinguished by the presence of 5–25 bp microhomologies, we analyzed the frequencies of NHEJ repair clones containing microhomologies within this size range. The frequency of repair events utilizing microhomology increased significantly in the heart and lung fibroblasts and showed a trend towards increase in the astrocytes with age (Figure 4B). The use of microhomology did not change with age in the skin, and was reduced in the kidney (Figure 4B). In summary, these results suggest that the use of MMEJ pathway increases with age in several mouse tissues. It is possible that the increased utilization of MMEJ is a compensatory mechanism to cope with the decline in the c-NHEJ function. Since MMEJ is a more error-prone pathway, this change in the pathway use may lead to a further increase of the mutation load by increasing the size and frequency of deletion events.
Aged tissues accumulate genomic rearrangements [41] and defects in DSB repair have been implicated in the aging process [42]–[44]. Here, we tested whether the process of NHEJ deteriorates during normal aging in the mouse. To address this question, we generated the R26NHEJ mouse model, in which a GFP-based NHEJ reporter cassette was knocked-in downstream of the ROSA26 promoter. To our knowledge, this is the first instance of a mouse model that can quantify NHEJ repair in multiple mouse tissues. The advantage of this model is that the NHEJ reporter counts completed repair events, rather than intermediate steps such as formation of γH2A.X foci or recruitment of DNA repair proteins. Furthermore, only the NHEJ events are scored, offering an advantage over comet assays, which cannot distinguish between DSB repair pathways. We found no age-related changes in the expression of the ROSA26 promoter (Figure 2), making this locus an ideal targeting site for aging studies. This is consistent with the fact that ROSA26 promoter has an open chromatin configuration and virtually no detectable methylation of CpG islands [45]. Another unique feature of the R26NHEJ mouse is that NHEJ events can be analyzed in the same chromosomal location across multiple cell and tissue types. ROSA26 expression was lower in the lung fibroblasts than in the other cell types, which may reflect the chromatin state of this locus in the lung. However, because our FACS parameters were set to include the majority of GFP+ cells, regardless of the fluorescence intensity, lower ROSA26 expression did not result in lower detected NHEJ efficiency in lung fibroblasts relative to the other tissues. This shows that the R26NHEJ system is suitable for comparison of NHEJ across tissues.
NHEJ repair events can be detected in vivo using this system by expressing I-SceI in the mouse to induce DSBs. Several approaches to induce DSBs in vivo were attempted including delivery of I-SceI endonuclease by adenoviral vector, adeno-associated (AAV) vectors, nanoparticles, or hydrodynamic tail-vein injections. The best results were obtained with an AAV9 vector but the frequency of the events was not sufficiently robust for quantitative analysis of repair events with reliable statistics. Therefore, we chose to examine the NHEJ efficiency ex vivo in freshly isolated fibroblasts from the heart, kidney, lung, skin, and brain astrocytes. We found that NHEJ efficiency declined significantly in all the cell types tested. The decline in repair efficiency ranged from 1.8 to 3.8-fold, with the strongest 3.8-fold decline observed in skin fibroblasts. Inefficient NHEJ can impede the cell cycle or lead to cell death. Furthermore, inefficient NHEJ may leave broken DNA ends exposed for a longer time, facilitating the formation of inappropriate junctions and large-scale genomic rearrangements previously observed in aged tissues [6], [8]. Indeed, we observed an increase in the deletion sizes (Figure 3A, B) with age. In addition to promoting tumorigenesis, such rearrangements could disrupt higher order chromatin structure and contribute to age-related dysregulation of gene expression patterns [46].
Fibroblasts and astrocytes play important roles in maintaining the tissue structure and function [47], [48]. Fibroblasts primarily form the extracellular matrix around cells, playing important roles in the structure, function, repair, signaling, and death of the surrounding specialized cells [49]. Astrocytes or glial cells are essential for providing structural, metabolic, and functional support to neurons as well as for memory formation, DNA repair, and scarring during trauma [50]. Impaired function of these cells, due to inefficient NHEJ, could contribute to age-related decline in the tissue function. Although the current study was limited to fibroblasts and astrocytes, the R26NHEJ mouse can be utilized to study NHEJ in all cell types, provided efficient delivery of the I-SceI endonuclease. As mice frequently develop lymphomas with age, it would be interesting to examine NHEJ in their hematopoietic tissues. We are currently generating a lentiviral vector to express I-SceI in hematopoietic cells.
The R26NHEJ mouse has allowed us to compare NHEJ efficiency in the same genomic locus across different tissues and different ages. We found that NHEJ is more efficient in the kidney, lung, and skin fibroblasts than in astrocytes and heart fibroblasts. The strongest decline in NHEJ with age was observed in the fibroblasts from two epithelial tissues, skin and lung. This could be explained by the higher number of cell divisions that occur over time in these tissue compartments, leading to faster cell aging and increased genomic instability. Understanding tissue specific differences in DNA repair pathways can shed light on the tissue specificity of familial cancer-predisposing mutations in genes such as BRCA1, BRCA2, or APC.
The Pem1 intron of the R26NHEJ construct can tolerate sizeable deletions and insertions, enabling the analysis of a wide range of NHEJ events at the repair junctions. Repair-associated aberrations exhibited great variation at different ages and in different tissues. Notably, NHEJ events in old lung and skin fibroblasts were associated with larger deletions, while NHEJ events in old heart fibroblasts were associated with smaller deletions. Large deletions could be a direct result of less efficient NHEJ resulting in broken ends being exposed to nucleases for a longer time. In addition to deletions, 33–67% of the junctions contained insertions. The insertion size increased with age in the astrocytes but decreased in the heart, kidney, and lung fibroblasts. The frequency of clones with insertions also decreased in the kidney and lung fibroblasts. Since insertion formation requires DNA synthesis, this decrease could be explained by a decrease in repair-associated DNA synthesis in old mice. We also found a considerable increase in MMEJ events in old heart and lung fibroblasts. The increased propensity for MMEJ suggests a possible switch from c-NHEJ to MMEJ, rendering the end-joining process more mutagenic, leading to genomic instability. Interestingly, compromised NHEJ has been implicated in tumorigenesis with an observed increase in alt-NHEJ mechanisms. Human urinary bladder carcinomas and myeloid leukemia exhibit high levels of MMEJ [51], [52]. As cancer incidence increases exponentially with age, the shift from c-NHEJ towards MMEJ observed in the heart and lung fibroblasts, may contribute to this process.
We previously examined the changes in NHEJ during replicative senescence of human fibroblasts [35]. Similar to the current study, we observed up to 4-fold decline in repair efficiency during replicative aging. Aged tissues contain a complex mixture of cells, including replicatively senescent cells, cells senescent due to stress, and proliferating cells; therefore, it was important to test whether NHEJ declines in aged tissues. Our findings, based on the NHEJ reporter mice, demonstrate that repair efficiency declines in aged tissues. Interestingly, the effect of replicative senescence and aging on the ratio of c-NHEJ to MMEJ was different between human and mouse cells. Although repair junctions from replicatively senescent human cells contained larger deletions, the use of microhomologies was markedly reduced [35]. On the other hand, MMEJ frequency was found to increase with age in mice. This difference may contribute to the higher genome stability and lower cancer incidence in humans compared to mice.
In conclusion, we show that NHEJ becomes less efficient with age, which may contribute to increased genomic instability and cancer incidence. The ubiquitously expressed, chromosomal NHEJ construct makes the R26NHEJ mouse a powerful tool for the analysis of NHEJ. These mice can be bred with various mice mutated for genes involved in DNA repair and aging. Furthermore, the effect of pharmacological and dietary interventions on NHEJ could be analyzed. These mice can also be used to compare NHEJ between different tissues and cell types, which could shed light on the tissue-specific differences in tumor susceptibility.
All mouse experiments were performed in accordance with the guidelines established by the University of Rochester Committee on Animal Resources (UCAR). All the experimental protocols were approved by UCAR and the approval number is 101423. Mice were euthanized by CO2 inhalation according to the approved protocol.
The NHEJ reporter cassette was inserted into the Multiple Cloning Site of the pBigT vector [40] and this entire construct was cloned into the pROSA26PA plasmid between the arms of the ROSA26 genomic sequence [39]. This pROSA26PA-NHEJ vector was then targeted into C57BL/6 mouse-derived embryonic stem (ES) cells. The transfection of ES cells and subsequent blastocyst injections were performed by inGenious Targeting Laboratory. We chose the C57BL/6 mouse strain because of its well-characterized aging pattern and relative longevity [37]. G418-resistant ES colonies were screened by Southern blot using BamHI digestion. ES cell clones that showed the desired chromosomal integration were injected into blastocysts, which were then transplanted into pseudopregnant female mice. Chimeric males obtained were mated with C57BL/6 females and the resulting 8 offspring were genotyped. Five out of eight founder mice confirmed positive for the targeted integration were then used to establish aging colonies of R26NHEJ mice.
R26NHEJ founder mice were genotyped using a GC-Rich PCR System (Roche) with the Forward 5′ primer 5′- CGGGACTCTGGCGGGAGGGCGGCTTGGTGC - 3′ binding to the ROSA26 promoter sequence involved in homologous recombination and the Reverse 3′ primer 5′- GTTCTAGAGCGGCCTCGACTCTACGATACC - 3′ binding to the internal sequence of the ROSA26NHEJ construct to selectively amplify a 1.3 kb band.
To distinguish between heterozygous and homozygous R26NHEJ mice, another primer pair was used in conjunction with the above PCR genotyping primers. The Forward Chr6-Sens primer 5′- AGTCGCTCTGAGTTGTTATCAGTAAGG - 3′ is homologous to the Chromosome 6 sequence upstream of the ROSA26 locus and the Reverse Chr6-Anti primer 5′- GGTTTCATGAGTCATCAGACTTCTAAGATCAGG - 3′ can bind only to the Chromosome 6 sequence, past the ROSA26 locus. Consequently, wild-type C57BL/6 (+/+) and heterozygous (N/+) mice with 1 copy of the NHEJ construct can amplify the 741 bp band with these primers while homozygous (N/N) mice with 2 copies of the NHEJ construct cannot.
Primary cultures were established from brain astrocytes and fibroblasts from heart, kidney, lungs, and skin of R26NHEJ mice. Briefly, mouse brains were mechanically digested with 0.25% Trypsin/EDTA in PBS for 10 min, washed, and plated on Poly-l-lysine- (Sigma) coated plates. Seven to ten days post-isolation, supernatant and debris were aspirated while confluent astrocytes remained adhered to the plates. Fibroblasts were isolated using protocols described previously [53]. All the cells were grown using DMEM/F12 media (Gibco) supplemented with 10% Fetal Bovine Serum (Gibco), and 1% Penicillin/Streptomycin (Gibco) under conditions of 3% O2, 5% CO2 at 37°C.
Fibroblasts were transfected with 5 µg pCMV-I-SceI for DSB generation and 0.025 µg pCMV-DsRed2 plasmid to normalize for transfection efficiency using program U-023 on Amaxa Nucleofector II using NHDF solution (Lonza). Similarly, astrocytes were transfected using program T-020 and PMGC solution (Lonza). Three days post-transfection, samples were analyzed for GFP+ and DsRed+ cells using Fluorescence Associated cell sorting (Canto II). Multiple transfections were performed for individual cell lines and NHEJ efficiency was calculated as the ratio of GFP+ to DsRed+ cells.
Mouse astrocytes and skin fibroblasts were seeded on slides, grown for 2 days, and fixed using 4% paraformaldehyde. Cells were permeabilized with 0.25% Triton X-100, washed with PBS, and blocked with 1% BSA for 1 h. Astrocytes and fibroblasts were incubated with the primary antibodies, anti-Glial Fibrillary Acidic Protein (GFAP) antibody (Abcam; ab7260) and ER-TR7 Fibroblast-specific antibody (MA1-40076; Thermo Scientific Pierce;) respectively, for 16 h at 4°C. After washing with PBS, the secondary antibodies, goat-anti-rabbit-FITC (Abcam; ab6717) and goat-anti-rat-Cy5 (Abcam; ab6955), respectively, were used, followed by 1 µg/mL DAPI staining for 2 min. Vectashield Mounting Medium (Vector Laboratory; H-1000) was added to the slides along with coverslip and cells were imaged on Leica Confocal Microscope.
Total RNA was extracted from astrocytes and fibroblasts using RNeasy Mini Kit (Qiagen). Following DNAse treatment, RNA was incubated with Oligo(dT) 18 primer and SuperScript III Reverse Transcriptase (Invitrogen) to produce cDNA. cDNA dilutions ranging from 1, 1/2, 1/4, 1/8, and 1/16 were set up to create standard curves. RT-PCR was performed using FastStart Universal SYBR Green Master (Roche) and control Actin primers (Ambion). Primers used for cDNA amplification were: GFP-Ex1 Forward, 5′- CTCGCGGTTGAGGACAAACTCTTCGCGGTCTTTCCAGTGGGG-3′ and GFP-Ex1 Reverse, 5′- GACTTGAAGAAGTCGTGCTGCTTCATGTGGTCGGGGTAGCGGCTGA-3′.
Equal number of cells (1×106) isolated from young and old mice were transfected with 5 µg of pCMV-I-SceI plasmid. Twenty four hours after transfection, SDS-PAGE was performed and I-SceI was detected using anti-HA antibody (Santa Cruz; sc-7392). β-Actin (Santa Cruz; sc-47778) was used as the loading control.
Genomic DNA was extracted from astrocytes and fibroblasts transfected with pCMV-ISceI and pCMV-DsRed2 plasmids, using the phenol-chloroform method. DSB repair sites in the NHEJ construct were amplified by PCR and GC-PCR using various primer combinations; PEM1 Forward, 5′-GCTAAGTGCTTAGTAAAGCAATAGACTGCAT-3′, 5′- GGCTACCTCCAGTTCTAAGGCTGCACTCCA-3′ and PEM1 Reverse, 5′- CTAGGTACTAGGAATTGAACCTAGG-3′, 5′- GGACTAGTAATTGTTTAACATGTGGGAAGTT-3′. Amplified DNA was electrophoresed in 1% Agarose gel and purified using QIAquick Gel Extraction Kit (Qiagen). These NHEJ fragments were cloned into a TA vector using the TOPO TA cloning kit (Invitrogen) and sent for sequencing. Sequenced TA-NHEJ clones were aligned and analyzed using the SerialCloner software.
Significance analysis for the percentage of deletions, insertions, and microhomologies was calculated using two sample t-test between percents using StatPac calculator. In all other cases, significance analysis was calculated using Student's unpaired t-test on GraphPad software.
|
10.1371/journal.pntd.0007058 | Maternal Leishmania infantum infection status has significant impact on leishmaniasis in offspring | Visceral Leishmaniasis is a deadly disease caused by Leishmania infantum, endemic in more than 98 countries across the globe. Although the most common means of transmission is via a sand fly vector, there is growing evidence that vertical transmission may be critical for maintaining L. infantum infection within the reservoir, canine, population. Vertical transmission is also an important cause of infant morbidity and mortality particularly in sub-Saharan Africa. While vertical transmission of visceralizing species of Leishmania has been reported around the globe, risk factors associated with this unique means of Leishmania transmission have not been identified therefore interventions regarding this means of transmission have been virtually non-existent. Furthermore, the basic reproductive number, (R0), or number of new L. infantum infections that one infected mother or dam can cause has not been established for vertical transmission, also hampering the ability to assess the impact of this means of transmission within reservoir of human hosts. Canine Leishmaniosis (CanL) is enzootic within a U.S. hunting dog population. CanL is transmitted within this population via transplacental transmission with no reported vector transmission, despite many repeated attempts to find infected sand flies associated with these dogs and kennels. This population with predominantly, if not solely, vertical transmission of L. infantum was used to evaluate the critical risk factors for vertical transmission of Leishmania and establish the R0 of vertical L. infantum infection. Evaluation of 124 animals born to eighteen dams diagnostically positive for infection with L. infantum showed that there was a 13.84x greater chance of being positive for L. infantum within their lifetime if the mother was also positive within her lifetime (RR: 13.84, 95% CI: 3.54–54.20, p-value: <0.0001). The basic reproductive number for vertically transmitted L. infantum within this cohort was 4.12. These results underscore that there is a high risk of L. infantum infection to transmit from mother to offspring. Targeted public health interventions and control efforts that address vertical transmission of L. infantum are necessary in endemic countries to eliminate visceral leishmaniasis.
| Canine leishmaniosis (CanL) is a deadly disease caused by Leishmania infantum parasite, it is found in animal populations, including people, in more than 98 countries across the globe. CanL was first identified within the US in hunting dogs 1980 and then again in 1999 when a large outbreak in a kennel in New York occurred. As the US is usually not considered a tropical country, there was much debate about how this neglected, vector borne, tropical disease had made its way into these dogs. We found that within the U.S. hunting dog population CanL is transmitted from mom to pup with no reported sand fly transmission in the population, despite multiple attempts to find infected sand flies associated with these dogs. While vertical transmission of this disease has been reported in case reports around the globe, risk factors associated with this unique means of Leishmania transmission are not known. Furthermore, the basic reproductive number, (R0), or number of new infections that one infected animal can cause has not been reported for vertical transmission of L. infantum. It is important to know the R0 as it helps identify how infectious a route of transmission can be and therefore how easy it might be to control this infection. A cohort of 124 dogs from 18 dams was analyzed from 1999 to 2016 for factors related to vertical transmission. Offspring from dams ever diagnostically positive for infection with L. infantum were 13.84x more likely to become positive for L. infantum themselves within their lifetime (RR: 13.84 95% CI: 3.54–54.20 p-value: <0.0001). The basic reproductive number for vertically transmitted L. infantum within this cohort was 4.12. These results underscore that an infected mom is highly likely to infect her offspring if treatment is not started to prevent transmission. There is a need for any public health prevention and control efforts to address vertical as well as vector transmission of canine leishmaniosis in endemic countries.
| Leishmaniosis is a disease caused by the obligate intracellular protozoan parasite Leishmania infantum [1–3]. Visceral Leishmaniasis (VL) can also be caused by Leishmania donovani which causes anthroponotic human visceral leishmaniasis in many countries including areas of Asia and Africa [4, 5]. Zoonotic visceral leishmaniasis (ZVL) occurs in countries where the disease is endemic/enzootic in both human and animal populations. Within these countries the parasite is transmitted primarily via the phlebotomine sand fly [6, 7], although the role of other means of transmission, particularly vertical transmission, is not known. Dogs play an important role in the ecology and control of ZVL as they are the predominant domestic reservoir for the disease, with greater than 10% seropositivity often evident in dogs prior to emergent VL observed in people [8]. Dog ownership is a risk factor of human visceral leishmaniasis in multiple endemic countries with ZVL including Iran, Ethiopia, and Brazil [9–11]. As such, control measures in locations where ZVL is prominent include insecticide treatment or culling of dogs. Dogs remain an important model system for understanding the ecology and epidemiology of VL [12–14].
In recent years vertical, and specifically transplacental, transmission of L. infantum has been shown to be able to maintain infection within population(s) of dogs [15, 16]. Dogs in Brazil have been shown to have infected in utero pups [17–19]. Multiple case reports and case series have identified vertical transmission of VL as an important cause of infant morbidity and mortality [20–22]. Compared to sand fly transmitted infection [23–25], there is very little known about the risk of vertical transmission in dogs or people [16, 26–29]. Therefore, understanding the impact and risk factors associated with parasite transmission in utero is important for education and treatment of infected mothers and for control of Leishmania infection within reservoir hosts.
In the United States leishmaniosis is enzootic in hunting dogs. CanL was first identified in a dog with no travel outside of the United States in 1980, but it was not until a large outbreak in a kennel in New York in 1999 that a larger scale study was performed to understand the broad burden of disease in the U.S. hunting dog population [30, 31]. Further examination found that the primary route of transmission in this population was vertical, from dam to pup [15, 32] and not via sand fly transmission despite many studies looking for infected sand flies associated with these infected dogs [33, 34]. Despite experimental studies that indicate that vector transmission of the Leishmania infantum found in US hunting dogs is possible, there is no evidence that vector transmission occurs naturally from the U.S. hunting dog population [34–36]. A decade of surveillance of this hunting hound population found that the prevalence of CanL from vertical transmission was higher than expected and similar to the rates seen in countries where VL is endemic [37, 38].
The basic reproductive number, R0, or the number of secondary infections one infected individuals can cause within a susceptible population is an important epidemiological value for public health officials interested in control and elimination of this disease in endemic countries [39]. Previous calculations of the R0 for leishmaniosis have been restricted as these studies did not include vertical transmission as a potential route of transmission or lacked data to assess the true rate of transmission in a population [40–42].
This study examines both L. infantum positive and negative dams their offspring over the course of their lifetime to determine risk factors associated with vertical transmission and the corresponding crude basic reproductive number of vertical transmission. We hypothesize that the crude R0 of vertical transmission will be greater than one: Leishmania will maintain infection by infecting at least one pup from a diagnostically positive dam. Understanding the risk factors associated with vertical transmission remains an important public health concern as elimination and control programs focusing on vector control does not show 100% reduction of VL in endemic countries with zoonotic disease [43–45], and vertical transmission appears to be a major risk for maintaining disease within an area or population.
A retrospective cohort study based on data collected regarding Leishmania infantum infection and exposure in U.S. hunting dogs since the 1999 outbreak [33, 34, 46] was completed. A subset of dams that were diagnostically positive and never diagnostically positive were identified. All pups from these two respective groups, ever positive or never positive, were tracked to determine their Leishmania diagnostic status. All historical data was collected from studies performed by Centers for Disease Control and Prevention [33, 34] and the our laboratory at Iowa State University and the University of Iowa [15, 32, 35, 47–49].
All dogs were enrolled in this retrospective study with informed consent from their caretakers and all protocols followed were approved by the University of Iowa Institution Animal Care and Use Committee (IACUC) an AAALAC accredited institution following the requirements for the US National Institutes of Health Office of Laboratory Animal Welfare Assurances which operates under the 2015 reprint of the Public Health service Policy on Humane Care and Use of Laboratory Animals, under protocol #6041721.
An active surveillance cohort of 4 large (>50 dogs each) kennels was established and observed over a 9-year period. Our laboratory visited each of these kennels biannually for at least three years, at which point two of the kennels elected to control visceral leishmaniasis in their kennel via euthanasia. Licensed veterinarians collected 1–5 cc whole blood and serum from all dogs present at these kennels. Demographic information regarding time of pregnancy, sex and age were collected. The active surveillance cohort testing period extended from 2007 to 2017. This surveillance effort started eight years, or at least one hunting-dog life-span, after the major L. infantum outbreak in 1999 with CanL surveillance performed on these same dogs passively by the CDC as reported in [33, 34].
DNA was isolated from canine peripheral whole blood samples collected in heparinized or ethylenediaminetetraacetic acid (EDTA) via the QIAmp DNA Blood Mini Kit (Qiagen, Valencia, CA) per manufacturer protocol. The quality and quanitty of DNA was assessed using a NanoDrop 2000 (Thermo, Scientific, Waltham, MA). Real time quantitative polymerase chain reaction (RT-qPCR) was performed as previously described with all samples run in duplicate with positive samples determined as samples with 1 or more positive wells and negative samples with no amplication in any wells [37, 49–51]. All RT-qPCR included both positive, negative control blood spiked with 106 Leishmania infantum parasites, and negative controls. Between 2007 and 2011 kinetoplastid primer and probe targets were used. The primer and probe sequences were as follows: F 5’-CCGCCCGCCTCAAGAC, R 5’-TGCTGAATATTGGTGGTTTTGG, (Integrated DNA Technologies, Coralville, IA) and TaqMan probe, 5’-6FAM-AGCCGCGAGGACC-MGBNFQ, were used (Applied Biosystems, Foster City, CA). From 2012 to present ribosomal primer and probe targets were utilized. The sequences were as follows: F 5’-AAGTGCTTTCCCATCGCAACT, R 5’ CGCACTAAACCCCTCCAA (Invitrogen, Life Technologies, Grand Island, NY), probe: 5’ 6FAM-CGGTTCGGTGTGTGGCGCC-MGBNFQ (Applied Biosystems, Life Technologies, Grand Island, NY). Assays were performed on ABI 7000 systems until 2016 when they were run on ABI 7900 systems (Applied Biosystems). Analysis was performed using ABI 7000 System SDS Software and ABI 7900 HT Sequence Detection Systems Version 2.4.1. (Applied Biosystems).
Serological status was determined via the Dual Path Platform Canine Visceral Leishmaniasis (DPP CVL) assay (Chembio Diagnostic Systems Inc., Medford, NY) or via immunoflourescent anitbody test (IFAT). The DPP CVL assay detects Leishmania-specific anitbodies via rK28 antigen, a Leishmania recombinant antigen. The assay was utilized as previously described with positives determined as dogs with a test and control line appearing at 4 minutes or less [51]. All positives or questionable samples were confirmed using the Chembio microreader system. The system detects the intensity of the control and test lines. Immunoflourescent antibody test (IFAT) was utlized on canine samples before 2015. This test was performed by the Division of Parasitic Diseases at the Centers for Disease Control and Prevention as previously described [33, 52]. Positive tests were determined as tests where immunofluorescence was reported in 50% of organisms at serum dilutions equal to or above 1/64. These tests were performed without identifying each dog (blindly) and were repeated four times at each dilution to determine positivity.
Univariate analyses were performed to determine unadjusted relative risk values for dam’s age at the time of birth, diagnostic status during the year of birth, and other variables. Pearson chi-squared test and Fisher’s exact test were used to assess categorical variables against disease status. Mann-Whitney test was used to compare dam’s age between disease states as age not normally distributed. An unpaired t-test with the Welch’s correction was utlized to compare litter size between infected and uninfected groups. For assessment purposes the dam’s diagnostic status via qPCR or serology during the same year she gave birth was utilized. Feasability restrictions, the fact the gestational period for a dog is two months, prevented the researchers from obtaining information on the dam’s diagnostic status during pregnancy.
Multivariable logistic regressions were performed to determine adjusted relative risk. Due to the fact that the dam’s diagnostic status can be determined via qPCR and serology, diagnostitc status was assessed different ways through three models. One model included the overall diagnostic status of the dam (ever diagnostic positive vs never diagnostic positive), the dam’s age at the time she gave birth (older than six years of age vs younger than or equal to six years of age), and the sex of the puppy (male vs female). To further assess the dam’s diagnostic status impact a second model was created with qPCR and serology as separate variables. A third model was created separating the dam’s serology and dam’s PCR status in the year she gave birth into two explanatory variables. P-values of less than 0.05 were determined as statistically significant. Each model was fit assuming a binomial distribution with a log link function.
Kaplan-Meier time to event analysis was performed to assess whether dam’s diagnostic status altered time to pup diagnostic positive.
Basic reproductive number was calculated using dams who were ever diagnostically positive for Leishmania, from which their average proportion of puppies per litter that became Leishmania diagnostic positive was determined. Hunting dogs are a medium size dog with average litter size in the study was between 6–7 [53]. Using the average litter size, the proportion of puppies in a litter that would become positive for Leishmania was determined as the basic reproductive number of vertical transmission in US hunting dogs.
For all analyses, as observation of transmission of L. infantum infection was the goal, L. infantum exposure/diagnostic result status for each dog was identified as “ever diagnostically positive” for Leishmania or “never diagnostically positive” for Leishmania. Positivity was determined as qPCR positive and/or serologically positive at any point during the dog’s lifetime.
All statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC) and Graph Pad Prism 6 (GraphPad Software, Inc, La Jolla, CA).
Compared to sand fly transmsision, little is known about the risk factors of vertical transmission of Leishmania infantum. Understanding these risk factors and the corresponding likelihood of transmission as measured by the basic reproductive number, R0, provides valuable information for assessing control and elimination programs for zoonotic leishmaniosis. We hypothesized that a dam’s positive Leishmania diagnostic status during pregnancy would be a risk factor of L. infantum transmission. A retrospective cohort study examined the health records from 130 dogs born to eighteen dams for risk factors associated with vertical transmisison and the corresponding indiviudal level basic reproductive number calculation. Six dogs were removed from analysis due to incomplete data to use in statistical models.There were eight dams identified as Leishmania diagnostic positive at some point in their lifetime and ten dams were diagnostically negative throughout their lives. Most dogs were not multiparous. The average litter size was 6–7 pups (Table 1).
Dogs that ever became diagnostically positive were born to dam’s that were slighly older in age, 5.10 years compared to 4.04 years (p-value = 0.0004) and were more likely to be born to dams who had previously had at least one litter (p-value <0.0001, RR = 3.351 95% CI:2.32–4.83). There was no significant difference between Leishmania diagnostic outcome in male vs. female dogs. Dogs ever diagnostically positive were more likely to be from large(r) litters. This difference have been skewed by on particulalry large litter of fifteen puppies from a dam that was diagnostically positive during her year of pregnancy at six years of age. When this litter is removed the signficance of dam age and litter size is reduced.
Additional analysis shows that dogs born to a dam that was qPCR positive for Leishmania infantum at the time of pregnancy had a relative risk of being diagnostically positive during their lifetime 10.46x greater than the risk than when the dam was PCR negative at the time of pregnancy (Unadjusted RR: 10.46, 95% CI: 3.57–31.82, p-value <0.0001). The dam’s serological status during the year she gave birth was also found to increase the risk of offspring testing diagnostically positive within their lifetime. Pups born to dams seropositve during the year they gave birth were 2.69x more likely to test positive for Leishmania within their life (Unadjusted RR: 2.69 95% CI: 1.32–5.52, p-value 0.0054, Table 2).
A series of three logistic regression models were created to determine the risk factors associated with vertical transmision of L. infantum. The models were labeled as A, B, and C. Whether the puppy became diagnostically positive within their lifetime or not was used as the outcome for these models. Model A assessed a dam’s diagnostic status as ever positive for Leishmania during their lifetime as an explanatory variable along with age at the time of pregnancy, and sex of the dog. When adjusting for all other explanatory variables it is found that dogs born to a dam that was ever positive for Leishmania have a relative risk 13.84x greater than dogs born to a dam that was never diagnostically positive (Adjusted RR: 13.84, 95% CI: 3.54–54.20, p-value 0.0002).
In order to assess the impact of seropositivity/ Leishmania exposure vs detectable parasite infection via qPCR from the blood in transmission two additional models were created; models B and C. Model B utilized a dam’s diagnostic status during the year she gave birth (positive vs negative), age of dam during pregnancy (older than six vs younger), and sex of the puppy as explanatory variables. In this model puppies born to dams diagnostically positive via qPCR or serology during the year of pregnancy were 2.27x more likely to become positive for Leishmania compared to dogs born to a dam that was diagnostically negative at the time of pregnancy. Model C used the dam’s qPCR status, serostatus and age during the year of pregnancy and progenys’ sex as explanatory variables. This model allows for the assessment of how parasite infection via qPCR from the blood vs seropositivity/ Leishmania exposure could affect Leishmania transmission. Pups born to a dam that was qPCR positive for Leishmania during pregnancy were 3.14x more likely to become positive for Leishmania in their lifetime (Adjusted RR: 3.14, 95% CI: 1.37–7.18, p-value: 0.0067, Table 3). Two dogs born to a dam that was never qPCR or serologically positive for Leishmania were found to be positive during their lifetime. One dog was identified as ever qPCR positive and one as ever serologically positive.
A dam’s serological status during the year of pregnancy was not statistically significantly associated with her offspring becoming diagnostically positive. This was an interesting finding as qPCR is a measure of parasite DNA within the peripheral blood. As the transplacental blood supplied each in utero puppy with nutrients, and apparently parasites, this may have increased the risk of the puppy becoming infected with Leishmania parasites.
Based on our findings via univariate and logistic regression, we were interested in evaluating the risk of becoming Leishmania diagnostic positive over years of a pup’s life based on it’s mother’s diagnostic status. To better assess when dogs became diagnsotically positive for Leishmania, time to event Kaplan-Meier curves were created. To visualize the overall relationship between age at which offspring became Leishmania diagnostically positive this was compared between the groups of dam Leishmania positve vs negative ever. Dogs born to positive dams (red) were statistically significantly more likely to become positive at a younger age than dogs born to negative dams (blue) (chi-square: 40.33, p-value <0.0001, Fig 1).
Based on the previous finding that dam qPCR status during the year she was preganant was also highly correlated with the pup becoming Leishmania diagnostic positive, dam’s qPCR status (negative during year of birth vs. positive) was utilized. Offspring born to dams who were qPCR positive (red) during the year they gave birth were significantly more likely to become positive for Leishmania via qPCR at younger ages than offspring from dams that were qPCR negative (blue) (Fig 2, chi-squared: 49.54 p-value <0.0001).
This was similar to the relationship between dams who were seropositve during the year they gave birth and the age at which their puppies became seropositive for Leishmania (Fig 3, chi-squared 18.43, p-value <0.0001).
Within this study cohort we found two instances and three litters in which three generations of infected dogs were identified. In these specific families, on average the second generation had evidence of infection in 79.2% of dogs (seropositive or PCR positive at some point of their lives). To date, dogs in the third generation were 60.4% sero- or PCR positive for L. infantum. It should be noted that one of these litters are dogs currently 3 years old. These younger dogs may become qPCR or seropositive as they age and experience immunosuppressive conditions.
A small subset of 20 dogs within the study were more closely followed through their entire lives and cause of death was established. Of the 20 dogs from infected dams for which a cause of death was identified 95%, or 19 of these dogs, died from clinical visceral leishmaniasis. The one dog identified as being diagnostically positive for Leishmania infantum but not dying from clinical visceral leishmaniasis died from a secondary infection with Ehrlichia spp. as identified at necrospy. Neither of the two dogs born to uninfected mothers found to be infected with L. infantum have died from VL, but this is a very small sample size.
The basic reproductive number for vertical transmission of Leishmania remains of interest in order to determine effectiveness of control efforts that are in many cases focused on vector transmission. R0 was calculated based on information regarding each litter from this population. On average, 64% of dogs born to a dam who were ever diagnostically positive for Leishmania will become positive in their lifetime. Using the average litter size of our population, between 6 and 7, we calculate an average R0 of 4.16.
A retrospective cohort study was performed to assess risk factors associated with vertical transmission of Leishmania infantum and a crude basic reproductive number was calculated for the population. The mother’s L. infantum diagnostic status during the year she was pregnant was a statistically significant risk factor for her offspring to be L. infantum positive during their lifetime, with a signficant 13 times greater risk of infection than dogs without maternal expsoure to Leishmania. Despite these dramatic findings in this retrospective cohort study, there is an overall paucity of reported cases of congential VL. This is likely for several reasons; first the diagnostic difficulties of confirming that a case is due to congential transmission vs. expsoure to sand fly transmitted disease in endemic areas. To date there is no way to distinguish L. infantum infection by route of transmission, so in endemic areas the presumption is that cases are vector borne, although this may not be true. The second reason is availability of treatment of mothers for ZVL during pregnancy reducing the maternal parasite load and therefore decreasing transmission to the child/offspring [27, 54]. This study is the first study to calculate the basic reproductive number and determine risk factors associated with vertical transmission of Leishmania infantum in a population where vertical transmission is the main route of transmission and there is no known vectorial transmission [55, 56].
Vertical transmission occurs not only in leishmaniosis but other infections as well, such as human immunodeficieny virus (HIV) and malaria [57, 58]. In HIV infection, anti-retroviral treatment during pregnancy and caesarean births have been associated with decreased risk of transmission likely due to a reduced exposure to the mother’s blood and virus [59]. In malaria, mothers with malaria during pregnancy are at risk of vertical transmission [60]. This is similar to CanL where dogs born to mothers that were qPCR positive during pregnancy had a much higher risk of becoming positive for Leishmania. This is likely due to the fact that a positive qPCR test identifies that there was parasite DNA in the blood which is shared between mother and pup across the placenta. The mother’s combined diagnostic status of seropositive or qPCR positive was a significant risk factor in predicting whether a puppy would become positive during their life. This was also reasonable as dogs become immunocompromised there can be increased disease progression and parasite replication with higher serological diagnostic values in dogs with more severe clinical disease [47, 50].
Within this study there were two sets of three generations of dogs that were followed and data indicating that transmission occurred across these generations. These results provide additional evidence that vertical transmission is capable of maintaining visceral leishmaniasis in a population over multiple generations.
Within this study two Leishmania-positive dogs were born to dams that were never qPCR or serologically positive for Leishmania. In the hunting dog community, dogs are commonly drafted or traded between groups and across international borders from endemic to non-endemic areas. Such movement of dogs greatly increases the difficulty of consistent testing across different locations and disease risk levels. This testing limitation may have led to a false negative status for the mother[61]. The two puppies that were identified as serologically/qPCR positive without maternal exposure could also have been exposed to Leishmania via fighting or wound cleaning of infected pen mates as blood to blood contact is possible due to the fact the dogs are housed in communal areas.
A small subset of 20 dogs (15% of the study population) were followed until death and a cause of death was identified. 100% of the dogs with an established cause of death were diagnostically positive for Leishmania infantum at some time throughout their life. Of those dogs with an established cause of death in this cohort, 95% died from clinical visceral leishmaniasis. These results highlight that without treatment many of these animals will progress with clinical disease. Therefore, it remains an important public health goal to identify ways to prevent L. infantum transmission from mother to child in both animals and people.
The basic reproductive number was calculated via an indivdual level model system, thus the number refers to the number of dogs in each litter that one mother could infect. This calculation provides a direct assessment of the R0 within this cohort. An R0 of approximately 4 (rounded to the nearest whole number to refer to number of puppies in the litter) shows that this disease is capable of maintaining at high levels within a population without vector transmission. The R0 of other diseases, such as influenza, which remain important public health concerns across the globe are as low as 2 [62]. Astonishingly, in comparison the R0 identified for an average canine litter coming from an infected dam was greater than 4, similar to the estimated basic reproductive number of smallpox [63]. As these studies all occur in an area where there is not holoendemic pressure of sand fly transmission, establishing the R0 and effect of vertical transmission in dogs from endemic areas would be valuable. These studies would all be limited by the inability to distinguish sand fly transmitted and vertical transmission once pups are born and it is hard to know the outcome of maternal infection on in utero pups.
Current control programs for leishmaniosis in countries where the disease remains endemic in both humans and animals include vector control, vaccination, and dog culling, which has been shown to be ineffective. Based on the data evaluated here, there is a significant need to also address vertical transmission through canine sterilization programs [64–66]. Recent studies have identified vaccination of infected/exposed asymptomatic dogs as safe, so vaccination to boost a better immunity prior to pregnancy may be of value to reduce transmission to the next generation [51]. Larger scale xenodiagnosis studies need to be performed to determine what skin burden of parasites is required to transmit CanL and the effectiveness of vaccination [67], allopurinol or additional (immuno)therapies to reduce parasite load immediately before or during pregnancy. Further analysis using Bayesian compartmental model techniques combining both vector and vertical transmission should be used to better understand the basic reproductive number for the full ecology of Leishmania infection in endemic areas and subsequently model how this number can be altered by public health control and prevention measures to assess elimination potential.
The findings of this study underscore the need for risk management through spaying and neutering animals by dog owners to reduce vertical transmision of L. infantum from their dogs. This action would decrease propagation of CanL within the canine reservoir for reduced transmission to people.
|
10.1371/journal.pgen.1005509 | Synergistic and Dose-Controlled Regulation of Cellulase Gene Expression in Penicillium oxalicum | Filamentous fungus Penicillium oxalicum produces diverse lignocellulolytic enzymes, which are regulated by the combinations of many transcription factors. Here, a single-gene disruptant library for 470 transcription factors was constructed and systematically screened for cellulase production. Twenty transcription factors (including ClrB, CreA, XlnR, Ace1, AmyR, and 15 unknown proteins) were identified to play putative roles in the activation or repression of cellulase synthesis. Most of these regulators have not been characterized in any fungi before. We identified the ClrB, CreA, XlnR, and AmyR transcription factors as critical dose-dependent regulators of cellulase expression, the core regulons of which were identified by analyzing several transcriptomes and/or secretomes. Synergistic and additive modes of combinatorial control of each cellulase gene by these regulatory factors were achieved, and cellulase expression was fine-tuned in a proper and controlled manner. With one of these targets, the expression of the major intracellular β-glucosidase Bgl2 was found to be dependent on ClrB. The Bgl2-deficient background resulted in a substantial gene activation by ClrB and proved to be closely correlated with the relief of repression mediated by CreA and AmyR during cellulase induction. Our results also signify that probing the synergistic and dose-controlled regulation mechanisms of cellulolytic regulators and using it for reconstruction of expression regulation network (RERN) may be a promising strategy for cellulolytic fungi to develop enzyme hyper-producers. Based on our data, ClrB was identified as focal point for the synergistic activation regulation of cellulase expression by integrating cellulolytic regulators and their target genes, which refined our understanding of transcriptional-regulatory network as a “seesaw model” in which the coordinated regulation of cellulolytic genes is established by counteracting activators and repressors.
| Cellulolytic fungi have evolved into sophisticated lignocellulolytic systems to adapt to their natural habitat. This trait is important for filamentous fungi, which are the main source of cellulases utilized to degrade lignocellulose to fermentable sugars. Penicillium oxalicum, which produces lignocellulolytic enzymes with more diverse components than Trichoderma reesei, has the capacity to secrete large amounts of cellulases. Meanwhile, cellulase expression is regulated by a complex network involved in many transcription factors in this organism. To better understand how cellulase genes are systematically regulated in P. oxalicum, we employed molecular genetics to uncover the cellulolytic transcription factors on a genome-wide scale. We discovered the synergistic and tunable regulation of cellulase expression by integrating cellulolytic regulators and their target genes, which refined our understanding of transcriptional-regulatory network as a “seesaw model” in which the coordinated regulation of cellulolytic genes is established by counteracting activators and repressors.
| Cellulolytic fungi have an inherent characteristic of cellulose deconstruction and can be used for bioconversion of insoluble plant cell wall polysaccharides into fermentable sugars [1–3]. The highly efficient production of their extracellular hydrolytic enzymes and other synergistic proteins [2,4], such as swollenin [5], plays a key role in reducing the cost of the biorefinery process [4]. However, incomplete knowledge of transcriptional regulatory networks for cellulolytic fungi has hampered the systematic improvement of cellulase production. These cellulolytic system genes are coordinately but differentially regulated in various cellulase producers [2,6]. Further characterization and manipulation of the cellulase regulatory network’s components will allow the rational engineering of cellulolytic fungi for improved enzyme production.
Transcriptional regulation of cellulolytic gene expression is central in controlling the carbohydrate hydrolysis process [6], and several positive or negative transcriptional factors of these degradative pathways were identified, such as the regulators encoded by the creA/cre1/cre-1 [7–9], xyr1/xlnr/xlr-1 [10–12], aceI [13], aceII [14], ace3 [15], clrB/clr-2/manR [16,17], and bglR [18] genes. The overexpression of these activators or deletion of some repressors is efficient in enhancing the cellulase and hemicellulase expressions [19,20]. However, the degree of cellulase induction differentially responds to these diverse regulator abundances. The transcription factor CreA, an ortholog of Migl from Saccharomyces cerevisiae [21], is a pivotal regulator mediating carbon catabolite repression (CCR) in filamentous fungi [7–9], and its deletion results in the obvious increase of cellulase expression and secretion. A transcriptional regulatory cascade that controls the xylanolytic genes between CreA and XlnR is also built in Aspergillus niger in response to preferred carbon sources [22]. In addition, the cre1 deletion mutant shows a conidiation formation defect [23]. Two novel zinc binuclear cluster transcription factors (CLR-1 and CLR-2) required for growth and enzymatic activity on cellulose were identified in Neurospora crassa [17]. The constitutive expression of clr-2 by the control of the promoter from ccg-1 is sufficient to drive cellulase gene expression when cultures are subjected to starvation [20]. In addition, the β-glucosidase regulator BglR and cellulase expression activator AceII were identified in Trichoderma reesei [18], but their orthologous encoding genes were absent in the Penicillium oxalicum genome [24]. Currently, the abilities to tune the expression abundance of just one transcription factor, as noted above, have profound effects on cellulase expression in these cellulolytic fungi [18–20,25]. However, whether such a simple mechanism could operate in the context of cellulolytic regulator combinations, including these characterized and novel transcription factors, remains unclear.
The P. oxalicum wild-type strain 114–2 was isolated from the soil in China more than 30 years ago [26]. A partially derepressed mutant JU-A10, which shows cellulolytic activity that is more than three times higher than that of its parent strain 114–2, was obtained after many rounds of mutagenesis and screening [26]. The mutant JU-A10 was further mutated to a cellulase hyper-producer JU-A10-T [26] and has been utilized in industrial processes for years. The clear genetic background provided by genome sequencing facilitated the rational improvement of these strains to enhance the expression of cellulolytic enzymes [24]. Currently, several structural genes associated with cellulase expression have been studied. The deletion of gene bgl2 (encoding the major intracellular β-glucosidase) [27] or PDE_01641 (the ortholog of N. crassa NCU05137) [28] results in the increase of cellulase production in P. oxalicum. In addition, three cellodextrin transporters (CdtC, CdtD, and CdtG) were identified, and their overexpression obviously increases the extracellular cellobiohydrolase activities [29]. However, evidence from diverse cellulolytic fungi showed that engineering cellulolytic transcription factors might have more efficacy in upregulating cellulase expression than merely manipulating the expression of structural genes for the major cellulases [1,19,20,30]. Subsequent studies to identify several specific regulators and their roles in regulating cellulase gene expression were conceptually appealing in cellulolytic fungi.
In this study, twenty transcription factors putatively involved in cellulase expression pathways were identified from a single-gene disruptant library. The single overexpression or deletion of these genes triggered cellulase expression to varying degrees, and synergistic and tunable cellulase expressions were observed in the combinations of the identified individual transcription factors. Furthermore, the responsiveness of the induction of cellulase expression by activator and relief from potential carbon catabolite repression to the internal signal cascades by the lack of the major intracellular β-glucosidase Bgl2 was also well-established. The suggested mechanisms of synergistic effects on cellulase expression might be general properties in cellulolytic fungi and broadly enable engineering strategies for the protein hyper-producers.
To decipher the transcriptional-regulatory network that governs cellulase expression in P. oxalicum, we first sought to identify the transcription factors (TF) that play roles in cellulolytic gene expression systematically. A total of 522 genes encoding sequence-specific regulators were predicted according to the protein sequence domain [24]. For the amplification of the flanking sequences of these TF-disrupting cassettes, primers were designed to meet the following criteria: GC-content 45%–60%, Tm: 50°C –60°C, and a length of 20 base pairs. The lengths of the 5’ and 3’ flanking regions ranged from 1.0 kb to 1.5 kb for each gene. The chimeric primers (S1 Table) for the amplification of upstream and downstream flanking fragments carried 25 bases of homologous sequence overlapping with the ends of ptra marker sequence [31]. A final fragment that contains target gene flanking sequences surrounding ptra was created by double-joint PCR [32] and transformed into the P. oxalicum Δpku70 mutant via protoplast transformation [33]. Pku70 and its homologs are involved in the non-homologous end joining (NHEJ) repair of double-strand breaks in diverse eukaryotes [32]. Considering the high homologous recombination frequency in the pku70 mutant (NHEJ-deficient background) [33], we selected three colonies per gene from these resulting transformants. The conidia from these primary transformants were purified by repeating the mono-spore isolation twice on the pyrithiamine resistance plates to obtain homokaryotic knockout mutants. The transcription factor gene replacement with ptra was verified by PCR-based screening. We found that the use of unpurified final amplicon of deletion cassettes resulted in almost 90% success in deleting the targeting genes in the Δpku70 mutant. Finally, a transcription factor mutant set, which bears a single deletion for 470 transcription factor genes in P. oxalicum, was successfully constructed.
The transcription factor deletion strains were screened and initially characterized for cellulose deconstruction on cellulose plates. According to the halos produced by the transformants on cellulose plates, 20 transcription factors that displayed putative roles in cellulase production were identified (Table 1). Twelve deletion strains exhibited increased cellulase activities and eight deletion strains exhibited decreased cellulase activities. These transcription factors represented negative and positive regulators of cellulose deconstruction, respectively. None of these transcription factors has been well characterized at the molecular level in P. oxalicum.
Among these transcription factor genes, PDE_05999, PDE_03168, PDE_07674 and PDE_03964 were previously known and encoded as ClrB, CreA, XlnR and AmyR regulators, respectively. The strongest effects on cellulose deconstruction observed in ΔclrB, ΔcreA, ΔxlnR and ΔamyR mutants indicated that these three genes encode the major regulators of some lignocellulolytic enzymes (Fig 1A). To clarify the mechanisms of lignocellulose deconstruction in P. oxalicum, we initially focused on the characterization of these central lignocellulolytic regulators ClrB, CreA, XlnR, and AmyR, and then identified their target genes involved in plant cell wall deconstruction. Up to date, mating assays in P. oxalicum have not been performed to remove the pku70 deletion through the crossing approach as T. reesei [34] and N. crassa [35] strains. We therefore constructed the corresponding mutants in the wild-type strain through the conventional transformation approach. Genomic DNA from putative transformants was analyzed by q-PCR (quantitative-PCR) and/or Southern blot (S1 Fig), and these transformants in which a single copy integration at the only a transcription factor gene locus were selected and further characterized.
The function of clrB (PDE_05999) was identified independently in our lab. The regulator protein sequence has 39% (BioEdit, E Value = 1e-131) of identity to the homolog of N. crassa and 56% of identify (E Value = 0) to that of A. nidulans [17] (Table 1). The P. oxalicum clrB gene encodes a protein of 780 amino acid residues. Two introns of 76 and 64 nucleotides, which follow the characteristics of the clrB homolog in N. crassa, were identified [17]. The deduced transcription factor ClrB contains normal characteristics of Zn(II)2Cys6 binuclear cluster DNA binding motif near the N-terminus (residues 40–71) and the middle homology domain (residues 351–453) that is related to fungal specific transcription factors, including XlnR/XYR1 [10,11] and yeast regulatory protein GAL4 [36]. In this study, several putative cellulolytic transcription factors, such as PDE_03268, PDE_03964, and PDE_09881, also contain normal characteristics of these zinc binuclear cluster proteins (Table 1).
To investigate the influence of ClrB on cellulase expression, we constructed a ΔclrB strain from P. oxalicum wild-type strain 114–2 (CGMCC 5302). The ΔclrB strain displayed significantly reduced growth on cellulose plate, but identical phenotype on glucose, xylan, or potato dextrose agar (PDA) plates relative to wild-type strain (Fig 1A). The ΔclrB mutant exhibited dramatically reduced cellulase activities when compared with the wild-type strain (Fig 1B–1D), similar to the recent findings with clr-2/clrB in N. crassa/A. nidulans [17]. Northern blot analysis was used to study the cellulolytic gene cbh1 (PDE_07945), eg2 (PDE_09226), and xyn1 (PDE_08094) expressions in the wild-type and ΔclrB strains grown on cellulose. Fig 2A shows that the mRNA levels of cbh1 and eg2 in the ΔclrB mutant significantly decreased and could hardly be detected. As a result, a slight decrease in the xyn1 transcription level was observed in the ΔclrB strain compared with that in the wild-type control, while a low transient increase expression of xyn1 was observed at the fourth hour following the shift (Fig 2A). The introduction of a wild-type copy of clrB at the clrB locus (strain RclrB) completely restored the growth defects of the ΔclrB mutant in cellulose, as well as the cellulolytic enzyme activities of culture supernatants (Fig 1B–1D). These results demonstrated that ClrB might be in the central part of the transcriptional-regulatory network of cellulase expression by controlling the transcription of cellulolytic genes.
The P. oxalicum genome contains approximately 10,000 genes and is predicted to encode 18 cellulases, 51 hemicellulases, and other cellulolytic enzymes involved in plant biomass degradation [24]. Therefore, to build a comprehensive picture by which P. oxalicum responds to cellulose, we adopted RNA-Seq to measure genome-wide mRNA abundances in the P. oxalicum wild-type strain and ΔclrB mutant when exposed to Vogel’s minimal medium containing 2% cellulose for 4 hours. The three biological replicates of each strains showed a high Pearson correlation (S2 Fig). A total of 224 genes were differentially expressed between the ΔclrB and the wild-type strains on cellulose (S2 and S3 Tables). Of these genes, 103 genes showed lower transcription levels in the ΔclrB mutant than in the wild-type strain (S2 Table). These genes of decreased expression in clrB regulon were subjected to gene ontology enrichment analysis. Percentages of the genes distributed within each functional category are shown in Fig 3. Among these downregulated genes, 24 genes encoding transporters were enriched, including PDE_00607 (p = 7.99e-173), encoding cellodextrin transporter CdtC [29], and PDE_06576 (p = 2.96e-58), encoding putative maltose permease, which suggests that ClrB might also be involved in the cellodextrin and maltose metabolisms. In total, 32 genes encoding carbohydrate-active enzymes (CAZymes) were included, including 9 main cellulase genes and two of 11 β-glucosidases PDE_00579 (p = 1.98e-109) and PDE_04251 (p = 2.39e-42) (S2 Table), which demonstrates that the genes involved in plant cell wall deconstruction were significantly enriched. Only 6 of the 51 hemicellulase genes showed obvious reduction in transcription levels in the absence of ClrB, including PDE_02101 (p = 0.0033), PDE_06649 (p = 0.00046), PDE_01302 (p = 5.63e-12), PDE_09710 (p = 3.09e-09), PDE_05998 (p = 1.36e-37), and PDE_06023 (p = 1.87e-41) (S2 Table). These data demonstrated that ClrB might play a significant role in activating cellulase gene expression but has differential regulatory effects on cellulolytic and xylanolytic genes in the early inducing phase on cellulose (4 hours post-transfer).
In total, 121 genes showed higher transcription levels in the ΔclrB mutant than in the wild-type strain (S3 Table). Among these upregulated genes, only 7 genes encoding CAZy proteins, including two predicted hemicellulase genes PDE_07585 (p = 0.02) and PDE_08238 (p = 0.013), showed altered expressions. No cellulase, β-glucosidase, and xylanase genes were included. Genes in this set were enriched in oxidoreductase activity (p = 8.1e-6).
To assess whether the overexpression of clrB enhanced the cellulase expression, two clrB overexpression recombinants were constructed under the control of its native promoter (strain OEclrB) and the A. nidulans gpdA promoter (strain gpdA(p)::clrB) in P. oxalicum wild-type strain [37], respectively. Both the OEclrB and gpdA(p)::clrB strains showed varied halos on 1% cellulose plates (Fig 1A) and showed almost 2.5- and 4.1-fold increases in filter paper enzyme activity (FPA), 2.5- and 4.0-fold increases in cellobiohydrolase (pNPCase) activity, and 8.7- and 16.5-fold increases in endoglucanase (CMCase) activity when grown on cellulose for 48 hours, respectively (Fig 1B–1E). Northern blot analyses also showed that the mRNA levels of cbh1 and eg2 in the gpdA(p)::clrB mutant were much higher than those in the wild-type strain on cellulose (Fig 2A).
To further test whether cellulase production was tightly responsive to clrB transcriptional abundance, we reconstructed the PDE_02864(p)::clrB expression cassette in which the clrB open reading frames and 3’ untranslated region were under the control of the novel promoter from the PDE_02864 encoding 40S ribosomal protein S8. The gpdA(p)::clrB-PDE_02864(p)::clrB strain was constructed and showed even higher cellulase expression than those in the single gpdA(p)::clrB and PDE_02864(p)::clrB mutants on cellulose (S3 Fig). These results demonstrated that dose effect of clrB transcriptional abundance is important for the high expression for cellulases, and tunable cellulase expression may be controlled by the ClrB concentration under cellulose conditions.
In addition to ClrB, P. oxalicum XlnR is another cellulolytic activator in the Zn2Cys6 binuclear cluster motif superfamily that was identified previously along with the homologous N. crassa XLR-1 [12] and T. reesei XYR1 [10]. To demonstrate whether XlnR showed a differential role in cellulolytic gene expression regulation, the ΔxlnR and gpdA(p)::xlnR mutants with constitutive expression for xlnR under the control of the A. nidulans gpdA promoter in the wild-type strain were constructed. The ΔxlnR mutant showed a slightly decreased growth on both cellulose and xylan media, but not on glucose or PDA plates (Fig 1A). Northern blot results also showed that weak expressions for cbh1 and eg2 transcripts and invisible xyn1 were observed in the ΔxlnR strain compared with that in the wild-type strain in the cellulose-containing medium (Fig 2A). These data indicated that P. oxalicum XlnR is a general transcription factor that regulates cellulase and xylanase expressions but not like T. reesei XYR1, which is the essential regulator that governed both cellulolytic and xylanolytic gene expressions [10].
To further analyze whether synergistic or additive effects for these two major cellulolytic activators ClrB and XlnR existed, the clrB was also overexpressed in the gpdA(p)::xlnR mutant, and the gpdA(p)::clrB-gpdA(p)::xlnR mutant that contains simultaneously overexpressed ClrB and XlnR was obtained. The gpdA(p)::clrB-gpdA(p)::xlnR strain showed 1.3-, 1.7- and 2.1-fold increased expressions in FPA, xylanase, and pNPCase activities compared with that in the gpdA(p)::clrB strain after shift to cellulose for 96 hours (Fig 4A–4C), but decreased production in the pNPGase activity (Fig 4D). Conversely, lack of both ClrB and XlnR (ΔclrB-ΔxlnR) led to a greater abrogation of cellulase and xylanase expression than each absence mutation under cellulose growth conditions (Figs 1B and 2A). These data revealed that ClrB and XlnR had additive effects on positively regulating the cellulase and hemicellulase gene expressions, and the high-abundance transcripts of ClrB and XlnR could facilitate the induction of cellulase expression under cellulose growth conditions.
PDE_03168 (CreA), as a homolog of the major carbon catabolite repressors in phylogenetically diverse fungi [7–9], played a negative role in the degradation of plant cell wall polymers. In this study, the ΔcreA mutant of P. oxalicum consumed cellulose faster than the wild-type strain (Fig 1A), similar to the findings in T. reesei cre1 [23] or N. crassa cre-1 deletion strains [38]. The ΔcreA strain grown on cellulose exhibited significantly increased cellulase activities compared with its parent strain, and showed almost 7.2-, 2.2-, 8.0-, and 4.4-fold increases in FPA, pNPCase, CMCase, and xylanase activities after shifting to cellulose for 96 hours, respectively (Fig 1B–1E). The ΔcreA mutant exhibited higher steady state amounts of cbh1 and eg2 mRNA than that in the wild-type strain by Northern blot analysis (Fig 2A) and q-PCR experiments (Fig 2B) under cellulose growth conditions. However, the gpdA(p)::creA mutant that contains an overexpression of creA showed significantly lower cellulase gene expression than that in the wild-type strain when grown on cellulose (Fig 2B), which indicates that cellulolytic gene expression under CCR mediated by the CreA in P. oxalicum was responsive to the creA transcript abundances.
Although the ΔcreA mutant produced higher cellulase expression than its parental strain P. oxalicum 114–2, we could not entirely exclude the possibilities that the increase of cellulolytic enzyme in ΔcreA mutant might be the partial cause of the enhancement of cellulolytic activators for ClrB and XlnR. To test these hypotheses, q-PCR was performed and the expression levels for clrB and xlnR in the ΔcreA mutant showed near to that in the wild-type strain on cellulose but obvious increase under glucose-repressing conditions (Fig 5A). These data indicated that CreA repressing the expression of clrB and xlnR was associated with the carbon source used in the medium, and CreA and ClrB, as well as CreA and XlnR might form a transcriptional cascade that regulates the cellulolytic gene expression in P. oxalicum.
Previously, we reported that a cellulase hyper-producer P. oxalicum mutant JU-A10, which bears a shift mutation at creA locus, has an extremely severe effect on morphology, including short and thicker hyphae [1]. The creA deletion in P. oxalicum wild-type strain 114–2 showed smaller colonies and reduced hyphae growth (Fig 1A). Under the microscope, the ΔcreA mutant on glucose plates exhibited considerably robust hyphae (Fig 5B). Thus, some hyphae morphology mutation in P. oxalicum mutant JU-A10 [1] might be specific results caused by the shift mutation at the creA location in the JU-A10 mutant by chemical mutagenesis. Similarly, the T. reesei Δcre1 mutant also displayed shorter and more robust hyphae than its parental strain [23]. We hypothesized that the morphology mutation caused by creA homologs may be general in filamentous fungi. However, whether hyphae morphology mutation was related to the increase in protein production and cellulase expression in these mutants containing deletion of creA needs to be further investigated in the future.
Cellulase genes were responsive to both major opposing regulators ClrB and CreA. First, whether the low cellulase expression exhibited by the ΔclrB mutant could be recovered when knocking out the CreA encoding gene remains unknown. To bypass this problem, we deleted the creA in the ΔclrB strain and obtained the ΔcreA-ΔclrB mutant. Northern blot, presented in Fig 2A, shows that ΔcreA-ΔclrB mutant exhibited a slightly higher amount of cbh1 mRNA than that in the ΔclrB strain. The ΔcreA-ΔclrB mutant also showed 3.1-, 0.5-, 3.1-, and 1.8-fold increases in FPA, pNPCase, CMCase, and xylanase activities relative to wild-type strain after shifting to cellulose for 96 hours, respectively (Fig 1B). In contrast to the ΔclrB mutant, the ΔcreA-ΔclrB strain produced visible halo in the cellulose medium plate when cultured for 9 days (S4 Fig). These data indicated that the lack of CreA partially rescued the cellulase expression defect in the ΔclrB mutant. Conversely, the gpdA(p)::creA-gpdA(p)::clrB strain that contains simultaneous overexpressions of CreA and ClrB also showed increases in the transcription expression levels for most of the cellulase genes compared with that in the gpdA(p)::creA mutant (Fig 2B).
The strong cellulase production observed in the ΔcreA and gpdA(p)::clrB mutants indicated that these two genes encoded the major transcription factors that oppositely regulate cellulolytic gene expression. Although, the triple-mutant RE-10 (Δbgl2-ΔcreA-gpdA(p)::clrB) was recently obtained [39], it is still not known why high cellulase expression in which occurs. Therefore, we deleted creA in the gpdA(p)::clrB strain to determine whether their functions in cellulase production were synergistic. To our knowledge, no reports have been described to perform this to determine their synergistic effects on cellulase induction expression in other cellulolytic fungi. Interestingly, the strain with the combination of creA deletion and overexpression of clrB displayed a much larger halo around its colony than that of the ΔcreA and gpdA(p)::clrB mutants on cellulose plate (Fig 1A). The differences in the FPA, pNPCase, CMCase, and xylanase activities were even more pronounced (11.6-, 11.6-, 58.6-, and 15.9-fold higher, respectively) between the ΔcreA-gpdA(p)::clrB and wild-type strains (Fig 1B). The ΔcreA-gpdA(p)::clrB mutant exhibited higher steady-state amounts of cbh1, eg2, and xyn1 mRNA than that of the ΔcreA and gpdA(p)::clrB strains under cellulose growth conditions by Northern blot analysis (Fig 2A). Q-PCR results also indicated that all three cellobiohydrolase and 10 endoglucanase genes showed strong synergistic increases in transcription levels with the exception of five endoglucanase genes for PDE_009267, PDE_00698, PDE_03711, PDE_09014, and PDE_06768 when grown on cellulose (Fig 2B). In addition, three of the 11 β-glucosidase genes (PDE_00579, PDE_04251, and PDE_04859) showed similar induction patterns as the above thirteen cellulase genes in the ΔcreA-gpdA(p)::clrB mutant on cellulose (S5 Fig). These data showed that the simultaneous overexpression of ClrB and lack of CreA have synergistic effects on cellulolytic and xylanolytic gene expressions.
To build a comprehensive picture by which P. oxalicum responds to cellulose, the genome-wide mRNA abundance in P. oxalicum wild-type strain was first measured on Vogel’s medium with no carbon or 2% glucose for 4 hours as an alternative reference. A total of 581 genes were differentially expressed between Avicel and no carbon cultures at the fourth hour (|log2(fold change)| > 1 and probability ≥ 0.8), 155 and 117 of which showed greater and lower expression levels on cellulose than that under either glucose growth or no carbon conditions, respectively (S4 and S5 Tables). We refer to this gene set, including 272 genes, as a “cellulose regulon” for P. oxalicum. The cellulose regulon encompassed 10 of 18 predicted cellulase genes, 30 of 51 predicted hemicellulase genes, and 3 of 11 predicted β-glucosidase genes PDE_02736 (p = 1.65e-07), PDE_00579 (p = 9.24e-174), and PDE_04251 (p = 1.02e-80) (S4 and S5 Tables). To further elucidate the regulation mechanisms for the opposing regulators CreA and ClrB in the gene expression of P. oxalicum under cellulose growth conditions, three biological replicates of ΔcreA, ΔcreA-ΔclrB, and ΔcreA-gpdA(p)::clrB mutants were also subjected to transcriptional profiling at the same condition. The three biological replicates of each mutant showed a high Pearson correlation (S6 Fig) and demonstrated the reliability of RNA-Seq. The gene ontology enrichment analyses for these 117 decreased genes of cellulose regulon showed no statistically significant results with the threshold (FDR < 0.05).
The hierarchical clustering of expression patterns for the 155 increased genes of cellulose regulon in the wild-type strain and the ΔclrB, ΔcreA, ΔcreA-ΔclrB, and ΔcreA-gpdA(p)::clrB mutants displayed four classes of genes with similar expression patterns. A heat map that depicts the relative expression levels of selected genes from groups 1 to 4 is shown in Fig 6A.
Group 1 consisted of 91 genes (S4 Table). Most of the genes in this set exhibited strong synergistic induction effects on transcriptional expression in the ΔcreA-gpdA(p)::clrB mutant (Fig 6A). In this cluster, we detected a significant enrichment of genes involved in hydrolase activities (p = 1.5e-30). This group included 9 of 18 cellulase genes (PDE_00507, PDE_09969, PDE_07929, PDE_07928, PDE_05633, PDE_09226, PDE_07124, PDE_07945, and PDE_01261), two β-glucosidase genes (PDE_00579 and PDE_04251), and three sugar transporter genes for cellodextrin transport-1 (PDE_00607), major facilitator superfamily (MFS) maltose transporter (PDE_06576) and monosaccharide transporter (PDE_04857), all of which showed significantly low levels in the ΔcreA-ΔclrB strain versus ΔcreA mutant, but similar transcription activities between the ΔcreA-ΔclrB and ΔclrB mutants, which indicates that the expression levels for these enriched cellulolytic genes were repressed by CreA but ClrB dependent under inducing conditions. Cellodextrin transport-2 genes (PDE_007257 and PDE_00753) were also CreA repressed and ClrB induced but not ClrB dependent. A total of 16 out of the 51 hemicellulase genes (PDE_06649, PDE_02101, PDE_00016, PDE_04478, PDE_01302, PDE_09278, PDE_06023, PDE_04182, PDE_08094, PDE_02514, PDE_00752, PDE_09710, PDE_06067, PDE_05998, PDE_02418, and PDE_07897) were enriched in this subset, and most of which were significantly repressed by CreA, but only four genes (PDE_05998, PDE_06023, PDE_01302, and PDE_09710) were ClrB induced. Noticeably, the expression levels for PDE_05998 (beta-mannosidase) and PDE_06023 (beta-1, 4-mannanase) were ClrB dependent. Interestingly, three genes involved in starch degradation, encoding starch binding domain-containing protein (PDE_01354), glucoamylase (PDE_05527), and α-amylase (PDE_01201) were also downregulated in the ΔclrB mutant and upregulated in the ΔcreA mutant, which suggests that the expression for these amylase genes were tightly co-regulated with the cellulase genes under cellulose growth conditions. A significant enrichment of genes (PDE_09832, PDE_04496, and PDE_09715) involved in the sporulation process (p = 8.2e-5) was also detected. In addition, 21 genes that encode hypothetical proteins were within this dataset (23%). The above data suggested that this cluster might represent the most central components associated with cellulose deconstruction, which indicates that this large synergistic activation by engineering CreA and ClrB might be specific to these cellulolytic target genes.
Group 2 consisted of 27 genes (S4 Table). The expression levels of this set were partially induced in the ΔclrB mutant and repressed in the ΔcreA mutant, but were significantly downregulated in the ΔcreA-gpdA(p)::clrB strain compared with that in the wild-type strain (Fig 6A). Although GO-term analyses revealed that no statistically significant results with the threshold (FDR < 0.05), genes that encode β-glucosidase (Bgl1, PDE_02736), endo-β-1, 4-glucanase (PDE_09267), and two β-xylosidases (PDE_07334 and PDE_08037) were included in this dataset.
Group 3 consisted of 23 genes (S4 Table), and the expression levels of which were partially repressed in the ΔcreA and ΔclrB mutants, and were cumulatively repressed in the ΔcreA-ΔclrB mutant, but were partially recovered in the ΔcreA-gpdA(p)::clrB mutant (Fig 6A). The GO-term analyses of the dataset of 23 genes showed a significant enrichment in the carbohydrate metabolic process (p = 3.6e-7). Five hemicellulase genes that encode PDE_06306, PDE_03572, PDE_07080, PDE_03573, and PDE_08036 and one α-glucosidase (PDE_00400) were also within this group.
Group 4 consisted of 13 genes (S4 Table). The group genes were slightly induced in the ΔclrB mutant and showed no induction in the ΔcreA mutant, but were significantly induced in the ΔcreA-ΔclrB mutant compared with that in the wild-type strain (Fig 6A). One gene that encodes β-xylosidase (PDE_00049) was within this dataset, and eight genes encoded hypothetical proteins.
Secreted proteins are expected to play a crucial role in cellulose degradation because of the nature of cellulolytic system for the deconstruction of plant cell walls. However, whether the cellulolytic protein concentration in secretome potentially correlates with their relative mRNA levels, and how this induction stimulus under cellulose conditions causes P. oxalicum to manipulate its secretome to facilitate cellulose degradation remain unclear. Thus, extensive secretome surveys using label-free LC-MS/MS were conducted to analyze the secretomes under cellulose conditions systematically. Herein, a supernatant from 4-day old wild-type culture grown on cellulose was digested with trypsin and analyzed by LC-MS-MS. For this purpose, 157 nonredundant proteins (P-value < 0.01) were identified based on a single or several peptide entries from the P. oxalicum protein database, the pI of which was concentrated in a pH range of 4–7, and 86 proteins were predicted to be secreted based on SignalP computational analysis (SignalP 4.1 Server, http://www.cbs.dtu.dk/services/SignalP/) (S6 Table). Proteins with predicted activities on carbohydrates in the P. oxalicum wild-type secretome dataset existed, including 10 of 18 predicted cellulases, one β-glucosidase Bgl1, and 10 of 51 predicted hemicellulases (S6 Table).
Subsequently, a question was posted with regard to the extent to which ClrB can be attributed to the regulation of extracellular protein abundances in the cellulolytic system. The total protein in the ΔclrB mutant culture supernatant was only 30% of that in the wild-type strain (Fig 6B). To characterize the secretome changes in response to the crucial regulator ClrB further, a total of 104 predicted secreted proteins in the gpdA(p)::clrB strain (S6 Table) and 61 predicted secreted proteins in the ΔclrB mutant (S6 Table) were identified after the shift to 2% cellulose for 4 days, respectively. These observations demonstrate that ClrB significantly increased the number of extracellular proteins on cellulose. The ΔclrB secretome dataset under cellulose growth condtions included only 5 of 18 predicted cellulases and 5 of 51 predicted hemicellulases in P. oxalicum genome. The gpdA(p)::clrB secretome dataset included 13 of 18 predicted cellulases and 10 of 51 predicted hemicellulases. A comparison between the secretomes of ΔclrB and those of gpdA(p)::clrB grown on cellulose showed that only 49 proteins overlapped (S6 Table), including 5 of 18 cellulases and 4 of 51 hemicellulases. These data indicate that ClrB enhanced various cellulolytic enzymes, including their secretion strength. More importantly, the observed changes of cellulolytic proteins in ΔclrB and gpdA(p)::clrB strains highly correlated with their corresponding mRNA abundances and broadly mirrored the ClrB-specific positive roles for the transcript expression of cellulolytic genes (S6 Table).
Concurrently, to evaluate the CreA-influenced extracellular proteins, we first used sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) to analyze the secretomes of ΔcreA and gpdA(p)::creA cultures under cellulose growth conditions (Fig 6B). The protein pattern of ΔcreA mutant showed more bands than that of gpdA(p)::creA mutant on SDS-PAGE (Fig 6B). The total protein concentration in ΔcreA mutant culture supernatant was 2.6-fold higher than that in the wild-type strain (Fig 6B). We further investigated the extracellular proteins influenced by CreA by adopting the label-free LC-MS/MS analysis to identify and quantify the proteins. A total of 85 and 30 predicted secretion proteins in ΔcreA and gpdA(p)::creA mutants were identified when grown on cellulose (S6 Table), respectively. The SDS-PAGE and LC-MS/MS analysis results for culture secretomes revealed that protein secretion, including protein abundance and distribution, was dramatically repressed by CreA under cellulose growth conditions.
The secretome of ΔcreA-gpdA(p)::clrB mutant was investigated under cellulose conditions to further identify the regulon with ClrB and CreA synergistic effects at the secretome levels. The total amount of the secreted protein in ΔcreA-gpdA(p)::clrB culture supernatant displayed a 7.5-fold increase compared with that in the wild-type strain (Fig 6B). Label-free LC-MS/MS analysis was used, and 97 predicted secretion proteins were identified in ΔcreA-gpdA(p)::clrB mutant (S6 Table). These proteins included 10 of 18 cellulases, and 15 of 51 hemicellulases. No β-glucosidase was detected in ΔcreA-gpdA(p)::clrB mutant.
To assess differences more accurately in protein distribution, the secretomes from ΔclrB, gpdA(p)::clrB, ΔcreA, gpdA(p)::creA, and ΔcreA-gpdA(p)::clrB mutants were combined to locate targets that were the basal components in P. oxalicum secretomes under cellulose growth conditions (Fig 6C and S6 Table). In these datasets, we identified that 25 proteins overlapped (S6 Table), including two cellobiohydrolases (i.e., PDE_07945 and PDE_07124), three endoglucanases (i.e., PDE_09969, PDE_07929, and PDE_09226), four hemicellulases (i.e., PDE_02101, PDE_06023, PDE_04182, and PDE_08094), two extracellular membrane proteins (i.e., PDE_08075 and PDE_02536) that contain common in fungal extracellular membranes domain, three amylases (i.e., PDE_01201 (alpha-amylase Amy13A), PDE_01354 (protein with starch binding domain), and PDE_09417 (glucoamylase GluA/Amy15A)), and five hypothetical proteins (i.e., PDE_03934, PDE_06089, PDE_07106, PDE_09289, and PDE_00667). All these proteins existed in the wild-type strain secretome under the same culture condition (S6 Table). After a 4-h shift from no carbon source, the transcription expression levels of 14 of these proteins increased in cellulose versus those in glucose (S6 Table).
The above data imply that β-glucosidase gene bgl2 (PDE_00579) was the major object regulated by ClrB and CreA at the level of transcription (S5 Fig). However, lack of CreA in the ΔclrB mutant could not recapitulate bgl2 expression level (S5 Fig), suggesting that bgl2 expression level was strictly ClrB-dependent under induction conditions. Considering the upregulation of cellulolytic genes in Δbgl2 mutant [27], we speculated that the expression levels for cellulolytic genes were further enhanced via the overexpression of clrB or deletion of creA in a bgl2 deletion background. In support of this hypothesis, we first constructed Δbgl2-gpdA(p)::clrB and Δbgl2-ΔcreA mutants. The cellulase gene transcription levels and cellulase activities, which were greater than those in each single mutation strain on cellulose (Fig 7A and 7B, and S7A and S7B Fig), as well as the Δbgl2-ΔcreA strains exhibited even more cellulase productions than Δbgl2-gpdA(p)::clrB mutant on cellulose (Fig 7A and 7B, and S7A and S7B Fig). These results suggest that the decrease of intracellular β-glucosidase activity may facilitate the transcriptional induction of cellulolytic genes.
Cellulase gene expression depends on the presence of the inducers and on the positive regulation of activators in cellulolytic fungi [2]. Recent studies indicated that the constitutive expressions of N. crassa clr-2 [20] and T. reesei xyr1 [40] could recapitulate the response to cellulose when incubated without carbon. To assess whether the constitutive expression of clrB, deletions of bgl2 or creA, or combination of these genetic manipulations was sufficient for the induction of cellulase genes independent of inducers, the cellulase expression levels in Δbgl2, gpdA(p)::clrB, ΔcreA, Δbgl2-gpdA(p)::clrB, Δbgl2-ΔcreA, and Δbgl2-gpdA(p)::clrB-ΔcreA mutants were evaluated as opposed to the wild-type strain when cultures were shifted from a glucose medium to a carbon-free medium for 4 hours. The findings revealed that Δbgl2-gpdA(p)::clrB mutant exhibited even more transcriptional abundances on carbon-free medium than on cellulose (Fig 7C and S7C Fig), whereas the “starvation response” for cellulase expression also occurred in Δbgl2-ΔcreA and RE-10 mutants (Fig 7C and S7C Fig). However, such a response was significantly low compared with that in Δbgl2-gpdA(p)::clrB mutant (Fig 7C and S7C Fig). Consistent with these results, the pNPCase and CMCase activities were more than 10-fold higher in Δbgl2-gpdA(p)::clrB strain than those in each single mutant and wild-type strain under carbon-free conditions (S8A and S8B Fig). The creA, amyR and xlnR transcription abundances in Δbgl2-gpdA(p)::clrB, Δbgl2-ΔcreA, Δbgl2-gpdA(p)::clrB-ΔcreA and wild-type strains were assayed to test whether CreA, AmyR and XlnR mediated the synergistic induction in Δbgl2-gpdA(p)::clrB strain when subjected to starvation. The findings indicated that amyR had a 7.7-fold decrease, whereas creA had a 4.3-fold increase and clrB had a 17.5-fold increase in Δbgl2-gpdA(p)::clrB mutant versus that in the wild-type strain (Fig 7D). These data signify that AmyR may share a key role in the “starvation response” for cellulolytic genes and provide a novel insight into the cellulase gene regulatory mechanisms during energy abstinence.
Given the dose-controlled or additive regulation of cellulase genes by ClrB and XlnR presented in gpdA(p)::clrB-PDE_02864(p)::clrB, gpdA(p)::xlnR, and gpdA(p)::clrB-gpdA(p)::xlnR mutants (Fig 4A–4D and S3 Fig), and the synergistic transcriptional induction of cellulolytic genes in Bgl2-deficient background (Fig 7A and 7B, and S7A and S7B Fig), we assessed whether the dose effects of ClrB and XlnR transcriptional abundance were feasible in further enhancing the cellulase expression in triple-mutant RE-10 [39]. We examined this hypothesis by reconstructing two overexpression cassettes (i.e., PDE_02864(p)::clrB-sur and PDE_02864(p)::xlnR-sur), in which the sur cassette (conferring resistance to sulfonylurea) was used as a resistance marker. These overexpression cassettes for clrB and xlnR were separately transformed into RE-10 [39]. The quadruple mutants RE-27 (Δbgl2-ΔcreA-gpdA(p)::clrB-PDE_02864(p)::clrB) and RE-29 (Δbgl2-ΔcreA-gpdA(p)::clrB-PDE_02864(p)::xlnR) were obtained, and their cellulase expression abilities were separately evaluated on cellulose and wheat bran media. Although all these experiments were performed in flasks, both RE-27 and RE-29 mutants showed more cellulolytic and xylanolytic enzyme activities and secretion abilities than RE-10 (Fig 8A–8C and 8E, and S9A–S9F Fig). When grown on a medium with 2% of cellulose as a sole carbon source for 120 h, RE-27 mutant displayed 62.3%, 34.8%, 288.5% and 28.0% greater FPA, pNPCase activity, xylanase activity and total secreted protein level, but 26.3% lower pNPGase activity, respectively, than RE-10 (S9A–S9E Fig). Similarly, the RE-29 mutant showed 55.3%, 44.1%, 255.2% and 20.6% greater FPA, pNPCase activity, xylanase activity and total secreted protein level, but 39.8% lower pNPGase activity, respectively (S9A–S9E Fig). We also observed a significant decrease in amyR expression level for both RE-27 and RE-29 mutants as compared to the wild-type strain on cellulose by q-PCR (Fig 8F). The findings further signify that AmyR may share a key role in the regulatory network for cellulolytic genes. When grown on a wheat bran medium, RE-27 mutant exhibited FPA (8.85±0.66 U/mL), CMCase activity (31.25±0.77 U/mL), xylanase activity (1341.97±172.94 U/mL), amylase activity (125.07±1.32 U/mL) and total secreted protein concentration (16.40±1.08 g/L), and RE-29 mutant also displayed FPA activity (7.58±0.34 U/mL), CMCase activity (31.03±0.29 U/mL), xylanase activity (1285.93±11.12 U/mL), amylase activity (148.82±3.19 U/mL) and total secreted protein concentration (16.16±0.72 g/L), respectively (Fig 8A–8E). These data signify that the dose-controlled regulation mechanisms of the cellulolytic regulators are a promising strategy for cellulolytic fungi to develop enzyme hyper-producers via the RERN technology.
In the above P. oxalicum cellulose regulon and basal secretome components, some enzymes involved in starch degradation were tightly associated with the cellulolytic protein expression on cellulose. The “starvation response” in Δbgl2-gpdA(p)::clrB mutant also dramatically decreased at the amyR expression level under carbon-free conditions. Therefore, P. oxalicum amyR (PDE_03964), an Aspergillus oryzae amyR homolog [41], was considered tightly associated with cellulolytic enzyme production. The strain with the deletion of amyR exhibited visible varying halos on cellulose and starch plates, as well as an identical phenotype on glucose relative to its parental strain (Fig 9A). As such, this strain demonstrated its differential roles in amylase and cellulase expressions. Moreover, this condition suggests that ΔamyR mutant has no defects in glucose uptake, sensing, or metabolism. The FPA in an amyR knockout mutant was about 1.6-fold higher than that in the wild-type strain of P. oxalicum (Fig 9B), while the amyR deletion reduced amylase under cellulose growth conditions (Fig 9C). ΔamyR mutant also displayed higher amounts of cbh1 and eg2 mRNA than that in the wild-type strain according to the results of northern blot (Fig 9D). Nonetheless, such mutant was deficient for transcribing the major glucoamylase gene gluA (PDE_09417) when grown on cellulose. This observation implies that AmyR was the main activator for amylase expression, and it repressed cellulase expression in response to the utilization of cellulose sources.
We constructed ΔamyR-gpdA(p)::clrB and ΔamyR-ΔcreA mutants to investigate whether AmyR plays a negative role in the synergistic/additive transcriptional activation of cellulolytic genes. As predicted, both ΔamyR-gpdA(p)::clrB and ΔamyR-ΔcreA mutants produced more cellulase activities than the strains that contain each individual mutation (Fig 9B) and showed higher transcription levels for cellulase genes under cellulose growth conditions (Fig 9E and 9F). The additive regulation for cellulase gene expression also existed in ΔamyR-ΔxlnR and ΔamyR-Δbgl2 mutants on cellulose (Fig 9G and S10A Fig). Therefore, the deletion of amyR in triple-mutant RE-10 (Δbgl2-ΔcreA-gpdA(p)::clrB) might further enhance cellulase expression under cellulose conditions. This premise also holds true in RE-27 and RE-29 mutants. Correspondingly, we constructed ΔamyR-Δbgl2-ΔcreA-gpdA(p)::clrB quadruple mutant (strain RE-30). However, the resulting strain RE-30 did not obtain greater cellulase expression than its parental triple-mutant RE-10 (S10B and S10C Fig), but AmyR still contributed to the functions of activating amylase expression in RE-10 on cellulose (S10D Fig). These results demonstrate that the deletions of amyR in RE-10 mutants were less effective for inducing cellulase expression than those in wild-type, gpdA(p)::clrB, ΔcreA, and Δbgl2 strains on cellulose.
The fact that RE-30 mutant produced cellulolytic enzymes near RE-10 implies that AmyR played a significantly different regulation activity for cellulase expression in RE-10 mutant than the wild-type strain under cellulose growth conditions. Considering the low expression of amyR in Δbgl2-gpdA(p)::clrB mutant under carbon-free conditions (Fig 7D), we hypothesized that the transcriptional abundance for amyR was also downregulated in RE-10 on cellulose. As predicted, we first observed a significant decrease in ΔcreA-gpdA(p)::clrB strain in RNA-seq data (RPKM: 124.5±5.8 in ΔcreA, 20.0±3.1 in ΔcreA-gpdA(p)::clrB for amyR versus 211.1±2.3 in the wild-type). The q-PCR experiments also revealed that the amyR expression was synergistically downregulated in the RE-10 mutant (Figs 7D and 10A). These data suggest that ClrB and CreA were supposed to participate in the control of the transcriptional response of amyR gene upon exposure to cellulose because the expression of amyR was decreased in gpdA(p)::clrB and ΔcreA mutants, and increased in ΔclrB and gpdA(p)::creA mutants (Fig 10A). The deletion of bgl2 also resulted in the decreased expression of amyR, which may also facilitate the decreased expression of amyR in ΔcreA-gpdA(p)::clrB (Fig 10A). In other words, no differential expression for cellulase expression between RE-30 and RE-10 mutants may be tightly related to the dramatic deregulation of amyR in RE-10 (Figs 7D and 10A). By contrast, the synergistic increase of cellulase induction in ΔcreA-gpdA(p)::clrB and RE-10 mutants may be partially a consequence of the decreased expression of amyR. The extent to which and how AmyR is involved in cellulase expression regulated by ClrB and CreA is still uncertain.
To gain insight into the molecular mechanism that underlies the AmyR-regulated cellulolytic gene expression on cellulose, we evaluated the global changes in ΔamyR mutants with three biological replicates by RNA-Seq. Consequently, 71% of the reads were mapped to the P. oxalicum 114–2 reference genome. The biological replicates of each ΔamyR mutant showed a high Pearson correlation (S6D Fig) and demonstrated the reliability of RNA-Seq. Given a |log2(fold change)|>1 and probability≥0.8 as the threshold, we determined that 131 genes (S7 Table) were upregulated and 579 genes (S8 Table) were downregulated in response to the deletion of amyR compared with the wild-type, respectively.
We then compared the RNA-Seq data for the 579 upregulated genes with that from the wild-type strain and ΔamyR, ΔclrB, ΔcreA, and ΔcreA-gpdA(p)::clrB mutants. The hierarchical clustering of these genes revealed nine groups of genes with similar expression patterns (S11 Fig). The expression levels for groups 1 and 2 increased in ΔcreA and ΔcreA-gpdA(p)::clrB mutants. Group 1 consisted of 195 genes. Within this subset, the proteins with dolichyl-diphosphooligosaccharide-protein glycotransferase activity (p = 3.0e-8) were enriched. Similarly, five genes in this group involved in starch degradation (i.e., PDE_04151, PDE_09417, PDE_05527, PDE_01201, and PDE_01354) were enriched. This case demonstrates that AmyR is the main activator for amylase gene expression. Group 2 composed of 122 genes. The GO-term analysis of these genes displayed a significant enrichment in the molecular function categories of the structural constituent of ribosome (p = 2.6e-101) and rRNA binding (p = 1.8e-5). These results signify that AmyR may play an important role in the translation process (p = 1.8e-38). Groups 3 to 6, 8, and 9 showed no statistically significant results with cutoff (FDR<0.05) via GO-term analyses. Group 7 contained 67 genes, in which 20 genes were enriched in the carboxylic acid metabolic process (p = 9.02e-12).
We also compared the RNA-Seq data for the 131 upregulated genes with that from the wild-type strain and ΔamyR, ΔclrB, ΔcreA, and ΔcreA-gpdA(p)::clrB mutants. The hierarchical clustering of these genes revealed four groups of genes with similar expression patterns (Fig 10B).
Group 1 included 84 genes (S7 Table), which were induced in ΔclrB mutant but repressed in ΔcreA mutant, particularly in ΔcreA-gpdA(p)::clrB mutant. The GO enrichment analysis revealed the induced expressions of the subsets of genes involved in the cellular amino acid metabolic (p = 5.4e-10) and organic acid biosynthetic processes (p = 1.6e-5). One gene cluster (from PDE_01212 to PDE_01220) encoding unclassified proteins was also involved in this dataset.
Group 2 consisted of 22 genes that were induced in ΔcreA mutant (S7 Table), especially in ΔcreA-gpdA(p)::clrB mutant. The GO function annotation of this subset genes revealed that the genes for hydrolase activity (p = 8.2e-12) constitute the largest group, including nine cellulase genes (i.e., PDE_07124, PDE_07945, PDE_05193, PDE_05633, PDE_09226, PDE_07929, PDE_00507, PDE_07928, and PDE_01261), five hemicellulase genes (i.e., PDE_06649, PDE_02101, PDE_09278, PDE_06023, and PDE_08094), β-glucosidase-encoding genes (i.e., PDE_00579 and bgl2), swollenin (i.e., PDE_02102), acetylesterase (i.e., PDE_05194), ABC multidrug transporter (i.e., PDE_07165), and formyltetrahydrofolate deformylase (i.e., PDE_07944). This group also contained cellodextrin transport-1-encoding genes (i.e., PDE_00607), a tetratricopeptide repeat protein (i.e., PDE_08095), and a hypothetical protein (i.e. PDE_06089).
Group 3 comprised 16 genes that were induced in ΔcreA mutant (S7 Table), particularly in ΔcreA-gpdA(p)::clrB and ΔamyR mutants. This gene set was categorized using GO terms. The results illustrated that the genes involved in the carbohydrate metabolic process (p = 1.06e-6) were enriched, including hemicellulases-encoding genes (i.e., PDE_01302, PDE_09710, and PDE_05998), endoglucanase (PDE_09969), β-glucosidase (PDE_04251), α-mannosyltransferase (PDE_09901), and α-xylosidase (PDE_06944). More importantly, a putative cellulose degradation regulator (PDE_05883) was observed within this set. The transcriptional expression for PDE_05883 was ClrB-dependent and repressed by CreA and AmyR. Moreover, this expression showed an additive increase in ΔcreA-gpdA(p)::clrB under cellulose conditions (RPKM: 6.9±0.8 in ΔclrB, 35.3±1.4 in ΔcreA, 3.4±1.1 in ΔcreA-ΔclrB, 62.9±9.6 in ΔcreA-gpdA(p)::clrB, and 51.4±3.8 in ΔamyR for PDE_05883 versus 16.1±0.6 in the wild-type strain). The variations in the expression levels of PDE_05883 in these mutants were further identified via q-PCR experiments (Fig 10C). PDE_05883 encodes a conserved fungal Zn2Cys6 binuclear cluster domain with a significant amino acid homology with N. crassa cellulase essential regulator CLR-2 (NCU08042) for cellulose degradation. PDE_05883 shares 37% identity with CLR-2 in N. crassa (BioEdit, Expect = 5e-096) and a 45% identity with P. oxalicum ClrB (BioEdit, Expect = 5e-0163). Therefore, the gene of PDE_05883 was named as clrB-2, and the corresponding protein was called ClrB-2. The P. oxalicum closest homolog ClrB (BioEdit, Expect = e-131) of N. crassa CLR-2 was identified in this study. The function of ClrB involved in cellulose degradation was also extensively characterized in the preceding discussion. Nonetheless, only limited information has been reported about the molecular mechanism of this ClrB-2. In sum, the above findings provide an additional objective assessment of the role of AmyR in regulating cellulase expression. Likewise, the preceding analyses suggest that the combinatorial cross-regulation of ClrB, CreA, AmyR, and PDE_05883 defining a regulatory network of cellulase expression must be further characterized.
Group 4 consisted of four genes that were repressed in ΔcreA mutant (S7 Table), but were significantly induced in ΔcreA-gpdA(p)::clrB and ΔamyR mutants. These genes involved β-1, 6-glucanase-encoding genes (PDE_02004), succinate semialdehyde dehydrogenase (PDE_05599), 5-nitroimidazole antibiotic resistance protein (PDE_08743), and 4-aminobutyrate aminotransferase (PDE_09301).
The results of the transcriptome analyses, Northern blot, and q-PCR experiments specified above demonstrate that the core cellulolytic genes are tightly regulated by ClrB, CreA, AmyR, and XlnR transcription factors under inducing conditions. The fine-tuned regulation mechanisms allowed us to hypothesize that these central transcription factors may directly bind to the promoters of their core targets. This case was expected because the CreA and XlnR homologs in T. reesei [42] have been determined to be capable of binding to cellulase gene promoters, corresponding to cbh1 in P. oxalicum. To support this hypothesis, GST-tagged ClrB (GST-ClrB), CreA (GST-CreA), and XlnR (GST-XlnR) binding domains were separately expressed in Escherichia coli and were purified. The nucleotide sequences of the putative target gene corresponded to 2 kb cbh1 promoter fragment. The abilities of the recombinant proteins to bind to cbh1 were assessed via electrophoretic mobility shift assay (EMSA). When the concentration of GST-ClrB, GST-CreA, or GST-XlnR fusion proteins increased, slower migrating shifted bands were evident (Fig 11A). However, no shifted band could be observed only with a high-concentration GST (negative control). These findings signify that ClrB, CreA, and XlnR could directly bind to cbh1 promoter region, and the number of binding sites may be more than one.
An important question is whether the direct interactions associated with these regulators play important roles in regulating their target genes by assembling into active transcription complexes, in addition to their direct binding to DNA segments, as observed in T. reesei [42,43]. To investigate this possibility, the full-length open reading frames (ORFs) of the transcription factors ClrB, CreA, AmyR, and XlnR were PCR amplified using cDNA from P. oxalicum 114–2 as the templates. All amplicons were cloned into plasmid pGAD-T7 and separately obtained fusion proteins (i.e., AD-ClrB, AD-CreA, AD-XlnR, and AD-AmyR). Similarly, the full-length creA, amyR, and xlnR were cloned into the partner plasmid pGBK-T7 and resulted in BD-CreA, BD-AmyR, and BD-XlnR. Protein–protein interaction assay was performed. The results showed that the strains with interactions between ClrB and AmyR, XlnR and ClrB, XlnR and AmyR, and XlnR and CreA could grow on synthetic drop-out (SD) plates that lack Leu, Trp, and His (QDO, Clontech) (Fig 11B). Likewise, the results revealed that XlnR interacted directly with ClrB, CreA, and AmyR as well as with ClrB and AmyR in vitro.
Several conserved and essential transcription factors for cellulase and hemicellulase genes have been recently described in cellulolytic fungi; yet, the regulatory patterns of these factors have not been thoroughly analyzed [10,15–17,23]. In this study, a single-gene disruptant mutant library of the transcription regulators in P. oxalicum was constructed, and 20 transcription factors that play a pivotal role in the activation or inhibition of cellulose deconstruction were identified. Of these screened transcription factors, ClrB, CreA, XlnR, and AmyR were selected for further analyses. The cellulase expression regulated by additive cellulolytic effectors was observed. The results correspondingly provided comprehensive information for deciphering and redesigning a cellulase expression regulation network for rational engineering of cellulase hyper-producers.
In this study, a master transcription factor ClrB was exceptionally identified in P. oxalicum transcription factor mutant set screening. This key transcription factor positively regulated the cellulolytic gene expression, and its deletion strain exhibited dramatically reduced cellulase activities (Fig 1B). However, this factor was not strictly required for xylanase gene expression. The homology search of P. oxalicum ClrB within some fungal proteomes showed that the homologs of ClrB (i.e., Clr-2 in N.crassa, ClrB in A. nidulans, and ManR in A. oryzae) [16,17] were recently determined and were required to induce major cellulases, some major hemicellulases, and mannanolytic enzyme gene expression. The search of T. reesei protein databases via a Basic Local Alignment Search Tool using a ClrB/Clr-2 query revealed that a protein (Trire2: 26163) with low sequence identity existed. However, the homology search of N.crassa Clr-2 within P. oxalicum proteome illustrated that the protein sequences for ClrB (BioEdit, Expect = 5e-0163) and PDE_05883 (BioEdit, Expect = 5e-096) have 45% and 37% identity with Clr-2 sequence [17], respectively. These phenomena for two Clr-2 homologs in P. oxalicum proteome were not observed in N. crassa and T. reesei proteomes. This case suggests that differential-inducing mechanisms for cellulase expression may exist among cellulolytic fungi.
The homologs of regulator XlnR were the most conserved in cellulolytic fungi [10–12]. However, P. oxalicum XlnR does not have the same transcriptional inducing ability for cellulolytic and xylanolytic genes as in others [10–12]. Significant differences in these regulatory patterns for cellulase genes were observed in P. oxalicum ΔxlnR and T. reesei Δxyr1 mutants [10]. The lack of XlnR homolog in T. reesei eliminated cellulase expression, but not in P. oxalicum ΔxlnR mutant (Fig 2A and 2B). The deletion of P. oxalicum xlnR slightly reduced the transcript levels of some cellulases and abolished the major xylanase expression under induction conditions (Fig 2A), which were similar to that in N. crassa Δxlr-1 mutant [12]. These findings suggest that the transcriptional regulation of lignocellulose-degrading enzymes mediated by XlnR homologs may be highly conserved in various filamentous fungi, but may also have interesting differences.
CreA/CRE1/CRE-1 is a wide-domain master regulator of carbon metabolism identified in filamentous fungi [7–9]. This regulator allows an organism to utilize a preferred carbon source but hinders it from metabolizing complex carbon sources, including cellulose [7–9]. In this study, the function of CreA homologs in repressing cellulolytic and xylanolytic gene expressions was conserved among cellulolytic fungi (Fig 2A and 2B). CreA homologs generally play an important role among cellulolytic fungi by linking CCR to developmental programs, including the conidia formation and hyphal morphology in T. reesei [23] and P. oxalicum (Fig 5B). This study determined that some transcription factors (i.e., PDE_07199, PDE_04095, and PDE_08372) are involved in both developmental programs and cellulase induction expression (Table 1). These regulators are conserved in cellulolytic fungi (Table 1). Therefore, the possible existence of an intimate crosstalk among certain developmental processes, such as sporulation and cellulase production pathways, is mediated by some regulators in ascomycete fungi.
By using RNA-seq data, we showed that expression of amyR was synergistically decreased in the ΔcreA-gpdA(p)::clrB mutant. The lack of AmyR significantly induced cellulase expression and decreased the expression for amylase genes involved in starch degradation (Fig 9A–9C). As such, AmyR may control the balance between starch and cellulose utilization by inducing and/or repressing cellulolytic and amylolytic gene expressions in P. oxalicum, respectively. The multiple-sequence alignment analysis showed that P. oxalicum AmyR shares a weak homology with N. crassa COL26 (NCU07788, 23% sequence identity, E value = 2e-038 by BioEdit) [44] and T. reesei BglR (Trire2: 52368, 24% sequence identity, E value = 3e-031 by BioEdit) [18], but is highly homologous to T. reesei, a functionally uncharacterized Zn(II)2Cys6-type fungal-specific transcription factor (Trire2: 55105, 38% sequence identity, E value = 9e-095 by BioEdit). Interestingly, the regulatory functions of AmyR gene were distinct from those of N. crassa COL26 [44] and T. reesei BglR [18]. During its initial response to cellulose, P. oxalicum ΔamyR mutant exhibited induction and did not decrease the cellulolytic gene expression as in N. crassa Δcol-26 mutant [44]. Moreover, N. crassa COL26 obviously repressed cre-1 transcription to promote the relief of CCR [44], but the creA transcript abundance only slightly increased in P. oxalicum ΔamyR mutant. N. crassa Δcol-26 mutant exhibited a severe growth defect on glucose, but not in P. oxalicum ΔamyR mutant. Although both T. reesei BglR [18] and P. oxalicum AmyR mutants displayed an elevated cellulase expression under inducing conditions, they demonstrated distinct regulatory trends to β-glucosidase expression. This difference might be related to the functional studies of BglR in T. reesei mutant PC-3-7 containing bgl2 mutation and other uncharacterized mutations [18]. Nevertheless, T. reesei 55105 (Trire2), which is much more homologous to P. oxalicum AmyR than T. reesei BglR (Trire2: 52368), may be a candidate regulator involved in cellulase expression regulation.
The cellulase expression in P. oxalicum is a highly coordinated process regulated by a suite of cellulolytic transcription factors (i.e., ClrB, CreA, XlnR, AmyR and ClrB-2) and other novel uncharacterized regulators. In this study, the cellulolytic regulators ClrB, XlnR, AmyR and ClrB-2 were significantly regulated at the transcriptional levels during their growth on glucose, but slightly at the early phase for under cellulose growth conditions (Figs 5A and 12A). The CreA tightly regulated the expression of clrB, xlnR, amyR and clrB-2 in response to environmental carbon. These data suggested that CreA might have a cascade regulation because it repressed the activator genes for AmyR, ClrB, ClrB-2 and XlnR as well as the structural genes whose expression was upregulated by ClrB, XlnR and ClrB-2 (Fig 12A). This “double-lock” regulation of cellulolytic genes mediated by regulator homologues in cellulolytic fungi might be general, which could facilitate the fast conversion of carbon metabolism from favored carbon sources to cellulose and hemicellulose utilization. This cascade regulation mechanism mediated by P. oxalicum CreA was similar to the pathway described Cre1-mediated double repression of xyr1 and xyn1 in H. jecorina [10]. Some similar situations with regard to the repression of the transcription of A. nidulans ethanol and xylan regulons have been previously reported [45,46]. The deletion of creA resulted in the slightly decreased expression of amyR (Fig 10A), whose absence further led to the upregulation of cellulase genes (Fig 12A). Similarly, the overexpression of ClrB led to a decreased expression of amyR, whose expression level was synergistically downregulated in RE-10 mutant, but increased in gpdA(p)::clrB-PDE_02864(p)::clrB mutant (Fig 10A). These data implied that regulatory function of ClrB and CreA on amyR expression may be important for cascade regulation for cellulolytic genes under cellulose growth conditions.
Another key finding of this study is that the transcriptional expression of ClrB-2, a novel regulator, is responsive to ClrB, XlnR, CreA and AmyR, which implied that ClrB-2 may mediate the cascade transcriptional regulation for cellulolytic genes by ClrB, XlnR, CreA and AmyR (Figs 5A, 10C and 12A). We did not systematically determine how cellulolytic genes could transform in clrB-2 mutant yet. Nonetheless, the variable expression levels on cellulose in these regulator mutants suggest that ClrB-2 is one of the most interesting target in P. oxalicum cellulolytic regulatory networks. Whether ClrB, CreA, XlnR, and AmyR converge to exert their partial regulatory function via ClrB-2 on cellulolytic gene expression must be urgently elucidated.
The synergistic and collective regulations of cellulase expression by cellulolytic regulators are still elusive and largely uncharacterized in cellulolytic fungi. Thus far, no research has systematically investigated whether or how the most central cellulolytic factors CreA and ClrB homologs perform the synergistic regulation of cellulase genes. In this study, ΔcreA-gpdA(p)::clrB strain yielded strong synergistic effects on cellulase expression (Figs 1A–1E, 2A and 2B). This observation indicates that the full induction of cellulase genes requires not only an exclusive inducer-induction and activation by positive regulators, but also the release of negative transcription factors. In accordance to these synergistic regulatory mechanisms in P. oxalicum, we constructed significantly higher cellulase hyper-producer RE-27 and RE-29 than the triple-mutant RE-10 (Fig 8A and 8B). We believe that the RERN technology presented here will be a valuable contribution in transforming some non-industrial model species (e.g., N. crassa and A. nidulans) into more industrially relevant species.
A very interesting finding in this study is that the cellulase expression increased evidently accompany with the increase of the copy number and the efficiency of its promoter of clrB or xlnR gene (Fig 8A and 8B, and S3A–S3D Fig), indicating that a tunable cellulase expression may be controlled by the activator concentration under cellulose conditions. These data signify that the cellulase expression is not only dependent on the presence of the activators ClrB and XlnR, but also severely dependent on their dose effects of ClrB and XlnR transcriptional abundances.
The other issue that must be explored is how cross-correlations occur between the cellulolytic regulators. To evaluate the role of CreA in CreA-mediated repression of cellulase and hemicellulase gene expressions, we developed three possible hypotheses. a) Putative CreA binding sites overlap with the putative ClrB or XlnR binding sites, and CreA could preferentially bind to the sites and/or block their binding to target promoters by competition. A recent study identified the presence of putative cis-regulatory elements recognized by both XYR1 and CRE1 and spaced in XYR1- and CRE1-dependent cellulase gene promoters [42]. b) CreA could stably associate with ClrB or XlnR components to form a heterocomplex, thereby making ClrB and XlnR completely non-functional for the induction of cellulolytic and xylanolytic genes. In this study, CreA and XlnR were assumed to directly interact with each other according to the yeast two-hybrid assay (Fig 11B and 11C). c) The activity of transcription factor is influenced by intracellular protein post-translational modifications, such as phosphorylation. These modifications may influence the ability of the transcription factor to bind to its binding sites [47]. In sum, the findings presented above support our proposition that biological relevances may exist between CreA and ClrB as well as between CreA and XlnR, which tightly regulate the expression of cellulase and hemicellulase genes.
Beta-glucosidases are the conserved components required in cellulose deconstruction, the number of which significantly varies among the genomes of cellulolytic fungi [24,48,49]. The lack of P. oxalicum intracellular Bgl2 contributes to the increase of cellulase expression [27], but not in T. reesei [50] and N. crassa [51]. These findings have raised the question as to how Bgl2 mediates the carbon metabolism involved in signal cascades in relation to the regulation of cellulase gene expression, which may reflect the general trend of cellobiose/cellodextrins-induced cellulase expression in cellulolytic fungi [27,50,51]. Some β-glucosidases in T. reesei [50] and N. crassa. [51] have been observed to play important roles in balancing cellobiose production and metabolism in intra- and extra-cellular environments. Given the general phenomenon that cellobiose induces cellulase expression in cellulolytic fungi [27,50,51], P. oxalicum Δbgl2-gpdA(p)::clrB mutant showed a strongly elevated cellulase expression (Fig 7A and 7B), which could be partially ascribed to the signal induction cascade mediated by cellobiose/cellodextrins from cellulose (Fig 12A). The robust induction of cellulase expression in Δbgl2-ΔcreA mutant was remarkably greater than that in each deletion mutant on cellulose (Fig 7A and 7B). This observation supports the premise that the CCR mediated by CreA increased when the major predicted intracellular β-glucosidase was absent under cellulose growth conditions. Although the lack of Cre-1 in N. crassa triple β-glucosidase mutant showed higher concentrations of secreted active cellulases than that in wild-type strain on cellobiose, it did not facilitate protein production and cellulase induction on cellulose [51]. In sum, ClrB positively regulated the transcriptional expression of bgl2 (S5 Fig), but its deletion conversely enhanced the signal cascade activation regulation by ClrB and/or the repression regulation of CCR mediated by CreA (Figs 7A, 7B and 12A). In addition, we also determined that the cellulase expression in Δbgl2-ΔamyR double mutant was further enhanced compared with each individual deletion strain (S10A Fig). These results revealed that the functional regulations for cellulase expression by these cellulolytic regulators may be sensitive to inducers in intracellular environments. This finding implies that the combination of intracellular cellodextrin induction and redesigned cellulolytic transcription factor regulation in P. oxalicum may be general for the full induction of cellulase expression on cellulose.
Many recent studies have attempted to robustly produce cellulases without inducers, but the molecular mechanism of cellulase induction under non-inducing conditions has remained elusive in diverse cellulolytic species [20,40]. Recently published data also showed that the misexpression of N. crassa clr-2 through Pccg-1 [20] and Ptcu1-driven expression of xyr1 [40] was sufficient for inducer-free cellulase expression. However, the cellulolytic gene transcript between P. oxalicum gpdA(p)::clrB and wild-type strains had no obvious differences under non-inducing conditions. Such case was similar in the XlnR activator in the gpdA(p)::xlnR strain under CreA-mediated CCR, and cbh1 transcription was still remarkably in low-level expression in the ΔcreA mutant on a carbon-free medium, which indicates the full induction requirement of cellulase gene expression, as evident in T. reesei [25] or N. crassa [51]. However, the cellulolytic gene transcription levels were obviously upregulated in the ΔcreA-gpdA(p)::clrB mutant under repressing conditions. These findings implied that the lack of CreA contributed to the activating function for ClrB on cellulolytic genes even in glucose. Notably, the Δbgl2-gpdA(p)::clrB strain exhibited an induction transcription of core cellulase genes for several orders of magnitude increase than the wild-type strain that shifted to a carbon-free medium (Fig 7C and S7C Fig). No bulk of cellodextrins was apparently transported into the cell to induce cellulase gene expression when subjected to starvation, but the induction abilities of these core cellulase regulons in the Δbgl2-gpdA(p)::clrB mutant under non-inducing conditions were comparable to under cellulose conditions (Fig 7C and S7C Fig). The amyR expression level had a 7.7-fold decrease, clrB had a 17.5-fold increase, and clrB-2 had a 9.3-fold increase in the Δbgl2-gpdA(p)::clrB mutant versus wild-type strain under carbon-free conditions (Fig 7D), which implied that AmyR, ClrB, and ClrB-2 were possibly tightly involved in cellulase expression regulatory during energy abstinence, as well as that under cellulose growth conditions. These were novel findings that indicated the “starvation response” for cellulase genes in P. oxalicum and other diverse cellulolytic fungi.
Cellulase formation apparently occurred because of consistent respective regulators, including the characterized or novel transcription factors identified in P. oxalicum in this context. Cellulase synergistic and dose-controlled regulatory systems mediated by diverse cellulolytic effectors were observed in the system mutation of this study for P. oxalicum regulators. Using this model (Fig 12A), the accumulation of intracellular cellodextrins can trigger signaling cascades that include expression of cellulase genes repressed by CreA and AmyR and activated by ClrB and XlnR. However, our data also support that the transcriptional regulation for CreA, AmyR, ClrB and XlnR genes is a powerful part of the regulatory network of cellulase gene expression. In the early cellulolytic induction, ClrB functions to repress expression of amyR, whose expression level is activated by CreA and reduced in Bgl2-deficient background (Fig 12A). Moreover, transcriptional expression for ClrB, AmyR, XlnR and ClrB-2 genes is also significantly repressed by CCR mediated by CreA in the presence of glucose (Fig 12A). The data established ClrB as a focal point to regulate cellulase expression by integrating other regulators and the target genes of these regulators, which refined our understanding on transcriptional regulatory network as a “seesaw model” in which coordinated regulation of cellulolytic genes was established through activators and repressors counteraction (Fig 12B–12D). These observations also suggested the hypotheses that the rational design of cellulase or high-value protein super producers might be guided in the future for the combinatorial effects of diverse cellulolytic effectors.
All the strains used in this study are listed in the S9 Table and are grown on Vogel’s medium that contain 2% glucose (mass/volume percent), unless otherwise noted. The P. oxalicum wild-type strain 114–2 (CGMCC 5302) was used as parental strain throughout this study. Hygromycin B, pyrithiamine, phosphinothricin, and sulfonylurea were added to the media with final concentrations of 200, 300, 1.6, and 4 μg/mL used for transformant selection, respectively. Vogel’s 50x salts (1,000 mL) was used: 125 g Na3Citrate•2H2O, 250 g KH2PO4, 100 g NH4NO3, 10 g MgSO4•7H2O, 5 g CaCl2•2H2O, 0.25 mg biotin, and 5 ml trace element solution (5 g Citric acid•H2O, 5 g ZnSO4•7H2O, 1 g Fe(NH4)2(SO4)2•6H2O, 0.25 g CuSO4•5H2O, 0.05 g MnSO4•1H2O, 0.05 g H3BO3, and 0.05 g Na2MoO4•2H2O, which were dissolved in distilled water; the resulting total volume was 100 mL). The wheat bran medium (mass/volume percent) was composed of corn cob residue (2.0%), Avicel (0.6%), wheat bran (4.6%), soybean cake powder (1.0%), (NH4)2SO4 (0.2%), NaNO3 (0.28%), urea (0.1%), KH2PO4 (0.3%), and MgSO4 (0.05%). Vogel’s medium that contained 1 M sorbitol was used in all the transformation experiments. Vogel’s medium with 2% Avicel was used as a sole carbon source to induce cellulase expression for cellulase activity assays, q-PCR, Northern blot, or RNA-seq analyses. The fluid mediums of the P. oxalicum strains were all cultivated in Erlenmeyer flasks at 30°C in constant light and 200 rpm agitation rate. The cultivation was performed on plates by adding agar in the fluid medium as a solidifying agent at 30°C in constant light.
P. oxalicum protoplast prepared according to modified methods, as described by Gruber et al. [52]. Seven to eight PDA plates were prepared, and 50 μL of fresh spore solution was streaked out on every cellophane-covered PDA plate. The plates were incubated at 30°C for 11–12 hours. Three milliliters of this protoplasting solution [0.075 g lysing enzymes up to 25 mL solution A (1.2 M sorbitol and 0.1 M KH2PO4, and pH = 5.6] were pipetted into a sterile petri dish, and one cellophane disc with freshly grown mycelium was added. Subsequently, 3 mL of protoplasting solution was readded. The petri dish was incubated at 30°C for 150 min. The final protoplast suspension was filtered into a sterile 50 mL centrifuge tube through a lens cleaning tissue in a glass funnel. The suspension was centrifuged for 10 min at 2000 rpm and 4°C in a swing-out rotor. The supernatant was cautiously decanted, and the pellet was resuspended in 10 mL of solution B (1 M sorbitol, 50 mM CaCl2, 10 mM TrisHCl, and pH = 7.5). The suspension was recentrifuged for 10 min at 2000 rpm and 4°C, and then the supernatant was cautiously decanted and the protoplasts were resuspended in 0.5 mL of solution B. The protoplasts were stored in ice. The following components: 200 μL protoplast suspension, 10 μL DNA fragment, and 50 μL of solution C (25% PEG 6000, 50 mM CaCl2, 10 mM TrisHCl, and pH = 7.5), were added into a 10 mL centrifuge tube and were mixed gently, and then the mixture was incubated on ice for 20 min. Two milliliters of solution C (room temperature) were added to the transformation mixture, which was mixed gently. The mixture was incubated for 5 min at room temperature, and then 4 mL of solution B was added and was mixed gently. All the transformation mixture was added to 30 mL Vogel’s medium (55°C) that contained hygromycin B, pyrithiamine, phosphinothricin, or sulfonylurea, and was then mixed shortly and was poured onto the bottom of the Vogel’s medium. Transformants would be visible for 3–4 days for hygromycin B, pyrithiamine, or phosphinothricin, and for 6–8 days for sulfonylurea at 30°C.
The P. oxalicum genome is available in DDBJ/EMBL/GenBank (under the accession number AGIH00000000) or at http://genome.jgi.doe.gov/Penox1/Penox1.home.html. Transcription factors were identified and annotated according to InterPro IDs in the Fungal Transcription Factor Database [53]. The transcription factor deletion strains, where each coding region of the transcription factor was substituted with a selective marker ptra gene [31], were constructed from the P. oxalicum strain Δpku70::hph [33] as follows. Deletion cassettes conferring resistance to pyrithiamine hydrobromide were constructed according to the double-joint PCR strategy [32]. Primer pairs for (x)-F1+(x)-ptraR and (x)-ptraF+(x)-R1 (S1 Table) were designed using the Primer 5 software, and the pairs were used to amplify the upstream and downstream fragments for 1000–1500 bp on either side of the encoding regions for each single target gene. The ptra selectable marker cassette was PCR-amplified from the pME2892 plasmid [54] with the PtraF1+PtraR1 (S1 Table) primer pair and was PCR product-purified. The upstream and downstream fragments contained 25 bp homology to the ptra cassette sequence. The primer pair (x)-F2+(x)-R2 (S1 Table) was used to produce final deletion cassettes through double-joint fusion PCR [the program used to fuse the three fragments: 94°C 2 min for 10 cycles (94°C for 30 s, 58°C for 10 min, and 72°C for 3 min), 72°C 5 min] [32]. PCR experiments were performed in final volumes of 50|μL that contained 1 unit of Trans HIFI DNA polymerase, 0.2|mM dNTP, and 0.4|μM of each primer. The program used to amplify the fused knockout cassettes was as follows: one cycle of 94°C (120 s), 30 cycles of 94°C (30 s), 58°C (30 s), and 72°C (1 min for every 1 kb of amplified product), followed by a final 10|min at 72°C. All the PCRs were performed in a Bio-Rad DNA Engine Peltier Thermal Cycler.
Each knockout cassette was independently transformed into P. oxalicum strain Δpku70::hph protoplasts [33]. The mature transformant conidia from the Vogel’s medium slant was streaked onto a Vogel’s medium plate that contained pyrithiamine hydrobromide. At least three Ptra-resistant colonies obtained from each assay were analyzed through diagnostic PCR to confirm the deletion using the primer pairs (x)-F1+ptraYZR or ptraYZF+(x)-R1 (S1 Table). Within the pair, the primer (x)-F1 or (x)-R1 (S1 Table) was located outside the transforming deletion fragment of the genome, and the primer ptraYZR or ptraYZF (S1 Table) was unique to the ptra sequence. These obtained strains constituted a stock of the P. oxalicum transcription factor deletion mutant.
The complementation cassette that conferred resistance to hygromycin B was used to transform into the corresponding mutant to identify that the interesting phenotype(s) observed for the transcription factor gene deletion mutants was indeed caused by the deletion of a relevant gene. The upstream and downstream fragments contained 25 bp homology to the hph cassette sequence. clrB and amyR wild-type allele complementation cassettes were obtained by amplifying the upstream fragments (encompassing the 1.5 kb promoter, the open reading frame, and 0.5 kb 3’ untranslated region) using the primer pairs clrB-F1+clrBHPH-R and AmyR-F1+AmyRHPH-R (S1 Table), respectively. The 1.5 kb downstream flanking segments of the 3’ untranslated region from the P. oxalicum genome DNA were amplified through the primer pairs clrBHPH-F+clrB-R1 and AmyRHPH-F+AmyR-R1 (S1 Table). The 1.8|kb hph gene fragment was amplified from the pSilent-1 plasmid [55] with the primers Hphs-F and Hphs-R (S1 Table). These three PCR fragments were ligated through Double-joint PCR and were amplified through the nest primer pairs clrB-F2+clrB-R2 and AmyR-F2+AmyR-R2 (S1 Table). The resulting clrB and amyR complementation cassettes were transformed into ΔclrB::ptra and ΔamyR::ptra mutants, and the complementation strains RclrB and RamyR of these cassettes were obtained, respectively.
The clrB promoter was replaced with the gpdA (glyceraldehyde-3-phosphate dehydrogenase) promoter from A. nidulans [37]. Moreover, 1,314 bp of the gpdA promoter was amplified from the plasmid pAN7-1 [56] using the primers PgpdA-F1 and PgpdA-R1 (S1 Table). Subsequently, 2,008 bp ptra selectable marker cassette was PCR-amplified with the primer pair PtraF1+PtraR1 (S1 Table) and was PCR product-purified. Furthermore, 3,148 bp of the clrB open reading frame and 3’ untranslated region were amplified with the primer pair clrB-Fa+clrB-Ra (S1 Table), and this fragment overlapped with the gpdA promoter and ptra fragment by 25|bp at the ends of this fragment. These 6375 bp PCR products were then ligated in the order of gpdA(p)::clrB-ptra by splicing through double-joint PCR with the nest primers PgpdA-F2 and PtraR1 (S1 Table).
A similar strategy was used to construct the creA overexpression cassette under the influence of the P. oxalicum gpdA promoter. Moreover, 1749 bp of the gpdA promoter was amplified from the P. oxalicum 114–2 genome DNA through the primers PGP-F1 and PGP-R (S1 Table). In addition, 1890 bp of the hph cassette was amplified from the plasmid pSilent-1 [55] with the primers Hphs-F and Hphs-R (S1 Table). Moreover, 1743 bp of the creA open reading frame and 3’ untranslated region was amplified using the primer pair GPcre-F+GPcre-R (S1 Table), and this fragment overlapped with the gpdA promoter and hph fragment by 25|bp at the ends of this fragment. These PCR products were then ligated in the order of gpdA(p)::creA-hph by splicing through double-joint PCR with the primers PGP-F2 and HPH-R1 (S1 Table). The resulting 5248 bp gpdA(p)::creA-hph overexpression cassette was transformed into P. oxalicum wild-type and gpdA(p)::clrB-ptra strain protoplasts, and the gpdA(p)::creA and gpdA(p)::creA-gpdA(p)::clrB mutants were obtained, respectively.
Similarly, the xlnR promoter was replaced with the gpdA promoter from A. nidulans. Moreover, 1314 bp of the gpdA promoter was amplified from the plasmid pAN7-1 [56] through the primer pair PgpdA-F1 and PgpdA-R1 (S1 Table). In addition, 1890 bp of the hph cassette was amplified from the plasmid pSilent-1 [55]. The xlnR open reading frames and 3’ untranslated region were amplified with the primer pair XlnR-Fa+XlnR-RH2 (S1 Table), and this fragment overlapped with the gpdA promoter and the hph fragment by 25|bp at the ends of this fragment. These PCR products were then ligated in the order of gpdA(p)-xlnR-hph by splicing through double-joint PCR with the primers PgpdA-F2 and Hphs-R1 (S1 Table). The resulting 6675 bp gpdA(p)::xlnR-hph overexpression cassette was transformed into gpdA(p)::clrB strain protoplasts.
The resulting gpdA(p)::clrB-ptra, gpdA(p)::creA-hph, and gpdA(p)::xlnR-hph overexpression cassettes were used to transform P. oxalicum wild-type strain protoplasts. The gpdA(p)::creA-ptra, gpdA(p)::clrB-ptra, and gpdA(p)::xlnR-hph overexpression mutants were selected on Vogel’s medium plate that contained hygromycin B or pyrithiamine hydrobromide.
Double-joint PCR was performed to construct the creA knockout cassette, with the hph cassette flanked by 1.5 kb upstream (CreA-F1+Crehph-R) (S1 Table) and 1.5 kb downstream (Crehph-F+CreA-R1) (S1 Table) of the creA ORF. Moreover, 1.8 kb of the hph cassette was amplified from the plasmid pSilent-1 [55]. The ΔcreA::hph final deletion cassette fragment was obtained through the primer pair (CreA-F2+CreA-R2) (S1 Table) with the three fragments above used as a PCR template. The ΔcreA::hph cassettes were transformed into P. oxalicum wild-type, gpdA(p)::clrB-ptra, and ΔclrB strain protoplasts. The selection of the ΔcreA, ΔcreA-gpdA(p)::clrB, and ΔcreA-ΔclrB transformants was consistent with the previous study.
The ΔxlnR cassettes conferred resistance to hygromycin B. The hph selectable marker cassette was PCR-amplified with the primers Hph-F1 and Hph-R1 (S1 Table) from the plasmid pAN7-1 [56]. The upstream and downstream flanking fragments were amplified with the primer pairs XlnR-F1+XyrHph-R and HphXyr-F+XlnR-R1 (S1 Table), which contained 25 bp homology to the hph cassette sequence. The primer pair XlnR-F2 and XlnR-R2 (S1 Table) was used to produce final deletion cassettes through a double-joint fusion PCR. The resulting ΔxlnR::hph knockout cassette was used to transform wild-type protoplasts, and the P. oxalicum ΔxlnR mutant was obtained. The ΔxlnR::hph knockout cassette was used to transform the ΔclrB protoplasts, and then the ΔxlnR-ΔclrB mutant was obtained. The above gpdA(p)::xlnR-hph overexpression cassette was used to transform the gpdA(p)::clrB-ptra strain protoplasts, and the gpdA(p)::clrB-gpdA(p)::xlnR mutant was constructed.
The upstream and downstream fragments of bgl2 ORF were amplified with the primer pairs Bgl2-F1+Bgl2hph-R and Bgl2hph-F+Bgl2-R1 (S1 Table). Moreover, 1.8 kb of the hph cassette was amplified from the plasmid pSilent-1 [55]. The final Δbgl2::hph fragment was obtained through the primer pair Bgl2-F2+Bgl2-R2 (S1 Table), with the three fragments above used as a PCR template. The Δbgl2::hph cassette was used to transform the protoplasts of the ΔcreA and gpdA(p)::clrB mutants, and then the Δbgl2-ΔcreA and Δbgl2-gpdA(p)::clrB mutants were obtained, respectively.
PDE_02864(p)::clrB-hph overexpression cassettes that conferred resistance to hygromycin B was constructed. The hph selectable marker cassette was PCR-amplified with the primers Hphs-F and HPH-R1 from the plasmid pSilent-1 [55]. The PDE_02864 promoter, and clrB open reading frame and 3’ untranslated region were amplified with the primer pairs DF1+DClrB-R and ClrB-F+ClrBHPH-R (S1 Table) from the P. oxalicum genome DNA, respectively. The primer pair DF2+HPH-R1 (S1 Table) was used to produce the final PDE_02864(p)::clrB-hph overexpression cassette via double-joint fusion PCR. The resulting PDE_02864(p)::clrB-hph overexpression cassette was used to transform the wild-type and gpdA(p)::clrB strain protoplasts, and then the PDE_02864(p)::clrB-hph and gpdA(p)::clrB-PDE_02864(p)::clrB mutants were obtained, respectively.
The upstream and downstream flanking fragments of the amyR encoding region were amplified with the primer pairs PDE_03964-F1+amyRHph-R and amyRHph-F+PDE_03964-R1 (S1 Table), respectively. The final ΔamyR::hph fragment was obtained through the nest primer pair PDE_03964-F2+PDE_03964-R2 (S1 Table) with the three fragments (amyR flanking sequences and hph encoding cassette) used as templates via double-joint PCR. The ΔamyR::hph cassette was used to transform the gpdA(p)::clrB-ptra protoplasts, and the ΔamyR-gpdA(p)::clrB mutant was obtained. The ΔamyR::ptra (ΔPDE_03964::ptra, from TF knock-out cassette set) knockout cassette was used to transform the Δbgl2::hph, ΔcreA::hph, and ΔxlnR::hph mutants, and the ΔamyR-ΔcreA, ΔamyR-Δbgl2, ΔamyR-ΔxlnR mutants were obtained, respectively.
First, the ΔamyR::sur knockout cassette, and PDE_02864(p)::clrB-sur and PDE_02864(p)::xlnR-sur overexpression cassettes that conferred resistance to sulfonylurea were constructed. The sur selectable marker cassette was PCR-amplified with the primers Sur-F1 and Sur-R1 (S1 Table) from the plasmid pCB1536 [57]. The upstream and downstream flanking fragments for the ΔamyR-sur knockout cassette were amplified with the primer pairs PDE_03964-F1+amyRsur-R and amyRsur-F+PDE_03964-R1 (S1 Table), which contained 25 bp homology to the sur cassette sequence, respectively. The primer pair PDE_03964-F2+PDE_03964-R2 (S1 Table) was used to produce the final deletion cassette ΔamyR-sur through a double-joint fusion PCR. The resulting ΔamyR-sur knockout cassette was used to transform the RE-10 (Δbgl2-ΔcreA-gpdA(p)::clrB) protoplasts, and the P. oxalicum and RE-30 (ΔamyR-Δbgl2-ΔcreA-gpdA(p)::clrB) mutants were obtained. The clrB promoter was replaced with the PDE_02864 (encoding 40S ribosomal protein S8) promoter from P. oxalicum. The PDE_02864 promoter sequence was amplified with the primer pair DF1+DP-R. In addition, the clrB and xlnR open reading frames, and the 3’ untranslated regions were amplified with the primer pairs DClrB-F+ClrBsur-R and DXlnR-F+XlnRSur-R, respectively. The primer pair DF2+Sur-R1 (S1 Table) was used to produce the final PDE_02864(p)::clrB-sur and PDE_02864(p)::xlnR-sur overexpression cassettes via double-joint fusion PCR. The resulting PDE_02864(p)::clrB-sur and PDE_02864(p)::xlnR-sur overexpression cassettes were used to transform the RE-10 protoplasts, and then the RE-27 (Δbgl2-ΔcreA-gpdA(p)::clrB-PDE_02864(p)::clrB) and RE-29 (Δbgl2-ΔcreA-gpdA(p)::clrB- PDE_02864(p)::xlnR) mutants were obtained.
Fungal genomic DNA was isolated, as described previously [58]. The Primer 5 software was used to identify the appropriate restriction enzymes for the Southern blot analysis of the gene replacement mutants. The fragments used for the probes were amplified with the primers presented in the S1 Table. A DIG-High Prime labeling kit was used to label the knockout cassette flank fragment probes. The homokaryons and integration patterns of the transforming cassettes in the genome were confirmed through Southern blot analysis, as described in the manipulations.
Colony morphology and conidiation were analyzed after inoculating the Vogel’s medium plates that contained 2% glucose, 2% xylan, 2% starch, or 1% cellulose as sole carbon source or potato dextrose agar (PDA) medium at 30°C for 5 days. The halo sizes that varied among the P. oxalicum strains were measured on cellulose or starch plates by adding Triton X-100 to a final concentration of 0.5%. Iodine solution (6 g of KI, 0.6 g of I2 in 100 mL of H2O) was used to indicate visualize starch-degrading colonies through the hydrolysis halo at room temperature for 10 min, which determined if the amylase expression was affected in the P. oxalicum strain. P. oxalicum hyphae and conidia were microscopically examined through lactophenol cotton blue staining (0.05 g cotton blue, 20 g phenol crystals, 40 mL glycerol, 20 mL lactic acid, and 20 mL distilled water).
Cellulase was produced in a 500 mL flask that contained 100 mL of fluid medium through a two-step cultivation procedure. Strains were first grown at 30°C in 100 mL of medium that contained 2 g of glucose as a carbon source and were then regulated at pH 5.5 and 200 rpm for 20 hours. The cultures were collected through vacuum drum filtration during this second step, and 0.5 g vegetative mycelia was added to 100 mL of Vogel’s medium that contained 2% cellulose as carbon source or wheat bran medium at an initial pH of 5.5 at 30°C and 200 rpm. Culture supernatants (crude enzyme) were diluted with sodium acetate buffer solution (SABF, 0.2 M, pH 4.8). Enzymatic hydrolyses of the polysaccharides were also performed in SABF (0.2 M, pH 4.8). The filter paper enzyme (FPA), endoglucanase (CMCase), xylanase, and amylase activities of the culture supernatants (diluted samples) were assayed using a DNS reagent (10 g 3, 5-dinitrosalicylic acid, 20 g sodium hydroxide, 200 g sodium potassium tartrate, 2.0 g redistilled phenol, and 0.50 g sodium sulfite anhydrous per 1000 mL DNS reagent) against Whatman No. 1 filter paper, carboxymethylcellulose sodium salt (CMC-Na), xylan (from beechwood), and soluble starch. CMC-Na, xylan, or starch was dissolved in SABF to a final concentration of 1% (mass/volume percent, m/v %), and then the mixture was left overnight and was shaken well before using. The following components were added in a 2.0 mL reaction mixture: 0.5 mL diluted culture supernatants and 1.5 mL CMC-Na, xylan, or starch solution for CMCase, xylanase, or amylase activity assays, respectively; and 2.0 mL diluted culture supernatants and 50 mg Whatman No. 1 filter paper for FPA assay into 25 mL colorimetric tube. The mixture was mixed gently and the reaction mixture was incubated for FPA measurement in a 50°C water bath for 1 hour, for CMCase and xylanase activity measurements at 50°C for 30 min, and for amylase activity measurement at 40°C for 10 min. Three milliliters of DNS reagent were then added to stop the reaction. A blank tube (with boiled crude enzyme) was used as control to correct any reducing sugar present in the crude enzyme samples. The tubes were placed in boiling water for 10 min, 20 mL distilled water was added, 200 μL of reaction mixture was pipetted, and the absorbance was determined at 540 nm. The cellobiohydrolase (pNPCase) and β-glucosidase (pNPGase) activities were measured by using 4-Nitrophenyl β-D-cellobioside (pNPC) and 4-Nitrophenyl β-D-glucopyranoside (pNPG) as substrates, respectively. The pNPC or pNPG was dissolved in SABF to a final concentration of 1 mg/mL. Moreover, 50 μL of pNPC solution (containing 1 mg/mL D-Glucono-δ-lactone) or 50 μL of pNPG solution and 100 μL of diluted culture supernatants were mixed, and then the mixtures were incubated in a 50°C water bath for 30 min. The reaction was stopped by adding 0.15 mL of sodium carbonate solution (10%, m/v), then 200 μL of these reaction mixtures was pipetted, and the absorbance was measured at 420 nm. One unit of enzyme activity was defined as the amount of enzyme required to release 1 μmol of glycoside bonds of the substrate per minute under defined assay conditions. Independent triplicate cultures were sampled and analyzed.
The total protein was determined using a Bradford assay kit according to the instructions of the manufacturer.
Freshly harvested conidia of the wild-type strain or mutants were inoculated with 106 conidia/mL into 100 mL Vogel’s medium that contained 2% glucose, and then grown for 22 hours at 30°C. Mycelia were harvested via vacuum filtration, and then washed with Vogel’s medium without a carbon source, followed by 2 h growth in 100 mL Vogel’s medium without a carbon source. Subsequently, mycelia were harvested via vacuum filtration, and then transferred into Vogel’s medium that contained 2% cellulose for 4, 8, 22, and 46 hours, with 2% glucose for 4 hours, or into a medium without any carbon source for 4 hours. Mycelia were harvested via vacuum filtration, and then immediately finely ground under liquid nitrogen, and then 1 mL of TRIzol reagent was added per 50–100 mg powder. The total RNA was isolated according to the instructions of the manufacturer.
The total amount (2 μg) of mRNA loaded was normalized by using rRNA as a loading control. The probes used for the Northern blot analysis were the partial cDNAs of cbh1 (PDE_07945), eg2 (PDE_09226), and xyn1 (PDE_08094) that were cloned from the P. oxalicum genome DNA via PCR with primer pairs (S1 Table). The probes were labeled using a DIG Northern Starter kit, and Northern blot analysis was performed according to the instructions of the manufacturer.
Putative targets were validated through q-PCR. Each strain was cultured independently from the Northern blot and RNA-seq experiments. cDNA was synthesized from the total RNA by applying a reagent kit with a gDNA eraser according to the instructions of the manufacturer. The obtained cDNA was applied for quantitative reverse transcription-PCR experiments. All q-PCR amplification was performed in 20 μl total volume with 7.4 μl distilled water, 0.8 μl of each primer (10 mM), 10 μl SYBR Premix Ex TaqII, and 1 μl template cDNA through the following program [59]: 95°C for 2 min, 40 cycles at 95°C for 10 sec, and 30 s at 61°C. The fluorescence signal was measured at the end of each extension step at 80°C. A melting curve program with a temperature gradient of 0.1°C per second from 65°C to 95°C was performed. The corresponding primers are shown in the S1 Table. The quantity and copy number of each target gene were calculated using a standard curve. Six 10-fold serial dilutions of purified DNA template (0.5 kb–1.0 kb) were prepared for the target genes to determine the standard curve of each target gene. The correlation coefficient (R2) for each standard curve was verified to be 0.99 or greater. The transcriptional expression of transcription factor genes were measured by q-PCR, and their expression levels were normalized to wild-type. Gene expression levels for cellulase genes were measured by q-PCR using actin (PDE_01092) as a control and normalized to expression levels by actin values/10000. Three biological replicates were performed on the same 96-well plate by using cultures grown in parallel. Data processing and statistical analyses were performed using Microsoft Excel.
A cDNA library prepared from mRNA was organized according to standard protocols. Quality control was implemented using the Real-Time PCR Systems. All the cDNA libraries were sequenced on the Illumina platform. Sequenced reads were mapped against predicted transcripts from the P. oxalicum 114–2 genome using the SOAP2 software for short oligonucleotide alignment (http://soap.genomics.org.cn/soapaligner.html) [60]. Transcript abundance (Reads Per Kb per Million reads, RPKM) [61] was estimated with RPKM = [# of mapped reads]/([length of transcript]/1000)/([total reads]/10^6). The differential gene expression was analyzed using DESeq software package [62] and NOIseq v2.10 (http://www.bioconductor.org/packages/release/bioc/html/NOISeq.html) with |log2(fold change)|>1 and probability≥0.8 as thresholds [63]. The biological replicates used for RNA-seq were highly reproducible. These datasets (RPKM) were subjected to hierarchical cluster analysis using the software HCE3.5 (http://www.cs.umd.edu/hcil/hce/) to determine the groups of genes with similar expression patterns for a different group of regulons. The Blast2go software v3.0 was used for the gene ontology analyses (https://www.blast2go.com/) [64]. Secreted proteins were predicted using SignalP 4.1 (http://www.cbs.dtu.dk/services/SignalP/). Secondary metabolism gene clusters were identified using annotated proteins in P. oxalicum [24]. Protein sequence alignments were performed among the P. oxalicum, T. reesei, and N. crassa proteomes using the BioEdit Sequence Alignment Editor software (http://www.mbio.ncsu.edu/BioEdit/bioedit.html).
Culture supernatants were collected after shifting to cellulose for 96 hours by filtrating using 0.22 μm PES membrane, and then the supernatants were desalted with 10 kDa molecular cut-off membrane, and were precipitated by acetone and trichloroacetic acid (20:1). The obtained protein powders were dissolved in denaturation buffer (0.5 M Tris-HCL, 2.75 mM EDTA, 6 M Guanadine-HCL), and were then reduced using 1 M DTT at 37°C for 1 hour. The following alkylation was performed using iodoacetamide for 2 hours away from light, and the samples were desalted and collected using a Microcon YM-10 Centrifugal Filter Unit. The obtained protein samples were digested thoroughly using trypsin for 12 hours, and these peptide mixtures were desalted with a ZipTip C18 column. These collected secretome samples were further separated on a C18-reversed phase column and then directly mounted on the electrospray ion source of a mass spectrometer. The peptides were subjected to nanoelectrospray ionization, followed by tandem mass spectrometry (MS/MS) in an LTQ Orbitrap Velos Pro coupled with high-performance liquid chromatography. Intact peptides were detected in the Orbitrap at 60000 resolution. Peptides were selected for MS/MS using a collision-induced dissociation operating mode with 35% normalized collision energy setting. Ion fragments were detected in the LTQ Orbitrap. A data-dependent procedure that alternated between one MS scan, followed by 10 MS/MS scans, was applied for the 10 most abundant precursor ions above the 5000 threshold ion count in the MS survey scan with the following Dynamic Exclusion settings: 2 repeat counts, 30 s repeat duration, and 120 s exclusion duration. An electrospray voltage of 2.2 kV was applied. For the MS scans, the m/z scan range was 350 Da to 1800 Da. Mass spectrometry data processing was performed using the Mass-Lynx software (version 4.1, Waters). LC-MS/MS analysis data were identified by searching the P. oxalicum protein database (http://genome.jgi.doe.gov/Penox1/Penox1.home.html).
The DNA-binding domain of ClrB (1–163 amino acids) was PCR-amplified from the P. oxalicum 114–2 genome DNA using the primers ClrB-RTF and ClrB-RTR (S1 Table). The resulting amplicon was digested using EcoRI and BamHI, and then inserted into the expression vector pGEX-4T-1 at the GST downstream with corresponding restriction sites. The correct fusion plasmid was confirmed via nucleotide sequencing, and then transformed into E. coli BL21 (DE3). Parental and recombinant strains were cultured in a lysogeny broth medium and were induced by 0.05 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at 30°C and 150 rpm for 8 hours to induce the GST alone and the GST-ClrB binding-domain production. Protein purification was performed by using Glutathione Sepharose 4B beads after cell ultrasonic decomposition according to the product manual. The protein concentration was then measured using a Bradford Kit according to the instructions of the manufacturer. A 2 kb promoter of PDE_07945 was then amplified from the P. oxalicum 114–2 genome DNA using the primers PDE_07945-F and PDE_07945-R (S1 Table), and then purified using a gel extraction kit according to the instructions of the manufacturer. The DNA concentration was determined using a UV-Vis Spectrophotometer Q5000. GST alone or GST-ClrB binding-domain was mixed in the binding buffer (containing 100 mM Tris-HCl, 100 mM KCl, 10 mM EDTA, 2.5 mM DTT, and 20% Ficoll-400, supplementing 1 μg Poly(dI-dC) to avoid unspecific binding) with purified DNA (20 ng) at room temperature for 10 min for electrophoretic mobility shift assay. The protein-DNA mix was then separated via gel electrophoresis, stained by ethidium bromide, and then visualized. The protein and DNA complex were retardant relative to free DNA. Binding reaction was performed by gradient increasing the protein content to avoid artificial results. GST alone was used as negative control.
The Matchmaker GAL4 two-hybrid system 3 was used for yeast two-hybrid assays. Full-length ORFs of the transcription factors ClrB, CreA, AmyR, and XlnR were PCR-amplified using the P. oxalicum 114–2 cDNA as templates (with primers listed in S1 Table). All the amplicons were cloned into the plasmid pGAD-T7 with the corresponding restriction sites, leading to AD-ClrB, AD-CreA, AD-XlnR, and AD-AmyR. Similarly, the full-length creA, amyR, and xlnR were cloned into the partner plasmid pGBK-T7, which generated BD-CreA, BD-AmyR, and BD-XlnR, respectively. All the fused plasmids were confirmed via nucleotide sequencing. The plasmids pGAD-T7 or AD-X coupling with pGBK-T7 or BD-X in pair were co-transformed into S. cerevisiae AH109 and were cultivated on an SD medium without Trp and Leu for 3 days at 30°C. Protein–protein interaction assay was performed on SD plates without Leu, Trp, and His, and on YPD plates to avoid artificial results. pGBKT7-53/pGADT7-T and pGBKT7-Lam/pGADT7-T pairs were used as internal positive and negative controls, respectively. Moreover, AD-ClrB, AD-CreA, AD-xlnR, and AD-AmyR paired with an empty pGBK-T7 introduced into AH109 were used to eliminate false positive results.
The RNA-seq data have been deposited in NCBI's Gene Expression Omnibus with accession number GSE69298 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE69298).
|
10.1371/journal.ppat.1000777 | Kaposi's Sarcoma Associated Herpes Virus (KSHV) Induced COX-2: A Key Factor in Latency, Inflammation, Angiogenesis, Cell Survival and Invasion | Kaposi's sarcoma (KS), an enigmatic endothelial cell vascular neoplasm, is characterized by the proliferation of spindle shaped endothelial cells, inflammatory cytokines (ICs), growth factors (GFs) and angiogenic factors. KSHV is etiologically linked to KS and expresses its latent genes in KS lesion endothelial cells. Primary infection of human micro vascular endothelial cells (HMVEC-d) results in the establishment of latent infection and reprogramming of host genes, and cyclooxygenase-2 (COX-2) is one of the highly up-regulated genes. Our previous study suggested a role for COX-2 in the establishment and maintenance of KSHV latency. Here, we examined the role of COX-2 in the induction of ICs, GFs, angiogenesis and invasive events occurring during KSHV de novo infection of endothelial cells. A significant amount of COX-2 was detected in KS tissue sections. Telomerase-immortalized human umbilical vein endothelial cells supporting KSHV stable latency (TIVE-LTC) expressed elevated levels of functional COX-2 and microsomal PGE2 synthase (m-PGES), and secreted the predominant eicosanoid inflammatory metabolite PGE2. Infected HMVEC-d and TIVE-LTC cells secreted a variety of ICs, GFs, angiogenic factors and matrix metalloproteinases (MMPs), which were significantly abrogated by COX-2 inhibition either by chemical inhibitors or by siRNA. The ability of these factors to induce tube formation of uninfected endothelial cells was also inhibited. PGE2, secreted early during KSHV infection, profoundly increased the adhesion of uninfected endothelial cells to fibronectin by activating the small G protein Rac1. COX-2 inhibition considerably reduced KSHV latent ORF73 gene expression and survival of TIVE-LTC cells. Collectively, these studies underscore the pivotal role of KSHV induced COX-2/PGE2 in creating KS lesion like microenvironment during de novo infection. Since COX-2 plays multiple roles in KSHV latent gene expression, which themselves are powerful mediators of cytokine induction, anti-apoptosis, cell survival and viral genome maintainence, effective inhibition of COX-2 via well-characterized clinically approved COX-2 inhibitors could potentially be used in treatment to control latent KSHV infection and ameliorate KS.
| Kaposi's sarcoma associated herpes virus (KSHV), with a 160 kb DNA genome, has evolved with two distinct life cycle phases, namely latency and lytic replication. KS, a complex angioproliferative disease, is regulated by a balance between pro-angiogenic and anti-angiogenic factors. In our previous study, we showed that KSHV modulates host factors COX-2/PGE2 for its own advantage to promote its latent (persistent) infection. The premise that COX-2 is involved in growth and progression of several types of solid cancers and inflammation associated diseases has been well documented but has never been studied in KS. Here, utilizing COX-2 inhibition strategies, including chemical inhibition and a gene silencing approach, we systematically identified the potential role of KSHV induced COX-2/PGE2 in viral pathogenesis related events such as secretion of inflammatory and angiogenic cytokines, MMPs and cell adhesion in de novo infected or latently infected endothelial cells. We report that COX-2/PGE2 inhibition down-regulates viral latent gene expression and survival of latently infected endothelial cells. The data emanating from our in vitro studies is valuable, informative and requires further examination using an in vitro angiogenic model and in vivo nude mice model to further validate COX-2 as a novel therapeutic to target latent infection and the associated diseases like KS.
| KSHV, the most recently discovered human tumor virus, is etiologically associated with Kaposi sarcoma (KS), primary effusion lymphoma (PEL) and multicentric Castleman's disease (MCD) [1],[2]. KS, an AIDS defining condition, is a highly disseminated unusual angiogenic tumor of proliferative endothelial cells and displays a very strong resemblance to chronic inflammation [1],[2],[3],[4]. KS is responsible for significant morbidity and mortality in HIV-infected patients in the developing world [1],[2]. KS lesions are characterized by proliferating spindle shaped endothelial cells, neo-vascular structures, inflammatory cells, and an abundance of inflammatory cytokines (ICs), growth factors (GFs), angiogenic factors and invasive factors such as basic and acidic fibroblast growth factor (bFGF, aFGF), interleukin-1α and β (IL-1α and -1β), granulocyte-monocyte colony stimulating factor (GM-CSF), platelet derived growth factor β (PDGF-β), vascular endothelial growth factor (VEGF), interferon-γ (IFNγ), interlukin 6 (IL-6), tumor necrosis factor α (TNF-α) [2], angiopoietin-2 (Ang2) [5], angiogenin [6], heme oxygenase-1 (HO-1) [7], transforming growth factor β (TGF-β) [8], adhesion molecules like inter-cellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule-1(VCAM-1), and matrix metalloproteinases (MMPs) like MMP-1, -2, -3, -9, and -19. Cell cultures composed of characteristic spindle-shaped tumor cells have been established from KS lesion explants by the addition of cytokines like TNF-α, TNF-β, IFN-γ, IL-1, IL-6, GM-CSF and oncostatin M [1],[2],[9],[10] highlighting the role of these paracrine factors in KS lesion cell survival. A crucial step in KS progression is its striking neovascularization and angiogenesis, which is regulated by aberrant production of angiogenic and anti-angiogenic factors from the infected host cells, uninfected neighboring cells or both [11]. It is believed that KSHV tumorigenesis and disease progression are predominantly driven by both paracrine and autocrine mechanisms, where KSHV infection could induce an angiogenic, GFs-, and MMPs- rich microenvironment and a strong cytokine network. These events, via their synergistic actions and communications, could support continued proliferation and migration of KSHV latently infected cells [2],[12].
KSHV encodes ∼86 putative open reading frames (ORFs) of which at least 22 are potentially immuno-modulatory and anti-apoptotic [13],[14]. Among these are the genes “pirated” from the host or cellular homologues like viral G-protein-coupled receptor (vGPCR), vIL-6, viral interferon regulatory factors (vIRFs 1-4), viral chemokines (vCCLs 1-3), MHC class I down-regulating E3 ligases K3 and K5 (MIR1 and MIR2), and Kaposin B [13],[14],[15],[16],[17],[18],[19]. These proteins are capable of regulating cellular cytokine expression, antagonizing host IFN mediated anti-viral responses and immune evasion, thus suggesting the importance of ICs in KSHV-associated pathogenesis [20]. In KS lesion endothelial cells, KSHV is in a latent form with about 10–20 copies of the viral episome per cell and lytic replication is observed in a low percentage of infiltrating inflammatory monocytes. Low percentages of KSHV-infected cells in KS lesions are typical spindle cells which are thought to represent neoplastic cells in these lesions and these cells occassionaly express lytic gene products, undergo lytic reactivation and may support productive replication [21]. During latency, KSHV expresses a battery of genes such as ORF73 (LANA-1), ORF72 (vCyclin), ORF71 (K13/vFLIP), and ORFK12 (Kaposin A, B and C), as well as 12 distinct miRNAs. These gene products obviously must be facilitating the establishment of lifelong latency in its host and in survival against the host intrinsic, innate and adaptive immune surveillance mechanisms [22],[23],[24].
Cytokines have been shown to play important roles in viral immune evasion and lytic replication. ICs like IL-1β, IL-6, and TNF-α have been shown to inhibit KSHV lytic gene transcription in endothelial cells [20]. Host immune responses against KSHV control viral replication and viral spread and exert a selective pressure on the virus to establish a latent state which allows the virus to evade the subsequent wave of adaptive immune host responses following an effective innate immune response. Therefore, studying KSHV infection linked cytokines is relevant to understand viral multifactor patho-biology, its mechanisms to induce neoplasia, and for developing therapeutic interventions.
Apart from viral genes, this virus has also evolved strategies to regulate host gene expression to create a microenvironment that is conducive for viral persistence. One of the host genes that is highly induced upon de novo infection of human microvascular endothelial cells (HMVEC-d) and human foreskin fibroblast (HFF) cells is cyclooxygenase-2 (COX-2) [25],[26]. KSHV-encoded early lytic-cycle membrane protein vGPCR and cell–cell contact deregulator protein K15 have also been shown to trigger COX-2 induction [27],[28]. COX, the rate limiting enzyme of prostaglandin synthesis has three isoforms identified to date, namely COX-1, COX-2, and COX-3. COX-1 is constitutively expressed and displays characteristics of a housekeeping gene in most tissues. In contrast, COX-2 is a key enzyme for prostanoid biosynthesis [29],[30]. COX-2 possesses pro-angiogenic, anti-apoptotic properties and is up-regulated by mitogenic and inflammatory stimuli [29],[30]. COX-2 has also been implicated in the progression and angiogenesis of several cancers [30],[31], and is widely regarded as a potential pharmacological target for preventing and treating malignancies [31],[32],[33].
In our earlier studies, we demonstrated robust COX-2 gene expression and high levels of PGE2 secretion by KSHV during primary infection of HMVEC-d and HFF cells [26]. Inhibition of COX-2 by NS-398 and indomethacin (Indo) did not affect KSHV binding, internalization of virus, or it's trafficking to the infected cell nuclei [26]. Intriguingly, latent ORF73 promoter activity and gene expression were significantly reduced by COX-2 inhibitors, and this inhibition was relieved by exogenous supplementation with PGE2 [26]. In contrast, lytic ORF50 gene expression and ORF50 promoter activity were unaffected indicating that KSHV has evolved to utilize COX-2 mediated inflammatory responses induced during infection of endothelial cells for the maintenance of viral latent gene expression [26].
Since COX-2 is linked to inflammation and KS is a chronic inflammation associated malignancy, we hypothesized that COX-2 is one of the virus's triggered pathogenic factors with key roles in inflammation, neo-angiogenesis, cell proliferation, and invasion associated with the KS lesions. When we tested this hypothesis by using chemical inhibitors of COX-2 or by COX-2 silencing, we uncovered evidence for the role of COX-2/PGE2 in viral latent gene expression, in pro-inflammatory, angiogenic and invasive events occurring during KSHV de novo infection of endothelial cells as well as the survival of latently infected endothelial cells. Effective reduction in secretion of autocrine and paracrine factors involved in KSHV pathogenesis during early and later time points of infection, along with cell cycle arrest observed in latently infected endothelial cells, suggested that COX-2 inhibition based therapy might provide an effective way to treat the angio-proliferative KS lesions.
HMVEC-d (CC-2543; Lonza Walkersville, Maryland) were cultured in endothelial basal medium 2 (EBM-2) with growth factors (Lonza Walkersville). HEK 293T (human embryonic kidney cells stably expressing SV40 large T-antigen) cells were grown in Dulbecco's modified Eagle's medium (Gibco BRL, Grand Island, New York) supplemented with 10% heat-inactivated fetal bovine serum (HyClone, Logan, UT), 2 mM L-glutamine, and antibiotics [26],[34],[35]. HUVECs (Lonza Walkersville) were cultivated in EGM-2 (Lonza Walkersville). Cells were typically used between 5 to 7 passages. TIVE (telomerase-immortalized human umbilical vein endothelial) and TIVE-LTC (long-term-infected TIVE) cells (a gift from Dr. Rolf Renne, Department of Molecular Genetics and Microbiology, University of Florida) were cultured in EBM-2 with growth factors. All cells were cultured in LPS-free medium. All stock preparations of purified KSHV were monitored for endotoxin contamination by standard Limulus assay (Limulus amebocyte lysate endochrome; Charles River Endosafe, Charleston, S.C.) as recommended by the manufacturer [26]. The COX-1 and COX-2 inhibitor Indomethacin and the COX-2-specific inhibitor NS-398 [N-(2-cyclohexyloxy-4-nitrophenyl)-methanesulfonamide] were purchased from Calbiochem, La Jolla, Calif. Both inhibitors were reconstituted in dimethyl sulfoxide (DMSO) and DMSO was used as solvent control for all experiments involving treatments with inhibitors.
Induction of the KSHV lytic cycle in BCBL-1 cells, supernatant collection, and virus purification procedures were described previously [26]. KSHV DNA was extracted from the virus, and the copy numbers were quantitated by real-time DNA PCR using primers amplifying the KSHV ORF 73 gene as described previously [26],[34],[35].
Lentiviral constructs expressing shRNAs against human COX-2 and control laminA/C were generated as described [36]. These shRNA transcription products are known to be processed by the cell to produce the functional siRNA sequence. AACTGCTCAACACCGGAAT (si-COX-2-1) and CACCATCAATGCAAGTTCT (si-COX-2-2) sequences were used as COX-2 shRNAs. Testing for reduction by shRNA constructs was done by transfection of target plasmids and shRNA lentiviral construct plasmids into 293T cells followed by protein extraction and immunoblot analysis to select the best candidates. Third generation lentiviral vectors were produced using a four-plasmid transfection system as previously described [36]. Briefly, 293T cells were transfected with vector and packaging plasmids. Culture supernatant was harvested 2 and 3 days post-transfection. Cell debris from the supernatant was cleared by filtration through 0.22-µm filters, concentrated by ultracentrifugation, and lentiviral vector titers were estimated by flow cytometery (eGFP expression). HMVEC-d cells were transduced with either si-COX-2 or si-lamin (si-C) to produce si-COX-2-HMVEC-d or si-C-HMVEC-d.
Total RNA was converted to cDNA, relative abundance of target gene mRNA was measured by qRT-PCR using the delta-delta method (ratio, 2−[ΔCt sample–ΔCt control]) as decribed previously [37]. Primer sequences are given in Table S2. PCR amplifications without cDNA were performed as negative controls.
Confluent HMVEC-d cells in eight-well chamber slides (Nalge Nunc International, Naperville, Il.) were either uninfected or infected (30 DNA copies/ cell) for 24h. For COX-2 and VEGF-A immunostaining, cells were fixed with 4% paraformaldehyde (PFA), permeabilized with 0.4% Triton-X 100 and stained with anti-COX-2 goat polyclonal antibody (Cayman chemical, Ann Arbor, Mich.) and anti-VEGF-A monoclonal antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) overnight at 4°C. Cells were washed and incubated with 1∶200 dilution of Alexa 594-coupled anti-mouse antibody or Alexa 488-coupled anti-goat antibody (Molecular Probes, Eugene, OR) for 1 h at RT. Nuclei were visualized by using DAPI (Ex358/Em461; Molecular Probes) as counter stain. Stained cells were washed and viewed with appropriate filters under a fluorescence microscope with the Nikon metamorph digital imaging system. HMVEC-d cells were either uninfected or infected (30 DNA copies/ cell) for 2 h, 4 and 5 days and stained for KSHV latency protein ORF73 (5d) and lytic protein ORF59 (2 h, 4d and 5d) using antibodies generated in Prof. Bala Chandran's laboratory. TIVE and TIVE-LTC cells were also co-stained for ORF73 and COX-2 using the above mentioned procedures.
Sections from lymph nodes and skin biopsy samples of healthy subjects and KS+ patients were obtained from the AIDS and Cancer Specimen Resource (ACSR). Sections were deparaffinized with Histochoice clearing reagent and hydrated with water before microwave treatment in 1 mmol/l EDTA (pH 8.0) for 15 min for antigen retrieval, and then blocked with blocking solution (2% donkey serum, and 0.3% Triton X-100 in PBS). Sections were incubated with the primary antibodies against COX-2 (Cell signaling technology Inc.) or ORF73 (generated in Prof. Bala Chandran's laboratory) overnight at 4°C. These sections were incubated with rat-polymer-HRP (Biocare medical) for 15 min, washed and developed using DAB reagent (DAKO). Counterstaining was done by hematoxylin. Similar procedure was followed for COX-2 staining of ACSR KS Screening tissue microarray (TMA) 09-1 (Table S1).
Sections from lymph nodes and skin biopsy samples of KS+ patients and control samples were deparaffinized and hydrated with water before antigen retrieval using DAKO target retriever solution in steamer for 20 min. Slides were cooled, rinsed, blocked using 1% BSA in 0.025% Triton X-100-PBS for 30 min and used for double staining of COX-2 and monoclonal mouse anti-human CD31 (DAKO, Denmark). Sections were washed and incubated with 1∶200 dilution of Alexa 594-coupled anti-mouse antibody or Alexa 488-coupled anti-rabbit antibody (Molecular Probes) for 1h at RT. Nuclei were visualized using DAPI and stained cells were viewed under an Olympus Confocal laser scanning microscope (Fluoview FV10i).
Conditioned medium was obtained from serum-starved, untreated, Indo, NS-398-pretreated HMVEC-d, si-COX-2-HMVEC-d or si-C-HMVEC-d cells either uninfected or KSHV (30 DNA copies/ cell) infected for different time points. Conditioned media were spun at 1,000 rpm for 10′ at 4°C to remove the particulates and assayed immediately. Total soluble protein was quantified by bicinchoninic acid (BCA) protein assay (Pierce, Rockford, IL) prior to use ensuring equal protein concentration for studying the cytokine profile by human protein cytokine arrays 3.1 and 5.1 from Ray Biotech (Norcross, GA) and Ray Biotech human MMP antibody array-1 which detects 10 human MMPs in one experiment. Uninfected HMVEC-d/si-COX-2-HMVEC-d/si-C-HMVEC-d cells were used as controls for KSHV infected HMVEC-d/si-COX-2-HMVEC-d/si-C-HMVEC-d cells, respectively. The cytokine detection membranes were blocked with blocking buffer for 1 h at RT and then incubated with conditioned media at 4°C overnight. The membranes were washed, incubated with 1 ml of primary biotin-conjugated antibody at RT for 2h, washed, incubated with 2 ml of horseradish peroxidase-conjugated streptavidin at RT for 45′, and developed using enhanced-chemiluminescence (ECL). Signal intensities were quantitated using an Alpha Inotech image analysis system. Signal intensities from all the arrays were normalized to the same background levels with positive and negative controls using Ray Biotech human antibody array 3.1/5.1 and MMP antibody array-1 analysis software.
Conditioned media used for the MMP detection by MMP-antibody array were also used for determination of active/total MMP-2 and MMP-9 using MMP-2 and MMP-9-enzyme-linked immunosorbent assay (ELISA) kits from Anaspec (San Jose, CA) as per manufacturer's protocols. These kits were optimized to detect levels of total MMPs and their activities using a 5-FAM/QXL™520 FRET peptide as substrate with its fluorescence monitored at Ex/Em = 490 nm/520 nm upon proteolytic cleavage. These novel assays use FRET substrates that incorporate QXL™520 non-fluorescent dyes, the best quencher available for 5-FAM and are designed for the specific quantitation of the activity of a particular MMP in a mixed biological sample, which may contain multiple MMPs. A monoclonal anti-human-MMP antibody was used to pull down both the pro- and active- forms of an MMP from the mixture, and proteolytic activity quantitated using a 5-FAM/QXL™520 FRET peptide. Similarly, active/total MMP-2 and MMP-9 levels were detected in the conditioned media obtained from TIVE cells and COX inhibitor pretreated or untreated TIVE-LTC cells.
Conditioned medium, obtained as described was used for quantitating VEGF-A and -C levels using QuantiGlo ELISA kits (R and D Systems, Minneapolis, MN) as per procedures recommended by the manufacturer. Each sample was run in duplicate and the assay repeated a minimum of three times. Quantities of VEGF-A or VEGF-C released were normalized by protein content.
Levels of PGE2 in the supernatants of uninfected and KSHV infected HMVEC-d cells, and inhibitor treated or COX-2/lamin silenced and then uninfected or KSHV infected HMVEC-d cells, or TIVE and TIVE-LTC, COX inhibitor treated or untreated TIVE-LTC cells were measured by ELISA (Cayman Chemicals) according to the manufacturer's instructions [26]. Data are expressed as the amount of PGE2 produced (pg/ml) per 105 cells.
Conditioned media were collected from the variously treated cells for analysis on matrigel and the assay was performed as per manufacturer's instructions (BD Biosciences, Mountain View, CA). Briefly, 5×104 HUVEC or HMVEC-d cells were plated on a Matrigel-coated 96-well plate with medium alone or medium obtained from cells treated with inhibitors alone or cells pretreated with inhibitors and then KSHV (30 DNA copies/ cell) infected or the cells silenced for COX-2 and then infected for 24 h. After 16 h in 5% CO2 at 37°C, the plate was examined for capillary tube formation under an inverted microscope and photographed. Each assay was done in duplicate and each experiment was repeated three times. Angiogenic index, a measure of tube formation, was calculated based on the number of branch points formed from each node per field at 10X original magnification. Differences between the numbers of tube formations in 3D-conditioned Matrigel assays were subject to student's t-test analysis. Similar assay was performed using the conditioned media obtained from 24 h serum starved TIVE cells and COX inhibitors or solvent pretreated or untreated TIVE-LTC cells.
Cell extracts were quantitated by BCA protein assay, then equal amounts of protein (20 µg/lane) were separated on SDS-PAGE, electrotransferred to 0.45-µm nitrocellulose membranes. The membranes were blocked with 5% BSA, probed with anti-COX-2, active-Rac, total-Rac , β-actin and tubulin antibodies and visualized using an ECL detection system [26].
Maxisorp II Nunc ELISA plates (Roskilde, Denmark) were coated with fibronectin (5 µg/ml), or poly-lysine (2.5 µg/cm2) overnight at 4°C and adhesion assays were performed. Briefly, HMVEC-d cells were resuspended in serum-free EBM-2 medium and plated at 3×104 cells in 200 µl/well and incubated at 37°C in a 5% CO2 with 100% humidity. At given times, unattached cells were removed by rinsing the wells with warm (37°C) PBS. Attached cells were fixed in 4% PFA, stained with 0.5% crystal violet and quantified by reading OD at 595 nm.
This assay is based on the principle that a Rho and Rac-GTP-binding protein is linked to the 96-well plates (RhoA and Rac-1 GLISA from Cytoskeleton, Inc.). The active GTP-bound Rho or Rac-1 in the cell lysates binds to the wells, while the inactive GDP-bound Rho or Rac-1 is removed during the washing steps. The bound active RhoA or Rac-1 is detected with a RhoA or Rac-1 specific antibody and quantitated by absorbance. The degree of RhoA or Rac-1 activation is determined by comparing readings from lysates prepared from various treatments.
Invasion through the extracellular matrix (ECM), an important step in KSHV pathogenesis, was measured by two methods. 1) Innocyte cell invasion assay was used to quantitate the invasive cells and is based on the principle that invasive cells would degrade the laminin layer and will migrate through the membrane and attach to the underside of the membrane. These invasive cells are dislodged from the underside of the cell culture insert and stained with a fluorescent dye in a single step and fluorescence is determined using a fluorimeter (Ex485/Em520 nm). Briefly, the upper chambers of Transwells (Corning Costar) precoated with ECMatrix was allowed to wet by incubating with serum free EBM-2. After an hour of hydration, 5×104 cells (HMVEC-d, TIVE, TIVE-LTC or HMVEC-d cells treated with various conditioned media) were plated in the upper chambers. The lower chambers contained complete growth medium. The inserts were incubated for 24 h and the invading cells were quantitated by fluorimetry. 2) Chemicon cell invasion assay was performed to further confirm the invasion of cells upon various treatments and the assay is based on staining the invasive cells on the lower surface of the membrane by dipping inserts into the staining solution, washing, drying the inserts and counting the cells by photographing the membrane through the microscope as described in the manufacturer's instructions. Both assays were used to assess the role of COX-2 in regulating the invasive potential of HMVEC-d, TIVE and TIVE-LTC cells. Human fibrosarcoma (HT-1080) cells with high invasive potential were used as positive control.
The in vitro effects of COX-2 inhibition, serum withdrawl on TIVE and TIVE-LTC cell numbers, and viability were determined by traditional trypan blue staining (evaluation of cell membrane integrity) in quadruplicate. As trypan blue staining is not a sensitive method for quantitation, the number of viable cells with their metabolically active mitochondria (an index of cell proliferation) was also determined by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT)–based colorimetric assay (ATCC, Manassas, VA) as per the manufacturer's instructions. The MTT assay detects living but not dead cells and signal generated depends upon the degree of activation of these cells. Briefly, 0.5×105 HMVEC-d, TIVE, and TIVE-LTC cells were allowed to grow in the presence of complete growth medium (EGM-2) or in basal medium without growth factors and serum (EBM-2) or EGM-2 containing the indicated amount of COX-inhibitor or solvent control for 24 h, 48 h, 72 h and 96 h. 10 µl of MTT Reagent was added to all the cells at the indicated time of treatment and further incubated for 4 h (development of insoluble purple precipitate), solubilized in detergent and then read at 570 nm. The amount of MTT (yellow tetrazolium salt), which is converted to insoluble purple formazan crystals represents the number of viable cells and the degree of conversion was assessed by measuring the absorbance at a wavelength of 570 nm.
TIVE or TIVE-LTC cells were either untreated or treated with drugs (Indo or NS-398) or solvent control as described for cell number and viability assays. Harvested cells were diluted to contain ∼106 cells/ml and DNA distribution analysis was performed. Cells were fixed with 70% ethanol overnight and DNA was stained with propidium iodide at a final concentration of 50 µg/ml with RNaseA (100 U/ml) prior to flow cytometry analysis using a LSRII (BD Biosciences). Data were analyzed using ModFit Lt V3 software (Verity Software House).
Despite clinical and epidemiological differences, the classic, epidemic (acquired immunodeficiency syndrome-associated KS), endemic and post-transplantation associated KS lesions show a similar histopathology characterized by spindle shaped endothelial cells with latent KSHV infection expressing endothelial markers (CD31, CD34, CD36, and EN4), extensive neo-angiogenesis and inflammatory infiltration [21],[38],[39],[40]. We analyzed the skin and lymph node tissue sections of healthy subjects and KS+ patients obtained from ACSR for the presence of COX-2 with anti-COX-2 antibody. Normal healthy control tissue sections (Figure 1A, panel 1) and normal healthy lymph node sections (Figure 1A, panel 5) showed negligible expression of COX-2. In contrast, abundant COX-2 expression was detected in KS skin tissue (Figure 1A, panel 2; Figure S1) and KS lymph node section (Figure 1A, panel 6; Figure S1). Intense, patchy COX-2 expression was detected in KS lymph node sections, especially surrounding neovascular structures (Figure 1A, panels 6 and 7; Figure S1). KS skin tissue and lymph node sections showed distinct nuclear staining for KSHV latency associated LANA-1 (ORF73) protein (Figure 1A, panels 4 and 8; Figure S1). Specificity of COX-2 staining was confirmed by the non-reactivity of isotype control for COX-2 antibody (Figure 1A, panel 12). Cytoplasmic COX-2 staining was observed in KS skin tissue sections, which were also observed in cells showing spindle phenotype (Figure 1A, panels 3 and 11). Strong COX-2 staining was also observed in the lining of neovascular structures in KS patient lymph nodes (Figure 1A, panels 6, 7, 9 and 10). We next assessed the phenotype of spindle cells for endothelial marker CD31 as well as COX-2. CD31 (red) was detected in spindle cells of KS lesions (Figure 1B, panels 1–6) and in KS patient lymph nodes (Figure 1B, panels 7–15). KS tissue sections (skin and lymph nodes) showed many strong CD31-COX-2 double positive cells (Figure 1B, panels 1–15). Many CD31 positive cells in KS skin tissue (Figure 1B, panels 1–6) and cells lining the neovascular structures in KS lymph node sections (Figure 1B, panels 7–15) displayed strong staining for COX-2. COX-2 (green) staining was not just limited to spindle cells present in KS tissues but also to other smaller cells whose morphological appearances suggested that they are most likely macrophages and/or lymphocytes (Figure 1B, panel 3). Strong COX-2 and CD31 co-staining was also observed in the lining of KS lymph node neovascular structures (Figure 1B, panels 7–15).
To define the prevalence of COX-2 up-regulation in KS, COX-2 staining was performed in KS-TMAs as described in Material and Methods. Varying level of COX-2 staining was observed in a variety of tissues from KS patients (Figure S2). Several sections from skin showed negligible staining (Figure 1C, panels 1 and 2) whereas the majority of them showed strong patches (Figure 1C, panels 4 and 5) of COX-2 staining. Very low COX-2 staining was observed in some sections from mouth (Figure 1C, panel 3). Strong COX-2 staining was observed in eye orbit (panel 6), tonsil (panel 7) and mouth (panel 8) and small bowel (panels 9 and 10) sections. Specificity of COX-2 staining was also confirmed by the non-reactivity of isotype control for COX-2 antibody even when tested on the colon cancer tissue (Figure 1C, panel 12) as compared to staining with COX-2 antibody (Figure 1C, panel 11). To further demonstrate that all TMA sections did not show strong or some nonspecific COX-2 staining, we provide some examples of sections with negligible, low and strong staining for COX-2 (Figure S2). A higher number of sections showed immunoreactivity for COX-2 but the levels of staining varied among the sections (Figure S2) suggesting a potential connection between COX-2 and KS. Magnified view of various KS sections described in Figures 1A and 1C are given in Figure S1 which clearly demonstrate COX-2 distribution in KS tumor cells with characteristic spindle phenotype. These results demonstrate that COX-2 is an abundant factor in the majority of KS lesions with a few exceptions (Figure 1C, panels 1–3, Table S1), and thereby suggesting that COX-2 might be playing a key role in KSHV pathogenesis.
Our previous study demonstrated that de novo infection of endothelial cells with 10 DNA copies/ cell of KSHV up-regulated COX-2 during early time points of infection which was maintained at 2–3 fold even at 72 h PI. Here, we extended this observation using higher KSHV DNA copies per cell (30) for infection of HMVEC-d cells and observed the cells until 5d PI. KSHV ORF73 gene expression as assessesd by qRT-PCR (Figure 1D) as well as by real-time RT-PCR with ORF73 gene specific primers and Taqman probes (data not shown) confirmed the successful infection of these cells. Compared to uninfected cells at all the respective time points, KSHV infection induced about 49, 44, 24, 15, 17, 15, 11, and 12- fold COX-2 expression at 2 h, 4 h, 8 h, 1d, 2d, 3d, 4d and 5d PI, respectively (Figure 1D). In addition, we also observed the concomitant induction of m-PGES-1, an enzyme converting PGH2 to PGE2, with about 13, 12.5, 11.3, 13, 10,11, 9, and 7-fold induction at 2 h, 4 h, 8 h, 1d, 2d, 3d, 4d and 5d PI, respectively (Figure 1D). To determine the percentage of cells expressing latent genes and undergoing spontaneous KSHV lytic replication, IFA was carried out using antibodies against ORF73 (latency marker) and ORF59 (processivity factor and a marker of lytic replication) proteins (Figure S3). Detection of a few lytic cycle positive cells at early time points (2h) of KSHV infection could be due to transient lytic burst in primary endothelial cells [41]. At 5 day PI, about 70–80% of cells stained positive for nuclear punctate pattern of ORF73 (Figure S3, panels 4 and 6) and about 9–12% stained positive for lytic ORF59 at early time (2h) (Figure S3, panels 10 and 12), whereas 15–17% cells displayed lytic cycle activation at later (5d) time point of infection (Figure S3, panels 16 and 18). We also detected a low level (8–10%) of lytic induction at 4d post KSHV infection (Figure S3, panels 13 and 15). The percentage of cells expressing ORF59 at 5d PI was significantly higher than at 4d PI and was reproducible. The spike of lytic burst at 5d PI in HMVEC-d cells could also be due to continued presence of pro-IC rich microenvirnment created by KSHV infection. These infected cells expressed high copy numbers for early lytic cycle switch protein ORF 50 at 2h post KSHV infection (data not shown). We emphasize that all analyses are based on IFA for lytic cycle ORF59 protein and hence, percentage of cells will not be identical for all endothelial cells (HUVEC cells) and might also depend on the number of KSHV DNA copies per cell used for infection.
TIVE-LTC cells are endothelial cells in culture with tightly latent KSHV gene expression supporting long-term episomal maintenance which is similar to viral-gene expression in the majority of KS lesion spindle cells [42]. KSHV-positive TIVE-LTC cells expressed very high levels of ORF73 gene expression. Compared to uninfected TIVE cells, TIVE-LTC cells showed increased expression of COX-2 (5-fold), m-PGES-1 (4-fold) and VEGF-A (8-fold) (Figure 2A). Compared to uninfected TIVE cells, KSHV-positive TIVE-LTC cells showed (4-fold) higher levels of COX-2 protein (Figure 2B). Punctate nuclear staining of ORF73 was observed in 50–60% of TIVE-LTC cells (Figure 2C; Panels 1 and 3). Distinct perinuclear COX-2 staining was observed in a majority of the TIVE-LTC cells (Figure 2C; Panels 2 and 5). Besides ORF73 positive cells, the majority of neighboring uninfected cells located in close proximity to the infected cells were also positive for COX-2 (Figure 2C; Panels 2, 3, 5 and 6). Overall, 70–80% of TIVE-LTC cells were positive for COX-2. Detection of COX-2 in uninfected cells could be due to paracrine COX-2 stimulation by the various cytokines and growth factors induced by KSHV. Similarly, COX-2 expressing uninfected cells were also seen in KSHV-infected HMVEC-d monolayer but were distinctly less in number and were in close proximity to the infected cells (data not shown). Similarly stained TIVE cells showed very faint cytoplasmic basal staining for COX-2 in a few cells (Figure 2C; panels 8 and 11) and no staining for viral latent protein ORF73 (Figure 2C; panels 7 and 10). COX-2 staining in TIVE cells (Figure 2C; panels 8, 9, 11 and 12) was not comparable to the strong perinuclear COX-2 staining seen in TIVE-LTC cells (Figure 2C; panels 2, 3, 5 and 6). Compared to uninfected TIVE cells, significantly higher levels of PGE2 (pg/ml) were detected in the supernatants of TIVE-LTC cells (Figure 2D).
Since COX-2 inhibition down-regulated ORF73 gene expression during de novo KSHV infection [26], we next determined the effect of NS-398 and Indo treatment on ORF73 gene expression in TIVE-LTC cells. First, we determined the concentrations of COX inhibitors affecting PGE2 secretion. TIVE-LTC cells pretreated with nontoxic doses of either Indo (500 µM or 250 µM) or NS-398 (50 µM or 75 µM) at 37°C for 1h did not completely inhibit PGE2 secretion (Figure S4). In contrast, by increasing the incubation period with these inhibitors to 8 h and 24 h, we observed a significant reduction (∼80%) in PGE2 secretion (Figure S4). This requirement for a higher dose of inhibitors to block COX-2 function and PGE2 secretion could be due to the continuous loop of COX activation leading to the maintenance of a constant level of PGE2 in latently infected cells.
NS-398 and Indo treatment of TIVE-LTC cells for 24 h down-regulated viral latent (ORF73) gene expression by 48% and 57%, respectively (Figure 2E). Significant detection of COX-2 in KS lesions (Figure 1A, B and C), long term KSHV infected endothelial cells (Figure 2) and in de novo infection of endothelial and fibroblast cells [26], as well as modulation of viral gene expression by COX inhibition (Figures 1 and 2), strongly indicated a role for COX-2/PGE2 in KSHV pathogenesis.
Pre-treatment of endothelial cells with chemical nonsteroidal anti-inflammatory drugs (N SAID) like Indo or COX-2 selective inhibitor (COXIB) NS-398 prior to KSHV infection abrogated the secretion of PGE2 [26]. These conventional NSAIDs have been shown to cause serious and significant complications [43]. Though the selective COX-2 inhibitors cause only occasional deleterious effects, they have also been shown to exhibit some COX-2 independent effects such as up-regulation of death receptor 5 (DR5) expression, inhibition of survival signal pathways, and augmentation of apoptosis [43],[44]. To determine the specificity of COX-2 involvement in KSHV pathogenesis and to avoid COX independent effects of chemical inhibitors, we used a COX-2 silencing method.
293T cells were co-transfected with COX-2 expression plasmid and si-COX-2-1, si-COX-2-2 and si-Control (si-C) plasmids. Transfection with pcDNA was used as a control (Figure 3A, lane 2). Western blots for COX-2 confirmed the silencing of COX-2 by si-COX-2 (Figure 3A, lanes 3 and 4) compared to si-C (Figure 3A, lane 1), and tubulin was utilized as a loading control. Co-transfection of 1 µg COX-2 expression plasmid and si-C showed 10-fold induction of COX-2 protein (Figure 3A, lane1). Transfection with 1 µg of either si-COX-2-1 or si-COX-2-2 along with COX-2 expression plasmid showed 85% and 90% reduction in COX-2 protein levels, respectively (Figure 3A, lanes 3 and 4). These results clearly implied that COX-2 silencing by these sequences was effective. We used both the plasmids to generate si-COX-2 lentiviruses which were used throughout this study. The lentivirus preparations were quantified for their titer and 30 DNA copies/ cell of all three lentiviruses [si-C, si-COX-2-1 and si-COX-2-2] were used for transduction in HMVEC-d cells. We observed very high transduction efficiency (>90% of HMVEC-d cells expressing GFP) by fluorescence microscopy.
Effect of COX-2 silencing in HMVEC-d cells was determined by infecting serum starved (8 h) si-C, si-COX-2-1 or si-COX-2-2 transduced cells for 2 h, 4 h, 8 h, and 24 h. These cells were treated with TNF-α for 30′ to serve as a positive control for COX-2 induction. Compared to uninfected si-C cells, KSHV infected si-C-HMVEC-d cells showed high COX-2 gene expression (Figure 3B). In contrast, KSHV infected si-COX-2-1 or si-COX-2-2 -HMVEC-d cells showed significantly reduced COX-2 expression (Figure 3B). Overall, si-COX-2-1 or si-COX-2-2 -HMVEC-d cells showed 82% and 93% reduction in COX-2 expression, respectively (Figure 3B).
We next assessed the functional consequences of COX-2 silencing by quantifying the secreted PGE2 levels in the supernatant of KSHV infected lentivirus transduced HMVEC-d cells (Figure 3B). In si-C-HMVEC-d cells, PGE2 secretion levels dramatically increased upon KSHV infection. Though there was induction in PGE2 levels in si-COX-2 transduced cells upon KSHV infection, this induction was lower than in si-C-HMVEC-d cells (Figure 3B). Similarly, TNF-α induced PGE2 secretion was reduced drastically in si-COX-2- HMVEC-d cells (Figure 3B) suggesting that COX-2 silencing could effectively abrogate KSHV infection induced PGE2 secretion in endothelial cells. COX-2 silencing did not change COX-1 expression (data not shown) further validating the specificity of the knock-down procedure.
We also assessed the consequence of COX-2 silencing on KSHV latent gene expression (Figure 3C). After 24 h KSHV infection, we observed about 61% and 59% reduction in ORF73 gene expression in si-COX-2-1 and si-COX-2-2 -HMVEC-d cells, respectively (Figure 3C). These results supported our earlier findings in HFF cells with chemical inhibitors [26] and demonstrated that COX-2 silencing effectively reduced KSHV latent gene expression.
In our earlier studies of oligonucleotide array analysis of KSHV-infected HMVEC-d and HFF cells at 2 and 4 h PI, we observed the reprogramming of host transcriptional machinery regulating a variety of cellular processes, including apoptosis, cell cycle regulation, signaling, inflammatory response and angiogenesis [25]. Since COX-2 has also been shown to regulate the majority of these factors, we next analyzed the role of KSHV-induced COX-2 in the modulation of these factors. Conditioned media (no serum) collected from KSHV-infected HMVEC-d cells at 2 h, 4 h, 8 h, 24 h, 4 days and 5 days PI were used to study the cytokine profile (Figure 4). Induction of cytokines was compared to the released cytokine levels in the uninfected cell supernatant at respective time points. Compared to uninfected HMVEC-d cells, KSHV infection triggered an appreciable (1.5-2) fold induction in the secretion of pro-ICs, such as growth regulated oncogene (GRO), GROα, IL-1α, IL-1β, ILs- (2,3, 6, 7, and 12-p40), TNF-α, TNF-β and IFN-γ at 4 h PI. 1.5 to 2-fold induction in these cytokine levels further up-regulated to 3 –3.5 -fold by 8 h, decreased to 2–2.5-fold by 24 h, 1.5–2 -fold at the 4d, and enhanced dramatically to 4–4.5 -fold at 5d PI (Figure 4, Table S3). Since, we did not observe an increase in these cytokines released at 2h PI, we used 4, 8 and 24 h PI time points throughout this study. Among all these cytokines, IL-8 levels did not increase at 5d PI. The drastic increase in the cytokine levels observed at 5d PI might be due to the spontaneous induction of KSHV lytic cycle replication observed in about 15–17% of these infected cells (Figure S3).
Compared to uninfected cells, KSHV infection induced the secretion of chemotactic cytokines (chemokines) that mediate leukocyte recruitment to sites of inflammation, fibrosis, and malignancy such as RANTES, macrophage chemoattractant protein-2 (MCP-2), MCP-3, thymus and activation-regulated chemokine (TARC), macrophage inflammatory protein (MIP-1Δ), macrophage derived chemokine (MDC), monokine induced by IFN-Gamma (MIG), epithelial neutrophil-activating peptide (ENA-78), and inflammatory cytokine 309 (I-309) (Figure 4B, Table S3). Among these chemokines, MCP-1 was the only one that was not up-regulated at all time points tested as the uninfected cells always showed some level of MCP-1 secretion in the culture supernatants.
KSHV infection stimulated the secretion of growth factors and angiogenic factors such as insulin-like growth factor-1 (IGF-1), platelet derived growth factor-BB (PDGF-BB), macrophage colony stimulating factor (M-CSF), granulocyte colony-stimulating factor (G-CSF), GM-CSF, angiogenin (Ang), oncostatin-M (Onco-M), TPO (thrombopoietin), VEGF, stromal cell-derived factor-1 (SDF-1), SCF (stem cell factor), TGF-β1 and leptin (Figure 4C, Table S3). EGF (epidermal growth factor) was very highly up-regulated at the early time points of KSHV infection with about ∼4 fold induction at 2 h PI, which decreased to 1.5-fold by 4 h PI, increased at 8h to ∼3- fold before decreasing to 2- fold by 24h and 4 days, and finally increased at 5 days PI (Figure 4C, Table S3). Endogenous levels of EGF were high as the supernatants obtained from uninfected cells also showed higher levels of EGF secreted (data not shown).
Compared to uninfected HMVEC-d cells, KSHV infection enhanced the secretion of anti-inflammatory cytokines, such as ILs (−4, −5,−10, −13 and −15) with 1.5 to 2.5-fold at 4 and 8h PI which decreased to 2-fold at 4d and was up-regulated at 5d PI (Figure 4D, Table S3). IL-10 levels were higher than all other anti-ICs with ∼4 fold from 8h PI to 5d PI (Figure 4D, Table S3).
To evaluate the specificity of KSHV infection induced cytokine secretion, virus was pre-incubated with 100 µg of heparin/ml which has been shown to block about 80% of virus binding and entry into the various target cells [26],[35]. Conditioned medium from serum starved HMVEC-d cells infected for 96h (Figure S5A, panel 1) showed a significant increase in the levels of various cytokines and inflammatory molecules, which were greatly reduced by pretreatment of the virus with heparin (Figure S5A, panel 2). Representative data from one time point of infection shows that there was complete inhibition of SDF-1, SCF, TGF-β and TARC with 60–70% inhibition of GM-CSF, GRO, GRO-α, ILs (−2,−3, −4,−5, and −10), MDC, MIG, MIP-1Δ, RANTES, IGF-1, angiogenin, oncostatin-M, and TPO and 30–40% inhibition of VEGF, PDGF-BB, IL-7, IL-1β, and IL-8. There was no detectable inhibition in secretion of MCP-1 and EGF (Figure S5A, panel 2) which might be due to their high endogenous levels of secretion even in the uninfected cell culture supernatant. These results demonstrated that the vast majority of the observed cytokine induction (Figure 4, A to D) was due to KSHV infection and not due to LPS or contaminating host cell factors in the virus preparations.
To understand the role of KSHV induced COX-2 in cytokine secretion detected in infected HMVEC-d cells, we used COX-2 inhibitors in conjunction with COX-2 silencing methods and examined cytokine gene expression. We prepared cDNA from serum starved (8h) cells either infected (4 h, 8 h, 24 h) or pretreated with either NS-398 (50 µM) or Indo (500 µM) (data not shown), and then infected with KSHV for 4 h, 8 h, and 24 h. Similarly, cDNA was prepared from serum starved (8 h) cells transduced with si-C, si-COX-2-1 or si-COX-2-2 and then uninfected or infected with KSHV for 4 h, 8 h and 24 h. cDNA was used for q-RT-PCR to quantitate the fold expression of specific cytokines, selected based upon our observations from Figure 4, such as IL-8, VEGF-A, GRO, VEGF-C, IL-1β, GM-CSF, RANTES, and SDF-1, normalized to the expression of endogenous HPRT and tubulin genes. Analyses of the results showed time dependent patterns of inhibition of gene expression by both methods of COX-2 inhibition (Figures 5 and S6).
In pattern one, KSHV infection induced nearly 5, 5.3, and 6.1-fold expression of IL-8 gene at 4 h, 8 h and 24 h PI, respectively compared to uninfected HMVEC-d cells, which were unaffected by pretreatment of cells with NS-398 for 1 h or by si-COX-2-2. These results suggested that IL-8 gene expression was not directly regulated by KSHV induced COX-2 (Figure 5A). In pattern two, in contrast, both COX-2 inhibitor treatment and COX-2 knockdown reduced KSHV induced VEGF-A and VEGF-C gene expression significantly at both early and late time points (Figures 5B and S6A), thus suggesting a role for KSHV induced COX-2 in the regulation of VEGF-A and -C at all time points of infection. In pattern three, a significant reduction of IL-1β and GM-CSF gene expression by COX-2 inhibitor treatment and COX-2 knockdown was observed at 4 and 8 h PI, and was moderately less at 24 h PI (Figures 5C and S6B). These results suggested a role for COX-2 in KSHV induced IL-1β gene expression at an early time point of infection. In pattern four, the expression of KSHV induced GRO, RANTES and SDF-1 genes were not significantly affected by NS-398 and si-COX-2 at 4 h PI, while significant reductions were observed at 8 and 24 h PI (Figures S6C, S6D, and 5D). This suggested that COX-2 plays a role in the regulation of these genes at later time points.
Overall, inhibition of cytokine gene expression by COX-2 inhibitor treatment and COX-2 knockdown were comparable. Both methods inhibited gene expression of VEGF-A, VEGF-C (angiogenic molecules), GRO (cytokine with inflammatory and growth-regulatory properties), RANTES (cytokine regulating T cell response) and SDF-1(a ligand for the chemokine receptor CXCR4) even at 24 h PI. Inhibition of IL-1β gene expression at early time points is not due to the inactivation of COX-2 inhibitor or less inactivation of COX-2 by the silencing method but demonstrates the specificity of the observed results. Absence of 100% reduction in the expression of the examined cytokine genes by both methods could be due to the inability to inactivate or deplete COX-2 completely, paracrine effects of the released additional factors and/or additional factors besides COX-2 in the regulation of these genes in KSHV infected cells.
To examine the role of COX-2 in the secretion of various cytokines during KSHV infection, we analyzed the cytokines from COX-2 inhibited cells. Data obtained from COX-2 inhibitor pretreatment followed by infection (4 h, 8 h, 24 h) or cells silenced for COX-2 and then infected for different time points (4 h, 8 h, 24 h) is presented as fold reduction compared to signals obtained from untreated KSHV infected cells, and KSHV infected si-C-HMVEC-d at respective time points (Table S4). Results shown in Table S4 can be divided into three groups. Group 1 includes cytokines inhibited by both kinds of COX-2 inhibition (chemical as well as silencing). Group 2 includes the cytokines inhibited by chemical inhibitor (NS-398) treatment alone but not reduced by COX-2 knock-down. Group 3 includes the cytokines up-regulated by COX-2 inhibition.
Group 1 cytokines are specifically dependent upon COX-2 which includes: pro-ICs like IL-1 (α and β), ILs (−2, −3, −p40 and −16) TNFα, IFNγ, LIGHT; chemokines including RANTES, MCP-2, MCP-3, TARC, MIP-1Δ, ENA-78, I-309, MIF, GCP-2, MIP-3-α, Eotaxin, Eotaxin-2, Eotaxin-3, IP-10, NAP-2, CK-β8-1; growth and angiogenic factors including PDGF-BB, MCSF, G-CSF, GMCSF, angiogenin, oncostatin M, thrombopoeitin, VEGF, SDF-1, SCF, TGF-β1, Leptin, FGFs (−4, −6, −7, −9), Flt3-ligand, Fractalkine, IGFBPs (−2, −3, −4), BDNF, PIGF, HGF (hepatocyte growth factor), Osteoprotegerin, NT-3, NT-4; and anti-inflammatory cytokines like IL-4, IL-13, and IL-15. In all these cytokines reported, although the fold reduction between the inhibitor treatment and COX-2 silencing were not identical, they were comparable. Overall, a profound reduction was observed in the levels of ILs (1β, −2, −3,−4, −13, −15 −16, −p40), IFNγ, MCPs (−2, −3), TARC, MIP-1Δ, ENA-78, I-309, GCP-2, Eotaxin, Eotaxin-3, PDGF-BB, G-CSF, angiogenin, Oncostatin M, TPO, VEGF, SDF-1, SCF, TGF-β1, Leptin, FGFs (−4, −6, −7), Flt3-ligand, Fractalkine, IGFBP-4, BDNF, PIGF, Osteoprotegerin, NTs (−3, −4) (Table S4). Group 2 cytokines including pro-ICs (GRO, GRO-α, IL-6, IL-7, IL-8, TNFβ), chemokines (MCP-1, MDC, MIG), growth and angiogenic factors (IGF-1 and IGFBP-1), as well as anti-inflammatory cytokines (IL-5) were only inhibited by chemical inhibitor (NS-398) treatment and not by COX-2 knock-down (Table S4). These results indicate some COX-2 independent effects of chemical inhibitors. Growth factor EGF, anti-inflammatory cytokine IL-10, regulators of MMP activity like TIMP-1 and TIMP-2 were up-regulated by COX-2 inhibition (group 3) (Table S4).
Over-expression of COX-2 is known to correlate with the aggressive and invasive potential of tumor cells by several mechanisms [32]. One of the mechanisms modulated by COX-2 during carcinogenesis is angiogenesis, presumably through increased production of the most potent and extensively studied pro-angiogenic factor, VEGF [45]. Here, we examined the role of KSHV induced COX-2 in the secretion of two angiogenic factors, VEGF-A and VEGF-C. VEGF-A is a dimeric glycoprotein with structural homology to PDGF, and is known to have several variants. We used an anti-human VEGF-A mouse monoclonal antibody that detects all isoforms, particularly the most commonly expressed 189, 165 and 121 amino acid splice variants. When examined by IFA, uninfected HMVEC-d cells showed a very low level of expression of COX-2 and VEGF-A (Figure 6A, panels a and b). In contrast, at 24 h PI, several cells were positive for cytoplasmic staining of COX-2 and VEGF-A (Figure 6A, panels d and e). Co-expression of VEGF-A and COX-2 was observed in 40–50% of infected cells (Figure 6A, panel f), which demonstrated that KSHV infection of primary endothelial cells up-regulated expression of both VEGF-A and COX-2 simultaneously. Uninfected cells in close proximity to the infected cells also showed expression of COX-2 and VEGF-A (data not shown).
Basal levels of VEGF-A (Figure 6B and 6C) and VEGF-C (Figure 6D and 6E) secretion in uninfected non-transduced and si-C, si-COX-2-1 or si-COX-2-2 transduced HMVEC-d were similar. Pretreatment of HMVEC-d cells with either NS-398 or Indo did not have any non-specific inhibition on the basal levels of VEGF-C secretion (Figure 6D). KSHV infection of serum starved HMVEC-d cells induced 1029, 1420, 2211, and 2371 pg/ml of VEGF-C secretion at 2 h, 4 h, 8 h, and 24 h, respectively. (Figure 6D), while 1239, 1457, 1897, and 2200 pg/ml of VEGF-C secretion in serum starved si-C-HMVEC-d cells was observed at 2 h, 4 h, 8 h, and 24 h, respectively (Figure 6E). Treatment of cells with either NS-398 or Indo prior to KSHV infection reduced VEGF-A and VEGF-C secretion at all the time points tested (Figures 6B and 6D). Analyses showed that NS-398 pretreatment inhibited VEGF-A and VEGF-C secretion which was higher than inhibition by Indo treatment (Figures 6B and 6D). COX-2 silencing also inhibited VEGF-A and VEGF-C secretion but to a lesser extent than chemical inhibitor treatment (Figures 6B-6E). The incomplete inhibition of VEGF secretion by COX-2 chemical inhibitors suggested that besides COX-2, other factors induced by KSHV infection might also be playing a role in VEGF release from infected cells.
To evaluate the role of COX-2 in regulating angiogenesis in latently infected cells, we checked the expression of VEGF-A and VEGF-C (Figure 6F) in TIVE-LTC cells untreated or treated with either 500 µM Indo or 75 µM NS-398 for 24 h. Compared to TIVE cells, TIVE-LTC cells showed 9.8- and 2.4 fold higher VEGF-A and -C gene expression, respectively (Figure 6F). Pretreatment of TIVE-LTC cells with either Indo. or NS-398 significantly inhibited the expression of VEGF-A, clearly demonstrating the role of COX-2 in regulating VEGF-A gene expression in latently infected cells (Figure 6F). Similar to VEGF-A gene expression, TIVE-LTC cells secreted appreciably high levels of VEGF-A (54 pg/ml) as compared to TIVE cells and this secretion was effectively reduced upon treatment with COX inhibitors (Figure 6G). VEGF-C secretion from TIVE (1800 pg/ml) and TIVE-LTC (2200 pg/ml) cells was comparable.
One important biological effect of PGE2, VEGF, and bFGF secretion is the induction of endothelial cell tube formation. Many chemokines are also recognized as important mediators of endothelial cell migration and tubular organization. For a comprehensive understanding of the role of KSHV induced COX-2 in VEGF-related angiogenesis of infected endothelial cells, conditioned media obtained either from serum starved (8 h) HMVEC-d cells infected with KSHV (24 h), or cells pretreated with COX inhibitors (Indo or NS-398) and then uninfected or infected with KSHV (24h) were tested for their ability to induce tube formation. Representative pictures are shown in Figure S7A. HMVEC-d cells spontaneously organized into a primitive vascular network even in the presence of EBM-2 alone (no serum) (Figure S7A, panel a). This network was still observed with supernatants from cells treated for 24h with COX-2 inhibitors which suggested that inhibitor treatment did not have any adverse effect on the secreted factors in the uninfected cells (Figure S7A, panels b and c). In contrast, highly organized enhanced capillary tube formations with strong branching networks were observed in endothelial cells incubated with conditioned medium from cells infected with KSHV for 24h (Figure S7A, panel d). When supernatants from cells pretreated with COX inhibitors and then infected with KSHV (24h) were used, we observed significant inhibition of tube formation (Figure S7A, panels e and f). Higher inhibition was observed with NS-398 pretreated infected cell supernatant with complete impairment and disintegration of tube formation (Figure S7A, panel f) which is in contrast to the strong, well communicating tubes observed with media after 24h KSHV infection alone (Figure S7A, panel d). These results suggested a direct role for KSHV infection induced COX-2 in the induction of factors mediating endothelial cell capillary tube formation.
When suramin, which possesses anti-angiogenic properties, was used as a negative control, cells failed to adhere to each other and remained either as single cells or clumps of cells with no tube formation (Figure S7A, panel g). In contrast, conditioned medium from cells cultured in the presence of medium with growth factors (EGM-2) showed intact and organized endothelial lattice formation comparable to the network formed in the KSHV infected culture supernatant (Figure S7A, panels h and d). This suggested a role for GFs in tube formation. Supernatant obtained from solvent treated and then KSHV infected cells showed a strong intact tube network that was comparable to the one observed in the presence of cells infected with KSHV for 24h (Figure S7A, panels d and i), thus ruling out the possibility of non-specific inhibition by solvent control on tube formation of endothelial cells. Similar results for tube formation were observed with HUVEC cells (data not shown).
The angiogenic index can be measured either by taking the sum of all the nodes (connection between various tubes on the matrigel) between the tubes formed on the matrigel or the length and width of tube formation between nodes. Results represented in Figure 7A show branch points/field, a measure of connections among cells. Supernatants from KSHV infected cells induced the number of branch points by about 2.5-fold as compared to EBM-2 only. Diffused nodes with incomplete branches were not counted. Branch points per field in the supernatants obtained from KSHV infected cells versus cells in the presence of EGM-2 were similar. Compared to supernatants from KSHV infected cells, supernatants from Indo or NS-398 pretreated and infected cells inhibited node formation by 45% and 80%, respectively (Figure 7A). Supernatants from cells pretreated with the solvent control before KSHV infection for 24h did not show any inhibition in node formation and was comparable to the branch points formed in the presence of EGM-2 or KSHV infection (Figure 7A).
A similar experiment was done for si-C, si-COX-2-1 and si-COX-2-2 transduced HMVEC-d (Figure S7B) and HUVEC cells. Transduction with si-C, si-COX-2-1 or si-COX-2-2 did not have any non-specific effects on the secretion of factors involved in angiogenesis (Figure S7B, panels a, b, and c). Tube formation in medium obtained from si-C transduced cells (Figure S7B, panel a) was comparable to the results from non-transduced uninfected cells (Figure S7A, panel a). Tube formation in the presence of medium obtained from infected si-COX-2-1 (Figure S7B, panels e and f) or si-COX-2-2 transduced cells (Figure S7B, panels h and i) was diminished, and nodes were diffuse compared to supernatants from infected si-C-HMVEC-d cells (Figure S7B, panels d and g). Consistent with the COX-2 inhibitor data, silencing by either si-COX-2-1 or si-COX-2-2 in endothelial cells prior to KSHV infection also reduced the ability of supernatants to induce node formation by approximately 22% and 24%, respectively, compared to si-C-HMVEC-d cells infected for 24h (Figure 7B). This inhibition was less pronounced when compared to the huge reduction (80%) observed with the supernatants in the presence of NS-398. Similar results for tube formation were observed with HUVEC cells (data not shown) implying that COX-2 plays an important role in regulating the angiogenic phenotype of KSHV infected endothelial cells.
As TIVE-LTC cells showed high VEGF-A gene expression and secretion, to further assess the biological role of secreted angiogenic factors, an endothelial cell tube formation assay was performed (Figure S8). HMVEC-d cells were seeded on a Matrigel-coated 96-well plate with conditioned medium obtained from 24h serum starved TIVE (Figure S8, panels 1–4), or TIVE-LTC cells (Figure S8, panels 5–8). After 16h of incubation with conditioned media, plates were examined for capillary-like tubular structures as described before. HMVEC-d cells spontaneously organized into a primitive vascular network even in the presence of medium obtained from TIVE cells (Figure S8, panels 1–4). Highly organized and intricate capillary tube formations with strong branching networks were observed in cells incubated with conditioned medium obtained from TIVE-LTC cells (Figure S8, panels 5–8). This network was also observed in the presence of solvent treated TIVE-LTC cells (Figure S8, panels 9–12), ruling out the possibility of non-specific inhibition by solvent on tube formation. In contrast, with supernatants from TIVE-LTC cells pretreated with COX inhibitors, we observed significant inhibition of tube formation (Figure S8, panels 13–20). Higher inhibition was observed with NS-398 treated TIVE-LTC cell supernatant which had complete impairment of tube formation (Figure S8, panels 17–20).
Quantitatively, supernatant obtained from TIVE-LTC cells induced roughly 3-fold more branch points/field than TIVE cells (Figure 7C). Pretreatment of TIVE-LTC cells with either Indo or NS-398 inhibited the secretion of angiogenic factors and thereby reduced node formation by 49% and 85%, respectively (Figure 7C). Together, these results suggest that KSHV induced COX-2 not only mediates the expression of growth and angiogenic factors from latently infected endothelial cells but also their functional properties, such as angiogenesis related tube formation.
MMPs belong to a family of secreted or membrane-associated zinc endopeptidases capable of digesting connective tissue ECM proteins as well as basement membrane constituents [46], and have been shown to play a critical role in orchestrating cell signaling, homeostasis of the extracellular environment via proteolysing their specific substrates [47], cell-cell and cell-matrix interactions, maintaining tight junctions, and thereby contributing to the malignant phenotypes of cancers, including cell invasion, metastasis, angiogenesis and inflammatory infiltration. So far, 23 MMPs have been identified in humans, and based largely on their substrate specificity, these are divided into collagenase like MMPs (−1, −8, −13), gelatinase like MMPs (−2, −9), stromelysin or proteoglycanase like MMPs (−3, −7, −10, −11), elastase (-12), membrane type-MMPs (1–4), and unclassified MMPs. Among them, MMP-2 and MMP-9 are known to be strongly correlated with the metastatic potential of cancer cells and in particular are prognostic factors in many solid tumors [46],[48]. MMP (−1,−2,−7,−9,−13) and MT-MMP-14 expression has been shown by immunohistochemistry in AIDS-related and classic cutaneous KS lesions at various histologic stages [49] implicating them in KS tumorigenesis and invasion. Here, we assessed the role of KSHV induced COX-2 on MMPs in uninfected and KSHV infected HMVEC-d cells.
Conditioned media collected from serum starved (8h) uninfected or KSHV infected HMVEC-d cells were used to probe for the presence of various MMPs and TIMPs using MMP antibody arrays (Figure 8A). Conditioned media from the uninfected cells had appreciable amounts of MMP-1 and -10 (Figure 8B). KSHV infection up-regulated the secretion of MMPs and TIMPs (Figures 8B and 8C) at all of the time points tested. Except for MMP-3, secretion of MMPs (-1, -2, -8, -9, -10 and -13) was enhanced in a time dependent manner with higher levels of secretion for MMP-9 and -2 (Figure 8C).
TIMPs are endogenous inhibitors of MMPs and the TIMP family consists of four distinct members, TIMP-1, -2, -3, and -4. Among these, TIMP-2 expression is constitutive and widely expressed throughout the body but TIMP-1, -3, -4 expression is inducible and often exhibits tissue specificity [46]. The balance between MMPs/TIMPs regulates ECM turnover, regulates tumor invasion and metastasis, wound healing and tissue remodeling during normal development and pathogenesis. The conditioned media from uninfected cells showed appreciable amounts of TIMPs 1 and 2 (Figures 8B and 8C). TIMP (−1, −2 and −4) secretion increased with the time post- KSHV infection (Figure 8C).
Effect of COX-2 inhibition was tested for a few select MMPs. MMP (−1, −2, −9 and −10) gene expression was induced during KSHV infection of HMVEC-d cells (Figures 8D-8G). NS-398 pretreatment reduced the expression of all MMPs tested (Figures 8D-8G), with the most significant inhibition of MMP-2 and MMP-9 (Figures 8E and 8F), suggesting that KSHV induced COX-2 plays a decisive role in controlling expression of KSHV infection induced proteases. Pretreatment of HMVEC-d cells with NS-398 slightly induced the expression of TIMP-1 and TIMP-2 by 1.4- and 1.3- fold, respectively as compared to 24h PI (data not shown). This data also supported the cytokine antibody array data (Table S4), where pretreatment of cells by NS-398 up-regulated the release of TIMP-2 by 2.1 fold at 24 h PI.
To evaluate the role of COX-2 in regulating angiogenesis and invasion, we checked the expression of MMP-9 and MMP-2 (Figure 8H) in TIVE-LTC cells treated with either 500 µM Indo or 75 µM NS-398 for 24 h. Compared to TIVE cells, TIVE-LTC cells showed nearly 4.5 and 5.7-fold higher MMP-2 and MMP-9 gene expression, respectively (Figure 8H). Pretreatment of TIVE-LTC cells with either inhibitor for 24h significantly inhibited the expression of MMP-9 and MMP-2, clearly demonstrating the role of COX-2 in gene expression during KSHV latency (Figure 8H).
Since the antibody array data presented in 8A-8C measured the total MMP pool (sum of inactive and active) secreted, we next used an MMP-9 ELISA to differentiate between the levels of the active form of the enzyme from the total released MMP-9. Compared to uninfected cells, about 2, 2.3, and 3.8-fold induction in the release of total MMP-9 was observed at 4h, 8h, and 24h PI (Figure 9A) which was consistent with the antibody array data (Figures 8B and 8C). NS-398 treatment prior to infection inhibited total as well as active-MMP-9 secretion implicating the role of KSHV induced COX-2 in regulating MMP-9 (Figure 9A). Similar to chemical blocking, compared to si-C HMVEC-d cells, COX-2 depletion by si-COX-2 (2) significantly decreased, total as well as active, MMP-9 secretion (Figure 9B). Compared to TIVE cells, about 2.8-fold and 2.1-fold induction in the release of total MMP-9 and active-MMP-9 was observed in TIVE-LTC cells (Figure 9C). COX inhibitor treatment of TIVE-LTC inhibited total as well as active-MMP-9 secretion (Figure 9D), indicating the importance of COX-2 in regulating MMP-9.
A similar analysis was performed for total and active-MMP-2. In agreement with MMP-antibody array data, we observed a 3.3, 3.3, and 3.5-fold induction in total MMP-2 release at 4h, 8h, and 24h PI (Figures 9E and 9F). NS-398 pretreatrment decreased the secretion of total MMP-2 but not that of active MMP-2 (Figure 9E). COX-2 depletion did not effectively inhibit either total or active MMP-2 induced by KSHV (Figure 9F). Latently infected TIVE-LTC cells demonstrated roughly about 1.8 and 1.5-fold induction in the release of total and active-MMP-2 as compared to TIVE cells (Figure 9G). Similar to total MMP-2 secretion, COX inhibition also regulated the secretion of active-MMP-2 and this activity was reduced in the cells treated with NS-398 (35%) or Indo (34%) for 24h (Figure 9H).
To assess the functionality of active MMP secretion upon KSHV infection, we performed invasion assays as described in the methods section. Figures 9I-9L represent the data obtained using Innocyte cell invasion assay while Figures S9 and S10 represent the invasive potential of the supernatants (used in Figures 9I-9L) as analyzed by Chemicon cell invasion assay. Similar results were obtained from both the methods used.
To evaluate the effect of KSHV infection on cell invasion, we infected HMVEC-d cells with KSHV at 30 DNA copies/ cell. At 24h PI, we assayed the ability of the cells to invade the ECMatrix barriers. Without chemoattractant gradients, the intrinsic invasiveness of normal HMVEC-d cells through an ECMatrix barrier was undetectable (data not shown). However, in the presence of complete growth medium as chemoattractant, some normal HMVEC-d cells succeeded in invading the ECMatrix barrier (Figure S9B; panel 1). KSHV-infected cells displayed increased invasiveness that was 4.5-fold higher than uninfected (169 cells/field versus 43 cells/field) (Figure S9B; panels 2 and 1). In contrast, NS-398 pretreated and then KSHV-infected cells showed reduced (68%) invasiveness (52 cells/field versus 169 cells/field) (Figure S9B; panels 4 and 2). This reduction in invasiveness was due to COX-2 inhibitor pretreatment rather than a non-specific effect of the NS-398 solvent as invasiveness in the solvent treated and infected cells was similar to that of cells infected with KSHV alone (164 cells/field versus 169 cells/field) (Figure S9B; panels 3 and 2).
To demonstrate whether KSHV infection could promote cell invasion in a paracrine fashion, we assessed the invasiveness of normal HMVEC-d cells in the presence of supernatants from uninfected- or KSHV-infected HMVEC-d cells. KSHV infection increased HMVEC-d cell invasiveness by 3-fold (47 versus 112 cells/field), 3.5-fold (52 versus 158 cells/field) and 3.5-fold (54 versus 180 cells/field) at 4h, 8h, and 24 h, respectively (Figures 9J and S9C; panels 1–3 versus 4–6). NS-398 pretreatment reduced KSHV promotion of cell invasion by 50% (77 versus 112 cells/field), 70% (43 versus 158 cells/field), and 70% (48 versus 180 cells/field) after 4, 8, and 24h, respectively. This indicated that KSHV induced COX-2 plays a critical role in regulation of MMPs and associated invasion (Figures 9J and S9C; panels 4–6 versus 7–9). Supernatant obtained from solvent pretreated infected cells was similar to KSHV infected cell culture supernatant (data not shown). Similar to NS-398 pretreatment, COX-2 silencing also reduced KSHV promotion of cell invasion (Figure 9K). This further validated the role of KSHV induced COX-2 in cell invasion, and suggests that KSHV infection could promote COX-2-dependent cell invasion through both autocrine and paracrine mechanisms. HT1080 cells showed maximum invasion (Figure S9C, panel 10).
Latently infected TIVE-LTC cells (169 cells/field) (Figures 9L and S10; panels 3 and 4) showed 3-fold increased invasiveness compared to TIVE cells (57 cells/field) (Figures 9L and S10; panels 1 and 2). This suggested that KSHV infection mediated secretion of proteases must be contributing to the invasive phenotype of TIVE-LTC cells. NS-398 pretreatment for 24h reduced TIVE-LTC cell invasion by 63% (62 cells/field versus 169 cells/field) (Figures 9L and S10; panels 7 and 8 versus 3 and 4) whereas Indo pretreatment reduced invasiveness by 43% (96 cells/field versus 169 cells/field) (Figures 9L and S10; panels 5 and 6 versus panels 3 and 4). Solvent treatment did not have any effect on TIVE-LTC cell invasion (176 cells/field versus 169 cells/field) (Figures 9L and S10; panels 9 and 10 versus panels 3 and 4) further validating the specific regulation of invasion by COX-2.
COX-2 expression and PGE2 secretion has been shown to accelerate integrin dependent cell adhesion, migration and cell-spreading [50]. Progression of KS from early stage to an invasive and metastatic phenotype is accompanied by a series of changes associated with cytoskeleton rearrangements as well as alterations in cell-cell and cell-matrix adhesion that allows cells to invade surrounding tissues and metastasize. To understand the role of KSHV induced COX-2 in the adhesion of endothelial cells, an adhesion assay was done using untreated maxisorp plates or plates coated with polylysine or fibronectin. Adhesion in the presence of polylysine was interpreted as the result of interaction between the polyanionic cell surfaces and the polycationic layer of adsorbed polylysine and reflective of charge based interactions rather than a response to secreted factors or surface expression of various integrins. Adhesion in the presence of fibronectin was interpreted as the result of interaction with integrins as it is an extracellular matrix glycoprotein that binds to integrins. Since we observed maximum PGE2 secretion during primary infection at 2h PI [26], we collected supernatants at 2h PI to demonstrate the paracrine role of PGE2 in the presence or absence of drug and used to test their ability to induce adhesion of uninfected cells. Conditioned medium collected during later time points of infection representing latency were not used for these assays.
Endothelial cells were allowed to adhere in the presence of the culture supernatant obtained from uninfected endothelial cells (2h) or the cells infected with KSHV for 2h, or cells pretreated with NS-398 for 1h and then infected with KSHV for 2h. Previously, we have shown that pretreatment of HMVEC-d cells with NS-398 inhibited the secretion of PGE2, suggesting that these supernatants would be depleted of PGE2. When plated on untreated plates, compared to the adhesion of cells in the presence of supernatant from uninfected cells, adhesion of uninfected HMVEC-d cells increased in the presence of culture supernatant from KSHV infected endothelial cells (Figure 10A). Treatment of the plate surface with polylysine increased the kinetics of binding irrespective of the presence of various culture supernatants (Figure 10B). When plated on fibronectin coated plates, cell adhesion kinetics were faster in the presence of infected cell culture supernatant suggesting a role for paracrine factors in the expression of integrins and its interaction with the integrin ligand, fibronectin (Figure 10C).
To address the role of PGE2 in HMVEC-d cell adhesion, we plated the cells in the presence of serum free medium containing 1 µM PGE2 which markedly enhanced HMVEC-d adhesion to the untreated as well as fibronectin coated plates (Figures 10A and 10C). The adhesion kinetics on the fibronectin plates was faster than adhesion to the untreated plates. Compared to the polylysine coated plates, PGE2 increased cell adhesion to the untreated and fibronectin coated plates suggesting its role in regulating the expression of surface molecules, possibly integrins or adhesion molecules, on the endothelial cells to facilitate rapid adhesion.
To understand the role of COX-2 inhibition and abrogated secretion of PGE2 in endothelial cell adhesion, we used the supernatant obtained from the cells pretreated with NS-398 and then infected with KSHV. Adherence of cells in the presence of NS-398 treated KSHV infected culture supernatant was comparatively less on all plates (untreated, fibronectin coated and polylysine coated) (Figures 10A-C) but decreased appreciably in the fibronectin coated plates (Figures 10A and 10C). This further confirmed the critical role of KSHV induced COX-2/PGE2 in endothelial cell adhesion. The role of PGE2 secretion in endothelial cell adhesion was further confirmed by using culture supernatants from KSHV infected si-C-, si-COX-2-1 or si-COX-2-2 -HMVEC-d cells (Figures 10D-F). Adhesion on untreated plates or fibronectin coated plates was reduced significantly in the presence of supernatants from si-COX-2-1 and -2 and KSHV infected HMVEC-d cells when compared to si-C-infected HMVEC-d cells (Figures 10D-F). This confirmed the involvement of KSHV infection induced COX-2/PGE2 in cell adhesion. Effects were more pronounced on fibronectin coated plates suggesting that COX-2/PGE2 mediates endothelial cell integrin expression or modulation of cell adhesion molecules. These results clearly demonstrated that COX-2/PGE2 play pivotal roles in cell adhesion to the matrix, an important event in KSHV pathogenesis that has seen little exploration.
Rearrangement of the actin cytoskeleton is primarily controlled by members of the Rho-GTPase family such as RhoA, Rac1, and Cdc42 [51]. Our earlier studies have demonstrated the activation of these GTPases by KSHV infection and stimulation by interaction of the KSHV envelope glycoprotein gB with adherent endothelial or fibroblast cell integrins [25],[35],[52]. RhoA-GTPases are implicated in regulating morphology and adhesion because interactions between the actin cytoskeleton and adherens junctions determine cell shape and motility [51]. RhoA and Rac have been shown to be critical regulators of cell adhesion and cell spreading while COX-2/prostaglandin production has been reported to be essential for integrin-dependent Rac activation in HUVEC cells [53]. Hence, we assessed the role of KSHV infection induced PGE2 in regulating these signaling molecules.
First, we asked the question whether secreted factors participate in the activation of RhoA- or Rac-GTPases. We quantified the RhoA-GTPase activity using a RhoA-GLISA kit on the lysates prepared from cells grown on untreated plates for different time points in the presence of culture supernatants from serum starved (8 h) uninfected HMVEC-d (2 h) and KSHV infected (2 h) cells. We observed 3.4, 3.9, 3.8 and 4.5-fold RhoA-GTPase activation upon plating cells in the presence of infected cell culture supernatant for 15′, 30′, 45′ and 60′, respectively (Figure 10G). This data suggested that the factors released during KSHV infection up-regulated RhoA-GTPase in the adhering cells.
Next, to understand the role of PGE2 in the regulation of RhoA-GTPase activity of infected cells, we tested the lysates from cells plated on untreated plates for 30′ in the presence of supernatant from cells uninfected or infected for 2h, and cells pretreated with 50 µM NS-398 for 1h and then infected with KSHV for 2h. Supernatant from the cells pretreated with NS-398 moderately inhibited activation of RhoA-GTPase (Figure 10G) suggesting that PGE2 secretion might not be involved in the stimulation of RhoA in HMVEC-d cell adherence. Activation of RhoA-GTPase was also measured in the lysate from cells grown in the presence of PGE2 for 30′. Induction of RhoA was ∼50% lower when compared to RhoA-GTPase activity in the presence of infected cell culture supernatant. This suggested that PGE2 alone is not enough to induce RhoA-GTPase in adherence of endothelial cells. Similar results were obtained from lysates prepared from cells plated on fibronectin coated plates thus ruling out the possibility of PGE2 participation in the interaction of integrins modulating RhoA-GTPase activity (Figure 10G).
As Rac-GTPases also play an important role in cell spreading and cell adhesion, we analyzed the activation kinetics of Rac1 by Rac1-GLISA. We observed 2.7, 3.8, 4.4 and 4.2- fold Rac1-GTPase activation upon plating cells in the presence of infected cell culture supernatant for 15′, 30′, 45′ and 60′, respectively. This data suggested that the factors released during KSHV infection up-regulated Rac1-GTPase in adherent endothelial cells. Supernatant prepared from cells pretreated with NS-398 drastically (65%) inhibited Rac1-GTPase activation (Figure 10H) suggesting that PGE2 secretion is involved in the stimulation of Rac1 in endothelial cells. To further confirm the role of PGE2, activation of Rac1-GTPase was measured in the lysate prepared from cells grown in the presence of PGE2 for 30′. Induction of Rac1-GTPase was 4.8- fold when compared to Rac1-GTPase activity in the presence of uninfected cell culture supernatant suggesting that PGE2 is enough to induce Rac1-GTPase in the adhering endothelial cells. Similar results were obtained from lysates prepared from cells plated on fibronectin coated plates thus demonstrating the possibility that PGE2 participates in the interaction of integrins modulating Rac1-GTPase activity (Figure 10H).
Rac1 activity was further confirmed using a PAK pull-down assay and fold activation was calculated by considering the Rac1-GTPase activity in the presence of uninfected supernatant at 60′ as one fold. About 1.8, 2, 3 and 2.5- fold activation of Rac1-GTPase was observed at 15′, 30′, 45′, and 60′, respectively (Figure 10I). Supernatant from the cells pretreated with NS-398 and then infected inhibited Rac1 by 75% (Figure 10J, lanes 3 and 1). This suggested that PGE2 secretion plays an important role in Rac1 stimulation, which was further supported by the 2.4-fold activity of Rac1 in the lysates prepared from cells plated in the presence of PGE2 alone for 30′ (Figure 10J, lane 4) which was comparable to the activity observed in the presence of infected cell culture supernatant (Figure 10J, lane 4). HMVEC-d cells plated on fibronectin coated plates cultured in the presence of the supernatant obtained from the 2h infected cell showed 2.5-fold activation of Rac1 which was completely abrogated (100% inhibition) in the presence of supernatant prepared from NS-398 pretreated and then infected endothelial cells (Figure 10K, lanes 1–3). Stimulation of Rac1 activity in PGE2 stimulated HMVEC-d cells grown on fibronectin plates was comparable to the activity observed in the presence of infected cell culture supernatant (Figure 10K, lanes 3 and 4). Collectively, these results clearly demonstrated that KSHV induced COX-2/PGE2 in the infected cell microenvironment plays an important role in endothelial cell adhesion by modulating the activity of Rac1-GTPases.
NSAIDs and derivatives of COXIBs are well documented for their anti-neoplastic activities such as inhibition of cancer cell line growth as well as initiation and promotion of apoptosis in various cancers [54]. Since KSHV induced COX-2 has an important role in regulation of viral latent gene expression that is linked to prolonged host cell survival, we assessed the effect of long term incubation of COX inhibitor (up to 96h) in latently infected endothelial cells (TIVE-LTC). To obtain the normal growth curve, cells were cultured in growth medium for 24 h, 48 h, 72 h, and 96 h before an MTT assay was performed. Results shown in Figure 11A depict growth kinetics of both cell types under normal conditions which clearly indicate that at any given time TIVE-LTC cells grow much faster than TIVE cells (Figure 11A). We next determined whether the longer duration of COX inhibitor treatment should be given in the presence of complete growth medium conditioned with serum or under serum starvation. Under serum deprivation, both cell types show reduced proliferation and the control TIVE cells appeared to be particularly dependent on serum growth factors for viability. This suggested that KSHV in TIVE-LTC cells must be inducing the secretion of growth factors that help target cell survival (Figure 11A). Even though the MTT assay measures mitochondrial activity in viable and in growth-arrested cells, its dynamic range is limited and can only be taken as an indicator for initial changes in cell survival. Therefore, we used traditional viable cell counting in similar experiments (Figure 11A). TIVE LTC cells displayed faster growth kinetics than TIVE cells thus validating the data obtained with the MTT assay (Figure 11A). Although, TIVE-LTC cells did not show profound death upon serum deprivation, TIVE cell viability was reduced by 8%, 32% and 66% at 48 h, 72 h and 96 h, respectively (Figure 11B). This further demonstrated the role of KSHV in secretion of various growth factors required for cell survival.
Next, we examined the effect of COX inhibitors on metabolism and growth index in latently infected cells. Neither fresh growth medium nor additional drug was added during the observation period for MTT and trypan blue exclusion assays. In MTT assays, we observed significant inhibition of TIVE-LTC cell metabolic activity with both drugs at all time points tested and the inhibition was marginally more with Indo for 72 h and 96 h (Figure 11C). Treatment for the same duration with solvent alone did not inhibit the metabolic activity of these cells (Figure 11C) thus validating the specific effect of COX inhibition. Similar to the MTT assay, we observed that treatment of TIVE-LTC cells with Indo reduced cell viability by 6%, 37%, 41% and 55% at 24 h, 48 h, 72 h, and 96 h (Figure 11D) while NS-398 treatment reduced cell viability by 4%, 11%, 31%, and 42% at 24 h, 48 h, 72 h, and 96 h, respectively (Figure 11D). It should be noted that TIVE and TIVE-LTC cells have hTERT which has been shown to be regulated by COX inhibitors [55]. To rule out the possibility that reduced cell viability and decreased cell metabolic activity observed in TIVE-LTC cells is not because of hTERT modulation of KSHV gene expression and related events in pathogenesis, we assessed the role of drug treatment for longer duration on the control TIVE cells (Figures 11E and 11F). NS-398 treatment did not reduce the metabolic activity of TIVE cells (Figure 11E), whereas Indo treatment affected TIVE cell metabolic activity only marginally, by about 9% and 15% at 48h and 72h of incubation, respectively (Figure 11E). Similarly, Indo treatment reduced TIVE cell viability only by 2% and 11% at 48h and 72h, respectively (Figure 11F).
Since longer incubation (48h, 72 and 96h) with the COX inhibitors reduced cell viability as observed by MTT assay, we assessed the effect of the COX inhibitor treatments on cell cycle profiling of TIVE and TIVE-LTC cells (Figure 11G, S11). To determine whether growth inhibition by COX inhibitors was attributable to cell cycle arrest, TIVE-LTC cells were treated with and without COX inhibitors for 24–96 h. According to the DNA profile, a significantly higher proportion of untreated TIVE-LTC cells were in S-phase compared to either Indo or NS-398 treated cells over longer incubation periods (48–96 h). We observed a clear anti-proliferative shift in the profile of the cell cycle parameters towards a reduced percentage of cells at the S and G2/M phases, together with an increased percentage of cells at the G1 phase. Approximately 70% reduction in S phase was observed in cells treated with COX inhibitors for 96 h (Figure 11G). There was not much change in the G2/M phase but in the drug treated cells, there was subsequent cell accumulation in the G0/G1 phase suggesting that COX inhibitors inhibit latently infected cells from crossing the G1/S boundary. Similar results were obtained with NS-398 treatment but solvent treatment did not affect the cell cycle profile of TIVE-LTC cells (data not shown) further validating the specific effect of the drug used for treatment. As 24 h treatment of TIVE-LTC cells with COX inhibitors could reduce ORF73 gene expression, we also assessed the long term effect of these drug treatments on KSHV latent gene expression. We observed 75–80% reduction in ORF73 gene expression in these cells after 96 h incubation with drugs (data not shown). Compared to TIVE–LTC, untreated TIVE cells showed shorter S phase (14%) and these cells were not affected by COX inhibitor treatment even after longer incubations (data not shown). These observations clearly suggest that COX-2 inhibitor treatment for longer durations slows proliferation of virally infected cells accompanied by reduced viral latent gene expression and thereby subsequently reducing the secretion of growth factors required for infected cell survival.
KS, a chronic inflammation associated tumor, is the most common aggressive malignancy among untreated HIV-infected patients. Progression of KS lesions, with its spindle shaped cells of endothelial origin, neovascular structures and inflammatory cells, is believed to be profoundly driven by the autocrine and paracrine loops of ICs, GFs and angiogenic factors present in the lesion microenvironment [2]. Hence, to effectively control and eliminate KS lesions, it is very important to understand the driving forces behind the initiation and maintenance of these secreted factors. KSHV latent gene expression observed in KS endothelial cells and the lytic gene expression observed in a limited percentage of KS inflammatory cells probably contributes to the up-regulation of cellular host factors and a synergy between host and viral factors could be contributing to the sustained induction of inflammatory molecules and progression of lesions. Our study unravels the multifactorial complexity of KSHV-host interactions governing KS progression/pathogenesis in conjunction with host factor COX-2 and its inflammatory metabolite PGE2 (Figure 12). The exhaustive studies presented here demonstrate that COX-2/PGE2 induction upon KSHV infection is an excellent example of synergy between host and viral factors, where COX-2 operates like a central player and performs dual functions of controlling downstream consequences of KSHV infection as well as latent viral gene expression. COX-2 not only helps maintain an angiogenic and inflammatory cytokine rich KSHV permissive microenvironment but also sustains viral latent gene expression critical for infected cell survival. Our findings also suggest the potential for COX-2 inhibition based therapies in treating angio-proliferative KS lesions as COX-2 silencing, as well as treatment with COX-2 chemical inhibitors, could effectively reduce inflammation, angiogenesis, cell adhesion, cell invasion and cell signaling events during KSHV de novo and latent infection.
Our study illustrates a continuous loop of events where COX-2 behaves like an intermediate in controlling KSHV pathogenesis and latency. However, the mechanism behind sustained COX-2 protein levels and tight regulation of its activity is still missing in this loop. Attractive answer to this question would be the interplay of viral proteins or, PGE2 per se or sustained levels of secreted factors or signal molecules like NF-κB [56]. Studies are underway to understand the transcriptional and post-transcriptional regulation of COX-2 during KSHV infection.
Expression of COX-2 in KS lesions, the detection of higher levels of PGE2 in KS tissue compared to surrounding normal tissue [57], significant over-expression of COX-2 in the early inflammatory/angiogenic stage as well as in the late nodular stage of classic and epidemic forms of KS lesions [58], together with our studies demonstrating COX-2 in KS lesions (skin tissue and lymph node), endothelial cells infected for 5 days and latently infected endothelial cells (TIVE-LTC) (Figures 1 and 2) emphasizes that prostaglandin cascade components are actively involved in KS pathogenesis and strengthens the role of COX/PGE2 in KSHV biology. Fold inductions for COX-2 and m-PGES-1 in HMVEC-d (5d inf.) cells with about 50–60% infection and TIVE-LTC cells expressing LANA in 50–60% of the cells were comparable. Our data demonstrate that 50–60% of TIVE-LTC cells were ORF73 positive and these cells were also positive for COX-2. In addition, COX-2 staining was also seen in the cells in close proximity to the uninfected cells which could be due to paracrine COX-2 stimulation by growth factors induced upon KSHV infection.
This study for the first time systematically evaluated the downstream consequences of KSHV induced COX-2/PGE2 by using COX-2 inhibition strategies involving parallel chemical inhibition and gene silencing approachs. This study also illustrates that cytokine secretion upon KSHV infection is not a random event but follows tightly regulated kinetics (Figure 4 and Table S3). It also indicates that inhibition of cytokines via inhibitor treatment is highly selective (Figure 5 and Table S4). For example, levels of IFN-γ, MCP-2, MCP-3, TARC, GCP-2, MIP-3α, Eotaxins, CK-β8-1, PDGF-BB, MCSF, G-CSF, GMCSF, angiogenin, VEGF, SDF-1, SCF, TGF-β1, leptin, and ILs (-3, -4 and -15) could be inhibited either by treatment with COX-2 inhibitor or NF-κB inhibitor [56]. Cytokines, like ILs (−5, −6 and −10) and GRO-α, were strongly inhibited by NF-κB inhibition, but not by COX-2 inhibition. In contrast, IL-1α, IL-1β, IL12-p40, TNF-α, IP-10, NAP-2, Oncostatin M, thrombopoeitin, FGFs (−4, −6, −7 and −9), Flt3-ligand, Fractalkine, IGFBPs and Osteoprotegerin were strongly inhibited by COX-2 inhibitor pretreatment demonstrating the specificity of downstream pathways regulated via COX-2 and NF-kB. Interestingly, KSHV infection induces sustained levels of NF-κB [56]. This, together with the fact that PGE2 itself can activate NF-κB, suggests the potential involvement of the COX-2-NFκB-COX-2/PGE2 axis during KSHV infection. Treatment of infected endothelial cells with Indo or NS-398 reduced the nuclear translocation of p65 (an indication of NF-κB activity) by 46% and 58%, respectively (data not shown), which suggests the involvement of COX-2/NF-κB in regulation of secreted factors. COX-2 inhibition could not impair cytokine secretion by 100%, which suggests the importance of viral and other host factors in controlling these cytokines.
To establish a lifelong successful infection in an immunocompetent host, KSHV must be utilizing an impressive array of immune modulatory mechanisms, one of which appears to be the induction of COX-2/PGE2. For example, the ability of COX-2/PGE2 to mediate regulation of IFN and RANTES (Table S4), involved in the recruitment of inflammatory cells, represents one strategy which KSHV utilizes to evade the host immune system. KSHV induced COX-2/PGE2 also regulates VEGF-A and VEGF-C (Figure 6), the multifunctional potent immunosuppressive cytokines that profoundly regulate cell growth, adhesion, angiogenesis, proliferation and differentiation, as well as FGF-4, PDGF, TGF-β, IL-1β and IL-6 which are known to up-regulate VEGF expression. Interestingly, increased COX-2 mRNA expression and PGE2 secretion has been shown to enhance VEGF mRNA expression suggesting a direct role for PGE2 in stimulation of angiogenesis [59]. VEGF-C and –A are also known to induce lymphangiogenesis and play key roles in lymphatic reprogramming involving the conversion of blood endothelial cells (BEC) to lymphatic endothelial cells (LEC) [60]; an important event in KS pathogenesis. In addition, reduced levels of IL-3 (Table S4), a known inducer of lymphatic markers Prox-1 and podoplanin in HMVEC-d cells, by COX-2 inhibition delineates a very significant role of COX-2 in KSHV lymphangiogenesis [61]. VEGF-A was found to be tightly regulated by COX-2/PGE2 in de novo infected HMVEC-d and TIVE-LTC cells (Figure 6). Similar to the involvement of COX-2/PGE2 in cytokine secretion in many inflammation related diseases [62], KSHV induced COX-2 also plays important roles in the expression and secretion of various chemokines, growth and angiogenic factors and thereby controls the angiogenesis and tube formation (Figure 7, S7, and S8) events of KS pathogenesis.
COX-2 has been implicated in invasiveness, angiogenesis and distant metastases of many cancers [63]. MMP secretion has been associated with many viruses, including EBV [64],[65],[66], hepatitis B virus (HBV) [67] and HIV-1 [68]. However, little is known about the functional role of MMPs and TIMPs in KSHV infection and KS. Our studies report for the first time the role of KSHV induced MMP-2 and MMP-9 (Figure 8) secretion in HMVEC-d and latently infected TIVE-LTC cells. HMVEC-d cell MMP secretion kinetics was different from the published kinetics in HUVEC cells [12], which could be due to cell type specific patterns of MMP secretion. TIVE-LTC cells secreted diminished levels of MMP-2 when compared to de novo infected cells (Figure 9E and 9G), therefore COX-2 inhibition could not effectively down-regulate active MMP-2 secretion during de novo infection (Figure 9E and 9F) compared to TIVE-LTC cells. Higher active-MMP-2 levels even upon COX-2 inhibitor treatment suggest that either MMP-2 might be controlled by factors other than COX-2/PGE2 or the inhibitor dose was insufficient to regulate its secretion. COX-2 inhibition specifically abrogated the expression and secretion of MMP-9 (9A-9D) in de novo infected as well as latently infected endothelial cells, a protease responsible for metastatic potential and triggering the angiogenic switch [16]. COX-2/PGE2 probably regulates MMP-9 at the transcriptional level by activating transcription factors like AP-1, Ets2, NF-κB, and Sp1 [33]. MMP-9 has the potential to increase VEGF release, its bioavailability to bind to VEGF receptors on endothelial cells, and thus leading to an angiogenic loop that eventually will result in cell migration, cell proliferation and angiogenesis [69]. COX-2 might be a key player regulating many feedback loops in cytokine-MMP interactions, including chemokines such as RANTES, TNF-α, GM-CSF and SDF-1 which induce MMP-9 that can cleave a spectrum of pro-cytokines like TNF-α, pro-TGFβ and IL-1β. COX-2 levels could modulate multifunctional TIMP-1 and -2, which can inhibit MMP activities and can activate the FAK/PI3-K or Src/PI3-K pathways [70],[71]. Along with the other above mentioned functions, PGE2 released early (2 h) during infection also enhanced the kinetics of endothelial cell adhesion (Figure 10). Increased adhesion could be attributed to various crucial factors controlled by PGE2, such as the activation of Rac1-GTPases (Figure 10), release of SDF-1 [72], IL-1β, TNF-α, along with the cell surface expression of adhesion molecules and integrins like αVβ3 and β1 [53]. Collectively, our study underscores the importance of the COX-2-PGE2-MMP-9 axis in KS pathogenesis and suggests that COX-2 inhibitors have tremendous therapeutic potential in KSHV biology.
Many viruses, such as herpes simplex virus (HSV), human cytomegalovirus (HCMV), pseudorabies virus (PRV), human herpesvirus-6 (HHV-6), EBV, murine herpesvirus 68 (MHV-68), and human T-cell leukemia virus type 1 (HTLV-1), have been shown to induce COX-2 and release PGE2 that participate in viral lytic replication. In contrast, the role of COX-2/PGE2 in KSHV infection appears to be different as our previous [26] and current studies demonstrate that COX-2/PGE2 not only regulates inflammation associated events via modulating cytokine secretion but also controls viral latency (Figure 2E) which has been shown to be essential for viral genome maintenance and host cell survival [23],[73].
COX-2 has been shown to play direct roles in the enchancement of tumorigenic and angiogenic factors in KSHV independent cancers. However, COX-2 appears to play an additional unique role in the context of KS pathogenesis and in creating a KS lesion microenvironment rich in cytokines since it also participates in KSHV biology by virtue of its ability to aid in the establishment and maintenance of latent gene expression. Since KSHV latent genes themselves are shown to be powerful mediators of anti-apoptosis, cell survival, as well as gene regulation including the induction of COX-2 and other cytokines and angiogenic factors [37],[56], our study exposes an interesting regulatory loop between KSHV induction of COX-2 expression, COX-2's role in the establishment and maintenance of KSHV latency and the induction of cytokines and angiogenic factors that sustains the KSHV-permissive microenvironment. Hence the observed effect of COX-2 inhibition on cytokines, angiogenesis and cell survival in the context of KSHV infected cells is probably not just due to COX-2's role as the direct activator of these processes but probably due to the combinatorial effect on reduction in KSHV latent gene expression and its downstream consequences including reduction in COX-2 expression, and cell survival. The observed regulatory loop of COX-2 in KSHV biology opens up a new avenue that could be potentially exploited for an effective control of KSHV and KS lesions.
Besides overcoming host intrinsic, innate and adaptive immune responses, survival of latently infected cells requires the constant blockage of apoptosis. Intriguingly, we observed that COX-2/PGE2 is involved in regulating latently infected TIVE-LTC cell survival (Figure 11). COX-2 inhibition for longer duration could shorten S phase, arrest TIVE-LTC cells at G1/S phase accompanied by further lowered ORF73 gene expression (Figure 11G and S11). An important question to be answered is whether exogenous supplementation of PGE2 in cells treated with COX-2 inhibitors will rescue cells from undergoing death or cell cycle arrest. Nevertheless, our study reveals that KSHV infection induced COX-2/PGE2 is an important anchor linking viral gene expression, GFs and cell survival in latently infected cells. We have demonstrated a reduction in ORF73 gene expression at early as well as later time points of COX-2 inhibitor treatment and a reduction in the cells in S phase at later times of drug incubation (Figure 11G). One of the key properties of LANA is to stimulate cells in the S phase entry [73] by relocalizing GSK3-β and stabilizing β-catenin, thereby manipulating the GSK3β-β-catenin complex. Decreased ORF73 gene expression upon COX inhibitor treatment and the shortened S phase of latently infected cells raises the possibility that COX inhibition might be disturbing the ability of LANA to interact with GSK-3β and the Rb protein required for G1/S progression. In other words KSHV might be utilizing COX-2 and PGE2 to stabilize these complexes required for successful latency. ORF73 gene expression is also critical to overcome the host chromatin-binding protein BRD4-and BRD2/RING3-stimulated G1/S arrest [74], therefore reduced ORF73 gene expression upon COX-2 inhibition could be pushing the cells to G1/S arrest.
This study also raises important questions including the role of COX-2/PGE2 in viral episome maintenance and their effects on LANA protein levels. As PGE2 is known to stimulate several signaling events (JNK-1, ERK1/2, PKC, PI3K-AKT, HPK1, Src), second messengers including cAMP, calcium and reactive oxygen species (ROS), and modulate various transcription factors (Ets-1, Sp1, Oct-1, STAT-3, AP-1, ELK-1, hypoxia inducible factor-1α, and β-catenin) [75],[76],[77],[78],[79],[80],[81],[82],[83],[84],[85],[86],[87],[88],[89], COX-2/PGE2 could be mediating their effect on KSHV latency via one or more of these factors. It is interesting to note that some of the PGE2 activated transcription factors (Sp1, HIF-1α and AP-1) are well established for their role in modulation of viral latency (ORF73) and lytic (ORF50) promoters [90],[91],[92]. Studies to decipher the molecular pathway of PGE2 mediated ORF73 promoter regulation and viral latency is under investigation. The ability of KSHV to utilize pro-inflammatory molecules to maintain latent gene expression demonstrates the plasticity of the KSHV genome and its adaptability to host surveillance.
Currently, there are no methods available to eliminate the latent infection of herpesviruses. Slow proliferation of KSHV latently infected cells accompanied by reduced viral latent gene expression upon treatment with COX-2 inhibitors (Figure 11) strongly demonstrates that COX-2 is an excellent target for controlling KSHV latency.
At present, two classes of COX inhibitors are currently available for use in humans. NSAIDs inhibit both COX-1 and COX-2 while COXIBs are COX-2 selective with very little effect on COX-1 and consequently have been described as “healthier, dedicated, and more targeted” [93]. Despite a few COX-independent actions [43] of chemical inhibitors, these are still the most promising drugs for treating inflammation associated cancers and are recognized as potentially effective antiviral, anti-mitogenic and anti-angiogenic compounds. Observations such as reduced b-FGF and VEGF secretion and MMP-9 regulation reveals that COX-2 inhibitors possess the potential to be exploited in the in vivo model to better understand their benefits as an adjuvant to the currently available chemotherapy for KS. Interestingly, PGE2 has also been shown to modulate several functions associated with rapamycin, a drug shown to be efficacious against PEL cell lines [94]. COX-2 inhibitors show additive effects when used as part of a combination therapy since they potentiate the effect of IFN-α in HCV infection [95] and IFN-γ in several tumors [96],[97]. Hence, their inclusion in combination with KS chemotherapy, radiation, and biological therapies might prove to be beneficial in the KS scenario. Effective inhibition of COX-2 could lead to reduced KSHV infection of endothelial cells which may in turn reduce the accompanying inflammation and KS lesion progression.
|
10.1371/journal.pcbi.1006141 | Ten quick tips for getting the most scientific value out of numerical data | Most studies in the life sciences and other disciplines involve generating and analyzing numerical data of some type as the foundation for scientific findings. Working with numerical data involves multiple challenges. These include reproducible data acquisition, appropriate data storage, computationally correct data analysis, appropriate reporting and presentation of the results, and suitable data interpretation.
Finding and correcting mistakes when analyzing and interpreting data can be frustrating and time-consuming. Presenting or publishing incorrect results is embarrassing but not uncommon. Particular sources of errors are inappropriate use of statistical methods and incorrect interpretation of data by software. To detect mistakes as early as possible, one should frequently check intermediate and final results for plausibility. Clearly documenting how quantities and results were obtained facilitates correcting mistakes. Properly understanding data is indispensable for reaching well-founded conclusions from experimental results. Units are needed to make sense of numbers, and uncertainty should be estimated to know how meaningful results are. Descriptive statistics and significance testing are useful tools for interpreting numerical results if applied correctly. However, blindly trusting in computed numbers can also be misleading, so it is worth thinking about how data should be summarized quantitatively to properly answer the question at hand. Finally, a suitable form of presentation is needed so that the data can properly support the interpretation and findings. By additionally sharing the relevant data, others can access, understand, and ultimately make use of the results.
These quick tips are intended to provide guidelines for correctly interpreting, efficiently analyzing, and presenting numerical data in a useful way.
| Data expressed as numbers are ubiquitous in research in the life sciences and other fields. The typical scientific workflow using such numerical data consists of analyzing the raw data to obtain numerical results, followed by interpreting the results and presenting derived findings.
In this article, we present some tips and tricks for working with numerical data. We discuss how data can be checked effectively for plausibility—and that this should be done frequently to spot potential mistakes early. Keeping clear records of the way raw numbers and intermediate or final results were obtained helps correct mistakes efficiently. To facilitate correct interpretation of data, we suggest paying attention to correct units of values and to understandable formulas. Appropriate statistical methods should be used for the specific question at hand. To avoid an over-interpretation of numbers, uncertainty of numbers should be respected, and one should be aware of cases in which bare numbers are actually misleading. Finally, we provide some recommendations for presenting data in a way useful for supporting interpretation and findings, ideally so that follow-up work can make use of the data.
| In many studies in the life sciences and other fields, findings are derived from some type of experiment via the analysis of numerical data in one form or another. The typical workflow, cf. Fig 1, involves multiple steps where good scientific practice needs to be followed so that sound results are obtained.
Experiments (regardless of whether they are in vivo, in vitro, or in silico) need to be designed and documented so that data of interest can be acquired reproducibly [1]. Data need to be stored in a reliable way that allows efficient finding and proper use [2]. Data processing and analysis need to be correct and reproducible in order to obtain meaningful results [3]. In particular, statistical analyses need to be done correctly [4]. Data need to be presented in an understandable form to show results and to support the conclusions drawn from it, which is typically achieved by presenting data in graphical form [5]. There are, however, many pitfalls when working with numerical data in these steps.
Mistakes when analyzing numerical data lead to incorrect results and incorrect interpretations. Correcting such mistakes can be time-consuming, depending on how much follow-up effort you or others have already spent, and realizing that one has wasted effort is usually frustrating. Not realizing mistakes yourself and presenting incorrect results to others can be rather embarrassing, in particular when the false results have already been published. Sources of errors in publications include but are not limited to applying inappropriate statistical methods [6] and incorrect interpretation of data by software [7]. Unfortunately, there is no way to completely avoid mistakes when working with data. (If you do know one, please let us know!) However, knowing about potential pitfalls, one can at least reduce the number of mistakes.
Properly understanding the data you are working with is probably the most important aspect when trying to avoid mistakes. Frequently asking yourself, “Does this make sense?” helps you spot mistakes early and avoids realizing at some later point, “Why didn't I see this stupid mistake earlier?!” Understanding the data at hand is also necessary to distinguish relevant effects in observed data from meaningless effects so that you can focus research and analysis efforts on interesting investigations to obtain valuable results.
The purpose of this article is to provide some guidelines for working with numerical data for researchers such as biologists, wet-lab experimentalists, computational scientists, and data scientists. We present a number of tips on how to make understanding numerical data easier, how to correctly work with it, and how to present results in a useful way to others. “Others” in this context might be a wide range of persons whom you might prefer not to annoy, e.g., a reader or reviewer of your research paper, your advisor grading your thesis, or your favorite colleague who will build on your work. This "other" person might also be yourself following up on previous work a few months from now.
The examples we chose for illustrating these guidelines range from rather generic applications to specific biological questions. Clearly, the effects we mean to point out by such examples may also occur more generally in various other contexts.
The article is structured in tips about how to keep data correct (Tips 1 to 3), how to correctly interpret it (Tips 4 to 8), and how to present it in a useful and nonambiguous form (Tips 9 and 10).
Frequent plausibility checking is probably the most useful approach to quickly spot mistakes—when acquiring data, in computations, and when interpreting results. Whenever you realize a mistake and you get the feeling you could have found it already much earlier, this can be a sign of too infrequent plausibility checking. Several techniques can help with plausibility checking.
Numerical results are typically based on literature data or on data acquired from your own experiments. In any scientific work, it is crucial to document in sufficient detail how these data were acquired and which calculations were performed so that the reported results can be reproduced. Below, we will discuss working with literature data and making calculations reproducible. For recommendations on how to document data acquisition in wet lab and in silico experiments, we refer to [10] and [1], respectively.
In order to get the implementation of data analysis right, it is necessary to know capabilities, limitations, and pitfalls of whatever tools you are using. Calculators require manual data input, which is error-prone and not automatic; software tools (Excel, R, Python, etc.) have their own peculiarities and possibly surprising behavior, e.g., in case of missing data or user input. In this tip, we discuss a few generic pitfalls without focusing on specific software.
The two previous pitfalls can be seen as instances of the problem that computer programs interpret data incorrectly. Another pitfall is spreadsheet software inadvertently interpreting input data as dates (possibly in a foreign language date format) or floating-point numbers in scientific notation. For example [7], the supporting information of publications containing gene names or identifiers in spreadsheets (e.g., Excel) is prone to errors like "SEPT2" (Septin 2) becoming "2-Sep" or "2006/09/02.” This and the two pitfalls above are cases of inappropriate data processed by functions that are correct and useful for appropriate input. In addition, there is always the possibility of incorrect function, i.e., bugs, in any software [17].
This type of problem can sometimes be detected by plausibility checking. To avoid these types of errors, using different functions or different software can help, or you may need to clean up the input data. In case of seemingly plausible but actually incorrect results (e.g., the maximum reported in Table 1), only a very detailed look at the data can help reveal the problem. In this case, one should be glad about every error message or program crash.
Unless numbers are clearly dimensionless (e.g., relations between the same type of measurements, counts), they are meaningless without knowing the unit. For instance, what does a temperature "in the 30s" mean? A warm summer day at 30°C in Paris and you have a problem if the fridge in your lab is broken? Should you expect snow at 30°F in Seattle, and you could just store your specimens (or the lunch you brought) outside? Are you experimenting with superconductors at 30 K in a physics lab and need protective gloves? Besides single numbers, equations also only become meaningful if their units are clear: "A [mathematical] model without units is not a model" (Matthias König).
Relations derived from numerical data (e.g., descriptive models) are typically written as formulas. Formulas by other authors found in the literature may be more or less cryptic at first glance and need to be understood correctly before making use of them. Besides tracing units in formulas (cf. Tip 4), further techniques can be employed to check formulas for plausibility or to interpret them correctly. This involves both the dependency on variables and constants in the formulas.
Descriptive statistics and statistical tests are frequently used when reporting numerical data and also represent a frequent cause of confusion. It is important to check the requirements for a given method or test, e.g., whether data are normally distributed when performing parametric tests that require normality. This already applies when reporting means and standard deviations, which has the implicit assumption that data are normally distributed. Otherwise, medians and inter-quartile ranges (or, graphically, box plots) might be more useful descriptions of the ranges of typical values. In this tip, we describe four further frequent causes of confusion when using statistical methods.
Besides tracking where values come from (cf. Tip 2), it is also important to keep track of how accurate they actually are. (Keep in mind that 68.432% of all statistics pretend to be more accurate than they actually are.)
Even when properly considering uncertainty in numbers (cf. Tip 7), quantitative descriptions of data should not be the only thing you look at. Clearly, it is typically necessary to simplify raw data for a useful presentation and to reach useful findings, but simplification may also lose relevant information and conceal parts of the bigger picture. For instance, Fig 3 shows a classical example of four clearly different data series which are indistinguishable based on common measures from descriptive statistics (mean and standard deviation in both dimensions) (see Table 4). The code for generating these plots is provided as Supporting Information S5 Data.
To properly present results based on numerical data, two aspects need to be considered: what are the best numbers to present for supporting your findings? What is the best way to present numbers?
Data analysis and presentation typically strongly reduces the amount of data that is finally reported (in a presentation, on a poster, in an article or thesis). Hence, it may make sense to also provide more of the original data (unless there are, e.g., legal restrictions or privacy concerns) and intermediate results to the interested reader, for example, as supporting information of publications, or via domain-specialized or generalist data repositories (see [55] for an overview; generalist repositories include Figshare [56], Harvard dataverse [57], the Open Science Framework [58], and Zenodo [59]). Providing data and code (e.g., scripts) used for data evaluation increases reproducibility and credibility of your results.
Publishing data via reliable platforms also ensures long-term availability of research results, which is a challenge otherwise [60]. Making research data, code, and results freely available ("open access" and/or "open data") is of increasing interest to funding agencies, e.g., the European Union [61]. Publishing research data is also encouraged to a different extent by many journals and publishers [62], including PLOS [63].
Preparing data and code for publication certainly requires some effort, but one also benefits from this work in different ways. Before being published, data and scripts need to be documented and prepared, e.g., in a suitable standardized format. This can help detect mistakes by simply looking at the material again but primarily helps make data and code understandable to other people (e.g., your immediate colleagues or even yourself a few months from now). Moreover, published data sets and scripts are also guaranteed to be available to yourself, e.g., after moving to a different lab. Using open source platforms for implementing data evaluation helps make such scripts reusable by yourself and others; cf. Tip 9 in reference [64]. Another aspect of sharing data and code is that other researchers might actually use it. This may help your work to get recognized and cited, fruitful new cooperations might evolve, and others might find gems in your data you will never even start looking for.
Following this recommendation, we provide the data and code used for our analyses in the supporting information to this article. Although we do not believe that there are further gems hidden in our data (the baseball data is not ours), we are looking forward to being proven wrong.
|
10.1371/journal.pntd.0002644 | Characterization of a Gene Family Encoding SEA (Sea-urchin Sperm Protein, Enterokinase and Agrin)-Domain Proteins with Lectin-Like and Heme-Binding Properties from Schistosoma japonicum | We previously identified a novel gene family dispersed in the genome of Schistosoma japonicum by retrotransposon-mediated gene duplication mechanism. Although many transcripts were identified, no homolog was readily identifiable from sequence information.
Here, we utilized structural homology modeling and biochemical methods to identify remote homologs, and characterized the gene products as SEA (sea-urchin sperm protein, enterokinase and agrin)-domain containing proteins. A common extracellular domain in this family was structurally similar to SEA-domain. SEA-domain is primarily a structural domain, known to assist or regulate binding to glycans. Recombinant proteins from three members of this gene family specifically interacted with glycosaminoglycans with high affinity, with potential implication in ligand acquisition and immune evasion. Similar approach was used to identify a heme-binding site on the SEA-domain. The heme-binding mode showed heme molecule inserted into a hydrophobic pocket, with heme iron putatively coordinated to two histidine axial ligands. Heme-binding properties were confirmed using biochemical assays and UV-visible absorption spectroscopy, which showed high affinity heme-binding (KD = 1.605×10−6 M) and cognate spectroscopic attributes of hexa-coordinated heme iron. The native proteins were oligomers, antigenic, and are localized on adult worm teguments and gastrodermis; major host-parasite interfaces and site for heme detoxification and acquisition.
The results suggest potential role, at least in the nucleation step of heme crystallization (hemozoin formation), and as receptors for heme uptake. Survival strategies exploited by parasites, including heme homeostasis mechanism in hemoparasites, are paramount for successful parasitism. Thus, assessing prospects for application in disease intervention is warranted.
| While isolating membrane-bound and secreted proteins as targets for Schistosoma japonicum vaccine, we identified a novel potentially functional gene family which had originated by a gene duplication mechanism. Here, we integrated structural homology modeling and biochemical methods to show that this gene family encodes proteins with sea-urchin sperm protein, enterokinase and agrin (SEA) –domain, with heme-binding properties. Typical of SEA-structural domains, the characterized proteins specifically interacted with glycosaminoglycans (GAGs), with implication in ligand gathering and immune-evasion. Consistent with modeled heme-binding pocket, we observed high affinity heme-binding and spectroscopic attributes of hexa-coordinated heme iron. Localization of the native gene-products on adult worm tegument and gastrodermis, host interfaces for heme-sequestration and acquisition, suggests potential roles for this gene family in heme-detoxification and heme-iron uptake.
| Schistosomiasis still ranks as the most important helminthic infection; second only to malaria in its socioeconomic burden in the resource constrained tropics and subtropics. It affects over 200 million people worldwide with more than 700 million people at risk of getting infected [1]. Although an effective treatment is available (praziquantel), the fact that reinfection occurs very rapidly after mass treatment renders chemotherapy alone inadequate for disease control. It is opined that a prophylactic alternative applied singly or in combination with other interventions, even with limited efficacy in limiting transmission is the optimum approach [2]. This intervention is especially needed in S. japonicum endemic areas, where non-human mammalian hosts are complicating control efforts.
Schistosomes inhabit host vasculature, where they ingest erythrocytes and catabolize the host hemoglobin as a source of amino acids for their growth, development and reproduction [3]. However, large quantities of potentially toxic heme (Fe-protoporphyrin IX) are released as ‘byproducts’ of hemoglobinolysis [3]–[6]. The parasite is thus faced with the challenge of maintaining heme homeostasis by evolving strategies to sequester and detoxify heme [3], [5]–[9], and at the same time maintaining a heme acquisition mechanism to harness the needed iron from the heme molecules [4], [10]. Indeed, effective mechanisms for detoxification of toxic heme and controlled acquisition of heme iron are paramount for parasite survival and establishment. Such mechanisms are major targets of effective drugs against hemoparasites, including malaria and schistosomiasis [11]–[13]. However, information on the exact mechanisms and molecules involved in this ‘weak link’ is either lacking or equivocal [3]. Such molecular targets should be localized at the host-parasite interfaces in contact with the host erythrocytes.
The tegument and gastrodermis are syncytial layers lining the entire parasite surface and the parasite gut, respectively [14]–[16]. Heme liberated during hemoglobinolysis is sequestered in the parasite gastrodermis lining the gut lumen [4], [17], and subsequently detoxified to non-toxic crystalline aggregates called hemozoin [8], [9], [17], [18] and regurgitated. The exact mechanism is not fully understood but it is thought that heme-binding proteins initiate the nucleation step of the crystallization, while lipids mediate the elongation step in an amphiphilic interface created by lipid droplets in the gastrodermis and gut lumen [17], [19]. Equally, schistosomes like other obligate parasites scavenge molecules from the host, including heme as the major source of iron needed for development and reproduction [4], [10]. Also, newly penetrated schistosomulae obtain iron via heme-binding proteins on their teguments before their guts are developed [20]. Thus, heme-binding proteins that are localized at these interfaces are most likely involved in the parasite heme acquisition and detoxification.
Over the years, enormous resources and technologies have been channeled towards identifying molecular targets involved in several biological mechanisms utilized by parasites for effective parasitism. The recently completed genome [21], transcriptome sequences [22] and proteomic studies [23] of this parasite represent invaluable feats towards identifying such targets. Although the functions of many sequenced genes are readily known or inferred from their amino acid sequences, many of the genes that are potential determinants of successful parasitism sometimes do not have readily identifiable sequence homologs. This is a major challenge for placing the vast amount of ‘omics’ data into functional contexts for identifying genes of interest [24], [25]. As a matter of fact, several of such proteins presently annotated as ‘hypothetical proteins’ may well represent the missing link to filling the gene ‘gaps’ in our understanding of host-parasite interactions. Indeed, over 30% of S. japonicum proteins are yet of unknown functions [21]. Therefore, adopting novel strategies for the characterization of otherwise ‘hypothetical proteins’ is highly needed and can provide valuable functional clues that may not be readily identifiable from sequence data alone [24], [25].
Our group had utilized a signal sequence trap (SST) to isolate secreted and membrane binding antigens from S. japonicum with appreciable success [26]. Among the SST isolated candidates, we identified a novel gene family which we found to have originated through a repetitive element mediated DNA-level gene duplication mechanism [27]. Although several transcripts from ∼27 duplicons were identified, no sequence homolog was readily identifiable in other organisms. We here utilized an integrated strategy combining comparative structural homology modeling and biochemical analyses to identify remote structural homologs, and characterize an extracellular domain in this family as SEA (sea urchin sperm protein, enterokinase and agrin)-domain. Similar approach was used to further identify and characterize a functional heme-binding site on the SEA-domain. SEA-module is an extracellular structural domain originally identified in sea urchin sperm protein, enterokinase and agrin, the basis for the nomenclature [28]–[30]. The domain is found in several functionally diverse proteins, and is known to assist or regulate binding to carbohydrate moieties. SEA-domain evolved from the ancestral ferredoxin-like fold, which is able to acquire various active sites including heme-binding sites [24]. The identification of a functional heme-binding protein in this hemophagous trematode is a significant contribution to our understanding of the host-parasite interaction as regards heme homeostasis. The biological significance of this finding and the potential role of this gene family in parasitism are discussed in terms of the parasite biology and prospects for application in disease intervention.
This study adhered strictly to the recommendations in the Fundamental Guidelines for Proper Conduct of Animal Experiment and Related Activities in Academic Research Institutions under the jurisdiction of the Ministry of Education, Culture, Sports, Science and Technology, Japan (Notice No: 71). All animal experiments were approved by Nagasaki University Board of Animal Research, according to Japanese guidelines for use of experimental animals (Approval No: 0809050699).
Six to eight weeks old Female BALB/c mice were purchased from SLC Inc. Labs, Japan. The CLAWN strain miniature pigs were from Japan Farm, Kagoshima, Japan. The miniature pigs were infected percutaneously with 200 S. japonicum cercariae.
Multiple alignments were performed using NCBI BLAST and Multialin Interface [31]. Post translational modifications were predicted using YingOYang 1.2 [32]. Molecular structure modeling was performed by fold recognition and ab-initio structure prediction methods using Protein Homology/Analogy Recognition Engine (Phyre v2.0) [33] and Rosetta Full Chain Protein Structure Prediction Server [34]. Ligand binding analysis to identify potential ligands and their binding sites in the folded protein was performed using 3DLigandSite server [35]. The modeled structures were analyzed using Discovery Suite 3.5 Molecular Visualizer, while the modeled receptor-ligand interactions were analyzed on the PyMol Molecular Graphics System, Version 1.6 (Schrodinger, LLC).
Total mRNA was purified from parasite egg, sporocyst, cercaria and schistosomula using Micro-to-Midi total RNA purification system (Invitrogen, USA), and from adult worms using NucleoSpin RNA II kit (Macherey-Nagel, Germany). Reverse transcription and amplification of the double stranded cDNAs were performed using Ovation Pico WTA System v2 (NuGEN, USA). For each candidate gene and the reference gene (S. japonicum β-Actin), PCR fragment was first cloned into pCR2.1 cloning vector and the resulting constructs used as templates for qPCR standards and for estimation of copy numbers. Relative expression of candidate genes in different developmental stages of the parasite was quantified using SYBR Premix Ex Taq II Reagents (Takara, Japan). Real-time PCR and data analysis were performed on AB 7500 Real-Time PCR Systems v2.0.5.
The complete coding sequences of the candidates were amplified and cloned into the TOPO TA cloning site of the expression vector pcDNA4/HisMax and expressed in BL21 E. coli cells, and FreeStyle 293 expression system (Invitrogen, USA) for binding assays. We took advantage of His6 tag to purify the recombinant proteins using TALON Metal Affinity Resins (Clontech, USA). Purified proteins were concentrated and imidazole elution buffer exchanged using Amicon Ultra Centrifugal Filters (Millipore, USA). Size exclusion gel filtration was performed using Sephadex G-50 medium (GE healthcare, USA). For heme-binding assays, purified proteins from FreeStyle 293 cells were treated with enterokinase to remove tags and purified with EK-Away resin (Invitrogen, USA).
Polyclonal mouse sera were produced against recombinant antigens by subcutaneous immunization of mice with 25 µg of purified recombinant proteins in 50 µl PBS, mixed with an equal volume of Gerbu Adjuvant 100 (GERBU Biotechnik, Denmark), on days 0, 21 and 42. Two weeks after the last inoculation, mice were exsanguinated to collect sera and spleens were aseptically obtained for monoclonal antibody preparation using the Clonacell-HY system (Stemcell Technologies, USA), according to manufacturer's instructions. The monoclonal antibodies were biotinylated using the one-step antibody biotinylation kit (Mitenylbiotech, USA).
Freshly perfused adult S. japonicum were washed three times in PBS (pH 7.4) and fixed in 4% neutral paraformaldehyde at 4°C until use. The samples were alcohol dehydrated, embedded in paraffin, cut into 5–7 µm thin sections and then mounted on microscope glass slides. Paraffin sections were deparaffinized by incubating for 10 min in two changes of xylene and rehydrated by sequential 10 min incubations in 100%, 95%, 70% and 50% ethanol, before rinsing in two changes of double deionized water. Schistosomulae were prepared by mechanical transformation and washed in Hanks solution. After washing with distilled water, the juvenile worms were fixed in cold acetone for 2 hours. Two drops of acetone fixed schistosomulae were added to poly-L-lysine coated glass slides and dried overnight. Immunoperoxidase technique was then performed as in adult worm sections.
Immunoperoxidase staining and immunofluorescence assays were performed using minor modifications to the method detailed by [36]. Briefly, the sections for immunoperoxidase staining were treated with 3% H2O2 in PBS for 30 min to destroy endogenous peroxidase. All sections were blocked for non-specific binding with 5% skim milk in PBS for 1 h, and then incubated for 2 h at room temperature with biotinylated monoclonal antibody or immune sera as indicated in each case. After washing three times in PBS pH 7.4 for 5 min each, the sections were incubated in FITC conjugated secondary antibody for immune sera IFA. For biotinylated mAB IFA and immunoperoxidase assays, sections were incubated for 30 mins with streptavidin-FITC (1∶500) and streptavidin-HRP (1∶500) solution respectively. The immunoperoxidase sections were washed in PBS and treated with diaminobenzidine tetrahydrochloride (DAB) chromogen, according to manufacturer's instructions (Dako, Japan). After counterstaining immunoperoxidase sections with Mayer hematoxylin, all the sections were washed, dehydrated by passage through alcohol and xylene, mounted, and viewed under Keyence All-in-one Fluorescence Microscope (Keyence, USA). Pre-immune serum was used as negative control.
For glycoprotein detection assay, SDS-PAGE fractionated purified recombinant proteins were stained using the Pierce Glycoprotein Staining Kit (Thermo Scientific, USA).
We utilized array type sugar chip (SUDx-Biotec, Japan); which is an array of 48 structurally defined sugar chains (glycans) immobilized on a thin gold chip to analyze the interactions of the SEA-domain proteins with glycans using SPR imaging [37]. The surface plasmon is excited when light is focused on the opposite side of the chip. The reflective light is measurable and is altered in response to binding of the proteins to the immobilized glycans. This alteration of the surface plasmon (expressed as resonance units, RU) is directly proportional to change in bound mass of analytes. Real time measuring of the SPR RU was used to monitor changes in the surface concentration or amount of bound analytes (protein). One of the benefits of this SPR system is that the weak interactions, which are easily washed out in the regular array technology and therefore not recognized, can also be monitored in real time. We used this method to detect real-time biological interactions between several glycans and the characterized SEA-domain proteins. For assessing the specificity and affinity of the protein-glycan interactions, we used chondroitin sulfate GAG chip to measure the association and dissociation kinetics in real time to determine KD of the binding.
Hemin-agarose binding assay was applied to study heme binding as detailed by [38]. Briefly, 200 µl of hemin-agarose (Sigma-Aldrich, USA) was washed three times in 1 ml of 100 mM NaCl-25 mM Tris-HCl (pH 7.4) with centrifugation done at 750×g for 5 min. Hemin-agarose was incubated with protein (20 µg) for 1 h at 37°C with gentle mixing. After 4 washes to remove unbound proteins, the beads were incubated for 2 min with elution solution (2% (wt/vol) SDS and 1% (vol/vol) β-mercaptoethanol in 500 mM Tris HCl, pH 6.8), boiled at 100°C for 5 min; centrifuged, and the supernatant analyzed by SDS-PAGE.
Binding assay based on the peroxidase activity of bound heme was performed as detailed by [38], [39]. Briefly, micro-titer plate coated with serial dilutions of the recombinant protein was incubated with hemin (20 µg/100 µl) at 37°C for 1 h. The unbound hemin was removed and the wells were washed three times with PBS (pH 7.3). 50 µl of ready-to-use substrate tetramethylbenzidine/H2O2 (TMB) (Bangalore-Genei, India) was added and the reaction stopped after 15 min with addition of equal volume of 1N H2SO4. The OD450 was determined in an ELISA plate reader (Bio-Rad, USA). The amount of hemin bound to protein was calculated from a linear graph of the peroxidase activities of known concentrations of hemin.
Optical absorption spectrometric studies were performed on Hitachi U-3900H spectrophotometer according to method detailed by [40]. Briefly, the binding of proteins to heme was titrated by adding increasing amount of the protein (0–28 µM) to 10 µM of heme in 40% dimethyl sulfoxide (DMSO) buffered with 20 mM HEPES (pH 7.4). Difference in absorption spectra over a range of 350 to 700 nm was recorded. We used the increase in absorbance at Soret peak (412 nm) to monitor the formation of the protein heme complex. The heme binding curve was constructed by plotting the change in absorbance at the Soret peak (ΔA412) versus the protein concentrations. The heme-binding curve was fitted using one site specific binding with Hill slope model on GraphPad Prism, v5.00.
Data analysis was performed on GraphPad Prism, v5.00. Mann-Whitney test was used to compare differences between two groups, while Kruskal-Wallis test was applied to compare differences among several groups. All plotted data are means with error bars representing standard deviation (SD). Statistical significance was designated as p<0.05.
PFAM: PF01390, SCOP: 82671, SCOP: 54861, PDB: 2e7v, PDB: 2acm, PDB: 1ivz, GenBank: AY570748, GenBank: AY570737, GenBank: AY570742.
We had identified a novel gene family with similar signal sequence and promoter regions among SST isolated cDNAs (Figure S1A) [26], and showed that this gene family had originated from retrotransposon-mediated gene duplication mechanism [27]. Although several transcripts from ∼27 duplicons were found to belong to this family, we could not readily identify the molecular functions of these genes since no sequence homolog was readily identifiable in any other organism [27]. Consequently, we utilized comparative structural homology modeling to identify features and domains that could predict the putative molecular functions of the encoded proteins. Firstly, protein topology indicated that while all the members of this family bear similar signal sequence and are thus trafficked to the surface; some also contain C-terminal transmembrane regions, akin to type-I transmembrane proteins (Figure S1B).
The molecular folding patterns of the proteins were modeled simultaneously in Phyre 2 and Rosetta using fold recognition and ab-initio structure predictions (Figure S1C). These programs create sequence alignment profiles from PSI-Blasts followed by scanning of ‘fold library’ to identify remote structural homologs from experimentally determined structures in PDB and SCOP databases [33], [34]. The secondary structure components showed antiparallel arrangement of β-sheets, backed by α-helices (Figure 1A), typical of ferredoxin-like folds. Interestingly, models from both programs identified an extracellular domain of ∼120 amino acids common among this family, with striking similar folding pattern as SEA-domain (sea urchin protein, enterokinase and agrin) [PFAM: PF01390; SCOP: 82671] (Figure 1A and Table S1). SEA-domain is a domain with ferredoxin-like fold [SCOP: 54861], found in several proteins of diverse functions in different organisms [28]–[30], [41], [42]. Notably, crystal structure of the SEA-domain of transmembrane protease serine II (TMPRSS2) of Mus musculus [PDB: 2e7v] was the highest scoring template at over 95% confidence, according to which the shown structures were modeled. For clarity, only the original SST identified candidates are shown as representative of the family (Figure 1A). The structural models for all members of the gene family are summarized in Table S1. Other high scoring homologs at over 95% precision were the SEA-domains of Mucin 1 [PDB: 2acm] and Mucin 16 [PDB: 1ivz].
To validate the models, rigid body superposition with the highest scoring template [PDB: 2e7v] was performed. The result showed Cα and main chain root mean square deviations (RMSD) of 0.680 Å and 0.838 Å respectively for SjCP3842, a representative member of this gene family (Figure 1B). Similar low RMSD values were recorded for the other candidates. Ramachandran plot (φ/ψ) of conformation angles for each residue showed over 98% of the residues in the favored region, with less than 2% in the outlier region. These results indicate the reliability of the predicted models (Figure 1B).
A reciprocal ‘BackPhyre’ using the modeled structures to scan over 25 genomes also mapped the domain to SEA-domains at over 95% confidence, albeit with limited protein sequence homology. The low sequence similarity (Figure 1C) observed from alignments of this extracellular domain with two major SEA-domains (MUC1 and TMPRSS2) could imply that this structural similarity is at least partly independent of amino acid sequence homology [29]. As a matter of fact, SEA-domains are primarily defined by their characteristic folding pattern, extracellular localization on transmembrane proteins, their ability to assist or regulate binding to glycans, and their presence in proteins with O-linked glycans [28], [29], [41]. As expected, multiple O-glycosylation sites were identified by posttranslational modification prediction. We also confirmed that the expressed proteins contain O-linked glycans using glycoprotein detection assay (Figure S2). Equally, two conserved cysteine residues are present in all the candidates (Figure S1A), which could be structurally important by providing disulfide bridges in the folded protein.
Further evidence to classify the identified domain as SEA-module was the identification of the typical glycine-serine amino acid consensus (frpG/Svvv) [30] auto-cleavage site of SEA-domains (Figure 1C). Some SEA-domain proteins have been shown to undergo auto-cleavage, although the resulting subunits remain non-covalently associated in the native state [30], [41], [42]. This cleavage site is usually located within the bend between β2 and β3 sheets [30] as we equally observed (red arrow in Figure 1 A and C). In addition, the SDS fractionated recombinant protein (shown later) contained extra bands of expected molecular weight as the potential cleavage products. Taken together, these results provide multiple grounds to classify this extracellular domain as SEA-domain.
To provide lead to the possible molecular function of the gene products, we subjected the modeled structures to ligand binding site identification using 3DLigandSite [35]. This program uses protein structure to search a structural library to identify homologous structures with bound ligands, which are then superimposed on the protein structure to predict potential ligand binding sites [35]. Interestingly, a binding site was observed for Fe-protoporphyrin-IX (heme) at significantly high precision (Figure S3). Binding sites for energy transfer coenzymes including ATP, and several metal ions (Mg, Zn, Cu) binding sites were also identified. The heme-binding site was predicted based on 178 heme ligands present in 177 homologous structures with bound heme (Figure S3).
Analysis of the modeled heme-binding pocket of SjCP3842 showed that the vinyl end of the amphiphilic heme is inserted into a hydrophobic cavity created between α2 and α3 helices, and β2 and β3 sheets (Figure 2 A and B). Many of the interacting residues in the binding pocket are conserved among the members of this protein family (labeled in red in Figure S4B), consistent with binding of a heme group. The hydrophilic propionate end (red sphere) of heme is rather facing away from the hydrophobic pocket (Figure 2 A and B), with one propionate group engaged in electrostatic interactions with a nitrogen atom in Arg-157 side chain (Figure 2C). The phenyl rings of three conserved phenylalanine residues (Phe-80, Phe-140 and Phe-156) and one other phenylalanine (Phe-143) engage in pi-stacking interactions with the heme Pyrrole rings, which further stabilize heme-binding (Figure S4B). There were also polar contacts between heme and Thr-79, Tyr-83, His-147 and His-149 (Figure S4B), and several hydrogen bond interactions within the binding site.
Consistent with binding to heme, we readily identified potential axial ligands for heme iron, indicating hexa-coordination state involving two possible pairs. The imidazole group on His-149 side chain (bond distance of 2.0 Å) is the putative proximal ligand with either His-147 (Figure 2C) or the thioether group on Met-50 (Figure S4C) as the distal ligand of heme iron. However, the exact pair of axial ligands or the possibility of simultaneously binding two molecules of heme needs to be experimentally clarified. Similar binding site characteristics were observed in another characterized candidate (SjCP1531). However, the iron is coordinated to Tyr-154 as its axial ligand (Figure S5).
We investigated whether this gene family is differentially expressed among developmental stages of S. japonicum by stage specific mRNA expression using real time PCR. All other in-vitro based characterization was limited to three candidates: SjCP3842 [GenBank: AY570748], SjCP1084 [GenBank: AY570737] and SjCP1531 [GenBank: AY570742]. Relative expression of each candidate gene was quantified and expressed as copy number per nanogram of cDNA (Figure 3 and Table S2). There was differential expression of the three genes among developmental stages of the parasite, with SjCP3842 expressed at higher levels relative to the other two characterized candidates (Figure 3 and Table S2). SjCP3842 was overtly expressed in the adult stage (5680±370.9), although at a higher level in female adult worm (4846±302.1) as compared to the male worms (2000±453.9). The expression levels in the snail intermediate inhabiting sporocyst (2474±627.2) and infective cercaria (2871±98.4) stages were also relatively high as compared to somula (543.4±64.1) and egg stage (252±370.1). SjCP3842 was expressed at the minimal level in the egg stage (Figure 3A). Conversely, SjCP1084 was mainly expressed in the egg stage in relation to other stages. However, the expression levels of SjCP1531 in all stages of the parasite were relatively low and mainly expressed at the egg and adult stages (Figure 3C and Table S2).
To confirm expression at protein level, we expressed recombinant proteins, generated and used specific immune sera to identify the native proteins in parasite crude extracts. The complete coding regions of the genes were amplified from S. japonicum adult worm cDNA library and cloned into the expression vector, pcDNA4-HisMax. For recombinant protein expression, the plasmid constructs were transformed into Freestyle 293 and BL21 E. coli cells. The recombinant proteins used for biochemical assays were expressed in Freestyle 293 cells to ensure proper folding and post translational modification. The proteins were found to exist as oligomers in the native state as seen in the multiple bands of additive ∼30 kDa subunits observed both on SDS-PAGE (Figure 4A), western blots probed with anti-HisG antibody (Figure 4 B and C), and by multiple peaks from size exclusion chromatography fractions (Figure 4D), all showing the tetramer as the native state. Similar oligomeric state was also predicted by structural modeling (Figure S1D). Oligomerization may have been mediated by the disulfide bridges on two conserved cysteine residues common among the members of this family (Figure S1A). Other extra bands are of same molecular weight as the expected SEA-domain auto-cleavage products (Figure 1C).
To confirm native expression and to show potential antigenicity of the candidates during schistosomiasis, immunoblotting and ELISA techniques were applied. Parasite egg (SEA) and adult worm (SWA) crude antigen preparations were blotted and probed with the polyclonal immune sera (α-SjCP3842, α-SjCP1531 and α-SjCP1084). Blotted protein fractions of sizes similar to both the subunits (∼30 kDa) and tetramer (∼120 kDa) reacted specifically with the immune sera (Figure 4E). Also, the recombinant proteins specifically reacted with sera from S. japonicum infected miniature pigs, with significantly high titers of IgG in ELISA (Figure 4F). These results indicate that this gene family is actually expressed in the parasite, appear functional and potentially antigenic during schistosomiasis.
In addition to their characteristic folding pattern, SEA-domains are known to assist or regulate binding to carbohydrate moieties. We assessed interactions of the characterized SEA-domain proteins with glycans using recombinant proteins and array type sugar chips in a Surface Plasmon Resonance (SPR) system [37]. The SPR signal (expressed in resonance units, RU) is proportional to the amount of protein analytes bound to the sugar chains immobilized on the sensor chip in a 48 glycans array. The SPR imaging showed specific binding to sulfated GAGs with relatively high affinity. There was disproportionately high specific binding to chondroitin sulfate, dermatan sulfate (CS-B), heparin, dextran sulfate and other sulfated GAGs (Figure 5). SjCP1084 and SjCP1531 have similar glycan binding pattern while SjCP3842 showed relatively less glycan binding capacity but also preferentially binds sulfated GAGs (Figure 5).
We further confirmed the specificity and affinity of protein-GAG interactions by using chondroitin sulfate GAG (CS-GAG) chip containing all possible sulfated disaccharides subunits of chondroitin sulfate, and different concentrations of the protein as analytes. The glycan array format of the CS-GAG chip used and the SPR imaging of the glycan binding assays are shown in a supplementary file (Figure S6 A and B). The binding kinetics of the carbohydrate-protein interactions showed significant binding affinity to CS-GAGs, with dissociation constant (KD) within the range of receptor-ligand interactions (Figure S6 C and D). Figure S6C shows the detailed sensorgram and the binding curve of the interaction between SjCP1084 and chondroitin sulfate E (KD = 9.84×10−9 M), as representative of the binding kinetics data. The other KD values for the interactions of SjCP1084 and SjCP1531 with different sulfated disaccharides of chondroitin sulfate are summarized in Figure S6D, showing values within nanomolar range. These results indicate the specificity and affinity of the observed protein-glycan interactions.
To validate the structure based heme-binding model, we showed heme-binding properties of this family in-vitro, by three independent methods: hemin-agarose binding assay, heme-dependent peroxidase activity of protein-hemin complexes and optical UV absorption spectroscopy. First, we showed using SjCP3842 that the purified recombinant protein has potential to bind heme on hemin-agarose beads. The eluted fraction showed evidence of specific binding of the protein to heme (Figure 6A). Same experiment performed using unconjugated Sepharose 4B as negative control did not show any trace of the protein in the eluted fraction. Heme binding assay was repeated using the three characterized candidates and similar specific binding was consistently observed after immunoblotting using immune sera (Figure 6B).
To confirm this observation in the native state, hemin-agarose beads were incubated with parasite adult worm crude antigen (SWA) to isolate the total heme-binding protein fractions in the parasite. The fractions were blotted and probed with monoclonal antibody against SjCP3842 (Figure 6C). The result clearly showed the presence of the protein in the parasite heme-binding protein fractions. The multiple bands are expected molecular weights of the monomer, dimer and tetramer. The fact that binding was ablated by the reducing effect of β-mercaptoethanol and denaturing effect of sodium dodecyl sulfate (SDS) used for elution suggests that the observed heme-binding property is at least partly non-covalently mediated by structure of the folded proteins.
To estimate the amount of heme bound by the protein, we assayed the heme-dependent peroxidase activity of the protein-hemin complex using SjCP3842. We first estimated the peroxidase activities of known concentrations of hemin, and used the resulting standard curve (linear graph) to estimate the amount of heme bound by the characterized heme-binding protein based on the peroxidase activity of bound heme (Figure 6D). The result showed that the amount of bound heme increased with increasing protein concentration, reaching saturation at about 2 µg of protein, when 1 µg of hemin was bound (Figure 6D).
To further assess the binding affinity of the protein-heme interaction, optical absorption spectra of the protein-heme complex was monitored by differential titration of 10 µM of heme with increasing concentrations of the protein (0 to 28 µM) (Figure 6E). The Soret absorption peak for heme alone was characteristically broad and was initially 388 nm prior to addition of the protein (broken lines). The Soret absorption maximum was red shifted to 412 nm on addition of protein and absorbance at this peak increased gradually depending on accumulation of protein-heme complex, until saturation at about 1∶1 molar ratio. The Q-bands (534 nm and 564 nm) and the isobestic points were also apparent, indicating the presence of two absorbing species (heme and protein-heme complex) in the solution. The UV-visualization spectral attributes of the protein-heme complex (Soret peak, 412 nm; Q-bands, 534 nm and 564 nm) were typical of heme with hexa-coordinated ferric iron [40], [43], consistent with the structural model of this study. However, this needs to be confirmed by electron spin resonance spectroscopy. The inset is the heme-binding curve constructed by plotting ΔA412 versus protein concentration (Figure 6E). The curve fitting indicates increasing accumulation of the protein-heme complex with saturation after about 10 µM of protein was added, thus suggesting a 1∶1 stoichiometry. The fitting yielded equilibrium dissociation constant KD = 1.605×10−6 M, indicating high affinity for binding heme. Taken together, these observations confirm the potential of the novel SEA-domain proteins to specifically interact with heme.
To ascertain the tissue distribution of the products of this gene family in the parasite, immunolocalization was performed by immunofluorescence assay (IFA) and immunoperoxidase staining. For clarity and because similar tissue localization patterns were observed, only the data for SjCP3842 is shown here. The results for the other candidates are presented in a supplementary figure (Figure S7). IFA on adult worm sections showed that the native SjCP3842 was localized on the adult worm tegument and gastrodermis of the parasites gut (Figure 7 A and D). Similar results were observed for all the three candidates as presented in a supplementary figure (Figure S7). No signal was observed in the ovary as shown in the cross section of the female adult worm probed with anti-SjCP3842 monoclonal antibody (Figure 7 B and E), which is consistent with minimal expression in the egg as earlier shown in the developmental stage specific gene expression (Figure 3A). The nuclei are stained with DAPI, showing staining both in the parasite tissues and the content of the ovary. No signal was observed in the sections incubated with sera obtained from control mice (Figure 7 C and F).
Equally, immunolocalization was repeated using immunoperoxidase-DAB technique with biotinylated monoclonal antibody detected with streptavidin-HRP. The result again showed localization on the adult worm teguments (Figure 7 G and H). The protein was also found localized on the tegument of the juvenile schistosomula stage (Figure 7I). No peroxidase activity was detected in the sections probed with pre-immune serum (Figure 7 J–L). Taken together, these results indicate localization on adult worm teguments and gastrodermis, and schistosomula teguments.
We have utilized comparative homology modeling to identify remote structural homologs, and successfully characterized a novel gene family encoding SEA-domain proteins from S. japonicum. Similar strategy was used to identify and characterize heme-binding property for this domain, thereby providing insight into the potential biological function of otherwise ‘hypothetical proteins’. Functional annotation of proteins routinely relies on sequence homology with already characterized proteins or at least domains with experimentally resolved functions. However, the degree of evolutionary conservation of the structural architecture of proteins is greater than the amino acid sequence conservation [24], [25]. Our results affirmed that absolute reliance on sequence homology for functional annotation of proteins is not exhaustive. In the post-genome era, the vast accumulation of sequence data has opened new frontiers for identification of intervention targets. However, determination of protein functions is one of the major challenges since sequence homology alone has proven insufficient for placing the vast amount of ‘omics’ data into functional context [24], [25]. It is necessary to explore other strategies that can effectively identify remote homologs, which are not readily identifiable from sequence data. The data presented here is a typical example of the possible application of molecular structural analysis to identify and characterize novel protein functions.
Like most previously characterized SEA-domain containing proteins, our candidates specifically interacted with sugar chains, especially glycosaminoglycans (GAGs) [28], [41]. GAGs are long linear polysaccharides composed of repeating disaccharide units, usually linked covalently to a core protein to form a proteoglycan. While the protein core keeps the proteoglycan localized on the cell surface or in the extracellular matrix (ECM), the GAGs components mediate interactions with a plethora of extracellular ligands and effectors. All cellular processes that involve cell surface molecular interactions including: ligand-receptor, cell-cell and cell-matrix interactions, will likely involve proteoglycans and GAGs because these molecules are ubiquitous and are shown to functionally bind proteins to regulate important developmental processes [44]–[46].
In addition to their space filling and organizational roles in the ECM, GAGs on proteoglycans can modulate the function of a repertoire of extracellular effectors by their roles in: ligand gathering, clustering and oligomerization of ligands and their receptors [45], [46], and their ability to act as storage depots for ligands by sequestering them and preventing their rapid degradation [45]. Proteoglycans are required as co-receptors for some growth factors and cytokines signaling in collaboration with the cognate signaling receptors in a ligand-receptor-proteoglycan ternary complex [45]–[47], and can also signal independently as a receptor via its cytoplasmic domain [47], [48]. Proteoglycans can also undergo proteolytic cleavage near the plasma membrane to shed their ectodomain as soluble regulators [49]. Specific interaction with GAGs of host (trans) or parasite (cis) origin as we observed here may suggest some functional role of this protein family as parasite receptors for accessing ligands and signals, especially of host origin. From the foregoing, and given that S. japonicum genome encodes many receptors and signaling molecules but sometimes not the ligands [21], it is plausible that parasite membrane receptors with GAG-binding potential could interact with its own or host proteoglycans in a receptor-proteoglycan-ligand ternary complex [45]–[47], as a means of accessing host molecules tethered on GAGs for signals for their growth, development, and maturation thus rendering them potential intervention targets.
The native proteins were localized at the parasite tegument and gastrodermis, sites that are of immunological significance being located at the host-parasite interface [14], [16]. These sites are rich in proteins that are often unique to schistosomes, some of which can directly interact with host derived molecules as observed in the characterized SEA-domain proteins [14], [16]. The ability of the parasite to bind GAGs on host secreted or shed proteoglycans [49] or proteoglycans on the surface of host immune cells [50] could result in masking of the ‘non-self’ status of the parasite, thereby evading attack by host immune system [51]. It is thus possible from the foregoing, that this gene family could also be involved in some immune evasion mechanisms. We are presently targeting the candidates that are expressed at the infective cercarial, schistosomula and adult stages for possible vaccine application.
Heme-binding properties have been described here for the first time for SEA-domain proteins from this hemophagous parasite. In terms of the parasite biology and host-parasite interaction, this finding represents a significant contribution towards elucidating heme detoxification and heme iron acquisition mechanisms of the parasite. Schistosomes inhabit the hepatoportal veins of the host, where they feed on host erythrocytes and catabolize the globin moieties of hemoglobin as a major source of the requisite amino acids for their growth, development and reproduction [3], [4]. However, the released heme moiety is potentially toxic due to its reactive nature and ability to produce free radical species, lipid peroxidation, and protein and DNA oxidation [3], [6]. Hemophagy-adapted parasites have therefore evolved strategies to sequester and detoxify heme [3]–[9]. Heme iron is arguably the major source of iron for this parasite, thus, the parasite also maintains a heme acquisition mechanism to harness the needed iron from heme molecules [4], [10]. These candidates are localized on adult worm gastrodermis, the site for heme detoxification [8], [17] and acquisition [3], [10]; and in the adult and schistosomula teguments, also potential sites for heme acquisition in these stages [20]. Indeed, effective heme homeostasis mechanism is paramount for parasite survival and establishment, and is a major target of effective drugs against hemoparasites including the quinines and artemisinine [11]–[13]. Unfortunately, the exact mechanisms and the molecules involved in heme-homeostasis are still controversial. However, there is a consensus on the involvement of heme-binding proteins both as nucleation agents for heme crystallization [6]–[8], [17], and as surface heme receptors in an ABC- (ATP binding cassette) transporters coupled heme uptake mechanism [38], [52], [53].
The developmental stage specific expression, especially of SjCP3842, clearly showed overt expression at the adult stage especially the female adult worms, which is consistent with the heme homeostasis requirements of this stage. There was also relatively high expression in the snail inhabiting sporocysts and the infective cercariae. The observation that the sporocysts also express this gene indicates expression at the snail stage as well, which may suggest similar or different function in the snail host. With regards to heme binding function, the sporocysts are known to absorb nutrients from snail host through their tegument for nourishment of cercariae in their germinal sac [54], and heme binding proteins have also been identified among secreted proteins from the sporocyst stage [55]. Since iron source in snail is mainly in the form of heme, it is plausible that heme binding proteins like the ones we characterized might be required for heme iron uptake from snail hosts, as well as other functions. SEA domain still do not have a well characterized function apart from interaction with glycans (GAGs), to which we and others have alluded several potential implications like ligand acquisition and immune evasion. The prospect that this gene family could perform more than one function in different developmental stages of the parasite implies that hemophagy might have been a major factor among other selection factors for this gene family.
SEA-domains are characteristically found in carbohydrate rich mucous environments [30]. The heme-binding SEA-domain proteins we described here are localized in the parasite gastrodermis and tegument. The gastrodermis is the syncytial linings of the parasite gut, the site for hemoglobin catabolism, heme sequestration, detoxification and acquisition. A similar structure called peritrophic matrix (PM) with heme-binding property has been described in the midgut of hemophagous insects. The PMs perform a central role in heme homeostasis by protecting the insects' midgut against damage from heme toxicity [56], akin to schistosomes gastrodermis. The PMs are complex matrices composed of heme-binding proteins, proteoglycans, chitins and chitin-binding proteins [56]. Specifically, Aedes aegypti Mucin I (AeMUC1) was identified as a major heme-binding protein in the PM [57]. MUC1 and the proteins we characterized here both contain SEA-domains. It is therefore plausible that similar mechanism mediated by heme-binding SEA-domain proteins may exist in schistosomes' gastrodermis. However, this hypothesis will need to be experimentally clarified by isolating and identifying all heme-binding proteins of the parasite and/or the parasite gastrodermis. We will design further studies to fully characterize the role of this gene family in the parasite heme-homeostasis and heme acquisition mechanisms, and explore prospects for its application in disease intervention.
|
10.1371/journal.ppat.1006901 | Picornavirus 2A protease regulates stress granule formation to facilitate viral translation | Stress granules (SGs) contain stalled messenger ribonucleoprotein complexes and are related to the regulation of mRNA translation. Picornavirus infection can interfere with the formation of SGs. However, the detailed molecular mechanisms and functions of picornavirus-mediated regulation of SG formation are not clear. Here, we found that the 2A protease of a picornavirus, EV71, induced atypical stress granule (aSG), but not typical stress granule (tSG), formation via cleavage of eIF4GI. Furthermore, 2A was required and sufficient to inhibit tSGs induced by EV71 infection, sodium arsenite, or heat shock. Infection of 2A protease activity-inactivated recombinant EV71 (EV71-2AC110S) failed to induce aSG formation and only induced tSG formation, which is PKR and eIF2α phosphorylation-dependent. By using a Renilla luciferase mRNA reporter system and RNA fluorescence in situ hybridization assay, we found that EV71-induced aSGs were beneficial to viral translation through sequestering only cellular mRNAs, but not viral mRNAs. In addition, we found that the 2A protease of other picornaviruses such as poliovirus and coxsackievirus also induced aSG formation and blocked tSG formation. Taken together, our results demonstrate that, on one hand, EV71 infection induces tSG formation via the PKR-eIF2α pathway, and on the other hand, 2A, but not 3C, blocks tSG formation. Instead, 2A induces aSG formation by cleaving eIF4GI to sequester cellular mRNA but release viral mRNA, thereby facilitating viral translation.
| When cellular translation initiation is stalled, translation initiation complexes aggregate in cytoplasm. We call these aggregations stress granules (SGs), and they can be marked by components such as TIA-1. SGs are always considered to be antiviral structures during viral infection, but viruses also regulate SG formation to facilitate their survival. Here, we show that the 2A protease of EV71 induced TIA-1 foci formation, and we analyzed these TIA-1 foci and found that they were different from typical stress granules (tSGs); thus, we named them atypical stress granules (aSGs). 2A alone could block tSG formation, and we found that protease activity of 2A was required for aSG induction and tSG blockage, but functioned in different ways. When the protease activity of 2A in EV71 was blocked (EV71-2AC110S), the tSGs but not aSGs appeared in infected cells. These tSGs contained cellular and viral mRNAs and translation initiation factors to inhibit viral translation, but aSGs contained only cellular mRNAs to promote viral translation. We propose a model revealing that EV71 escapes cellular antiviral response by manipulating SG formation: 2A transforms the overall translation shutdown system to a selective virally beneficial system by switching from tSGs to aSGs.
| Stress granules (SGs) form in response to a variety of stresses such as oxidative stress, heat shock (HS), hypoxia, nutrient deprivation, and viral infection [1]. During SG formation, messenger RNA (mRNA) translation initiation is inhibited, and polysomes are disassembled. Thus, SGs contain stalled pre-initiation complexes (PICs) consisting of translationally silent mRNAs, 40S ribosomal subunits, canonical eukaryotic initiation factors (eIFs) such as eIF4E, eIF4G, eIF4A, eIF4B, and eIF3, and RNA-binding proteins (RBPs). Two aggregation-prone RBPs, T-cell-restricted intracellular antigen 1 (TIA-1) and the RasGAP SH3-domain binding protein 1 (G3BP), appear to be critical for SG formation and are recruited to SGs [2,3].
SGs in general are considered to be transient and dynamic. Compounds such as cycloheximide (CHX) stabilize mRNAs on polysomes and inhibit SG formation and foster their disassembly [4,5]. The SG formation induced by the phosphorylation of eIF2α is well characterized. The protein kinase R (PKR), PKR-like ER kinase (PERK), general control nonderepressible 2 (GCN2), or heme regulated inhibitor (HRI) can phosphorylate eIF2α under different stress conditions. For example, PKR can be activated by viral dsRNA, and HRI can be activated by arsenite (AS) or HS. The phosphorylation of eIF2α interferes with the formation of the eIF2-GTP-tRNAiMet ternary complex and thereby stalls translation initiation [1,6–8]. However, eIF2α-independent SG formation also exists and includes eIF4A inhibition by either pateamine A or hippuristanol and inhibition of eIF4G-eIF4E interactions during hydrogen peroxide-induced oxidative stress [9–12].
Thus, SG formation can be caused by a variety of mechanisms that impair translation initiation. Although SGs formed in response to diverse stresses share many of the same components, certain factors appear to be recruited in a stress-specific fashion. For example, HS protein 27 (HSP27) is found in SGs in HS cells but not in cells undergoing oxidative stress [13]; the p68 src-associated protein in mitosis (Sam68) is recruited to SGs in picornavirus-infected cells but not to those formed in response to oxidative stress or HS. Thus, SGs may be compositionally different depending on the type of stress [14], and distinct SGs may be regulated differentially and have multiple roles. However, the mechanisms by which SGs form have not been identified completely.
Enterovirus 71 (EV71), a member of the Picornaviridae family, is widely spread and causes severe hand-foot-mouth disease in infants [15]. Thus, understanding the host factors that influence viral pathogenesis is critical to designing improved antiviral strategies. The EV71 genome (∼7.5 kb) can be immediately translated into a single polyprotein via an internal ribosome entry sequence (IRES)-mediated, cap-independent mechanism of translation initiation, and this polyprotein is subsequently processed by proteases 2A and 3C into the structural and nonstructural proteins [16]. Furthermore, the IRES can drive the viral genome translation in the absence of functional eIF4F complex, which is disrupted due to cleavage of eIF4G in picornavirus-infected cells [17,18].
SGs are thought to be antiviral, and many viruses have hence evolved various strategies to disrupt SG formation to maintain efficient translation of their proteins and to prevent their genomes and transcripts from being stalled in SGs [19,20]. Poliovirus (PV) infection was initially indicated by the recruitment of HuR and G3BP to SGs in early phases [10]. Subsequently, White et al. found that eIF4G and polyA-binding protein (PABP) were also recruited to SGs early in PV infection [21]. However, further examination revealed that eIF4G, G3BP, and PABP were no longer found in SGs at later times, indicating that PV may actively disrupt SG formation at later times [21]. Furthermore, the discovery that G3BP was cleaved by 3C, which coincided with SG disassembly in infected cells, provided a possible explanation for these findings [21]. White et al. suggested a model whereby PV initially induces SG formation but induces SG disassembly at later stages via cleavage of G3BP, thus preventing SG formation even in the presence of external stress [21].
However, because most of the SG markers used in the study by White et al. can be cleaved by 2A or 3C, whether other SG components are also released from SGs is unclear. Therefore, Piotrowska et al. further examined SG formation in PV-infected cells and found that infection induced stable, compositionally unique SGs containing TIA-1 but lacking G3BP and eIF4G and that these SGs did not disassemble at late times in infected cells, which raised the possibility that PV might not induce the complete disassembly of SGs. As G3BP, eIF4G, and PABP are all cleaved by viral proteases, proteolysis may trigger their release from SGs. Alternatively, the release may be triggered by a mechanism not yet identified [14]. Furthermore, recent studies showed that 2A of picornaviruses induced SG formation and that no other viral proteins induced SG formation [22]. However, the molecular mechanism of these 2A-induced SGs and the cellular components that 2A targets to trigger SG formation remain unknown.
Therefore, several questions raised in the aforementioned studies have not been resolved. (1) How do picornaviruses induce SG formation at early times and block SG assembly at later times? (2) Is cleavage of G3BP by 3C critical for the inhibition of SG formation at later times during infection? (3) What is the molecular mechanism of 2A-induced SG formation? Are they typical stress granules (tSGs)? (4) What are the organization and role of 2A-induced SGs in picornavirus-infected cells?
In this study, we demonstrate that the 2A protease of EV71 blocks tSG formation but induces atypical stress granule (aSG) formation to facilitate viral translation. These aSGs are induced by cleavage of eIF4GI and are different from tSGs in that they are devoid of G3BP and a series of eIFs, they are independent of eIF2α and PKR phosphorylation, they cannot be dissolved by CHX, and they can specifically sequester cellular mRNAs but not viral mRNAs. On the other hand, infection with a 2A protease activity-inactivated recombinant virus, EV71-2AC110S, induces tSG formation via the PKR-eIF2α pathway, and these tSGs are antiviral structures. These findings provide a new conceptual mechanism for SG regulation during picornavirus infection.
To explore the formation of SGs during picornavirus infection, we infected HeLa cells with EV71 for 6 hours (h) and visualized SGs via immunofluorescence (IF) with antibodies against Sam68, TIA-1, and G3BP. EV71 infection led to the formation of SGs containing Sam68 and TIA-1, but devoid of G3BP, and as a control, AS induced the formation of tSGs containing TIA-1 and G3BP, but not Sam68 (Fig 1A).
To explore EV71-induced SG formation in more detail, we tried to clarify the dynamics of SG assembly during EV71 infection and visualized SGs at different time points post-infection (pi). An antibody against EV71 was used to visualize EV71-infected cells, and TIA-1 and Sam68 were also visualized. When cells were mock- or EV71-infected for 2 hours, EV71 infection was undetectable via IF, and localization of TIA-1 and Sam68 did not change. As infection proceeded, IF revealed EV71 in 90% of the cells at 4 hpi and 95% of the cells at 6 hpi, TIA-1 and Sam68 assembled into foci and colocalized with each other in infected cells (S1A and S1H Fig). Therefore, the appearance of TIA-1 and Sam68 foci can be used as indicators of EV71 infection. However, although TIA-1-related protein (TIAR) aggregated and persisted in infected cells with TIA-1 and Sam68, the other tSG markers such as G3BP, PABP, 40S ribosomal protein S3 (RPS3), eIF1a, eIF3a, eIF4G, eIF4A, and eIF4E aggregated in less than 30% of cells at 4 hpi and were evenly distributed at 6 hpi (S1B–S1H Fig). Similar results were observed in EV71-infected rhabdomyosarcoma (RD) cells (S1H Fig), suggesting that the persistent SGs containing TIA-1, TIAR, and Sam68 might not be tSGs.
To confirm which protein of EV71 induced the formation of persistent SGs, we expressed all the viral proteins in HeLa and RD cells and found that only 2A protease induced the formation of TIA-1 foci (S2A Fig), which also contained Sam68 and TIAR but did not contain PABP, G3BP, RPS3, eIF1a, eIF3a, eIF4G, eIF4A, or eIF4E (Fig 1B and S2B Fig), suggesting that 2A expression alone is sufficient to trigger the formation of persistent SGs during EV71 infection and the formation of Sam68 or TIA-1 foci can also be used as an indicator of 2A expression.
Next, we sought to determine whether 2A-induced persistent SGs share features with tSGs. First, a previous study showed that tSGs are dynamic structures of which TIA-1 rapidly shuttles in and out [23]. Thus, we analyzed the dynamics of TIA-1 in tSGs and persistent SGs via fluorescence recovery after photobleaching (FRAP) assay. To rule out the impact of size, typical and persistent SGs with similar size (1.5–2 μm in dimeter) were analyzed. We found that GFP-TIA-1 in typical and persistent SGs were rapidly recovered (Fig 1C), suggesting that persistent SGs are also dynamic structures. Second, previous studies also showed that stresses not only induce tSG formation but also induce adjacent processing body (p-body) formation [24–26]. Thus we also analyzed the location of the p-bodies in EV71-infected cells by using a widely used marker of p-bodies—mRNA-decapping enzyme 1A (DCP1A), and found that p-bodies were adjacent to EV71-induced persistent SGs at 4h post-infection, but disappeared at 6h post-infection, suggesting that EV71-induced persistent SGs were distinct from p-bodies but similar to tSGs (S2C and S2D Fig). Third, we evaluated the effect of eIF2α phosphorylation on the formation of TIA-1 foci and found that expression of eIF2αS51A (an eIF2α non-phosphorylated mutant) [27–29] blocked AS-induced tSG formation but had no effect on EV71- and 2A-induced formation of persistent SGs (Fig 1D and 1E), indicating that 2A induced the formation of persistent SGs in a phospho-eIF2α-independent manner. Fourth, we found that CHX dispersed AS-induced tSGs but had no effect on EV71- and 2A-induced persistent SGs (Fig 1F and 1G). Therefore, we defined EV71- and 2A-induced persistent SGs as atypical stress granules (aSGs).
Next, we sought to determine the role 2A plays in aSG formation. Since 2A is an important viral protease, we determined whether 2A protease activity is required for the formation of aSGs. Previous studies showed that a 2A mutant, 2AC110S, was catalytically inactive [30,31]. We also found that 2AC110S indeed lost the ability to cleave eIF4G (Fig 2A); subsequently, 2AC110S also lost the ability to induce aSG formation (Fig 2B and 2C), demonstrating that 2A protease activity is essential for aSG formation.
Furthermore, based on three facts, (1) both eIF4G and PABP were cleavage substrates of 2A and excluded from 2A-induced aSGs; (2) both eIF4GI and eIF4GII play similar functions in translation initiation, but cleavage of eIF4GI by 2A of picornaviruses is more sensitive than cleavage of eIF4GII [32,33]; (3) the mRNA level of eIF4GI was previously shown to be much higher than that of eIF4GII in mammalian cells [34], therefore, we hypothesized that cleavage of eIF4GI/PABP by 2A might be critical for 2A-induced aSG formation.
To this end, HeLa cells transiently expressing eIF4GIG689E (a 2A cleavage-resistant eIF4GI mutant) [35] or PABPM490P/Q540N (a 2A and 3C double-resistant mutant) [36,37] were infected with EV71 or treated with 2A. We found that eIF4GIG689E and PABPM490P/Q540N were indeed resistant to 2A cleavage (Fig 2D and S3A Fig), and eIF4GIG689E expression dramatically blocked EV71- or 2A-induced aSG formation (Fig 2E and 2F, cells marked by “yellow arrow” and “+” indicate eIF4GIG689E expression with EV71 infection or 2A expression), whereas PABPM490P/Q540N expression had no effect on the formation of aSGs (S3B and S3C Fig), suggesting that cleavage of eIF4GI is critical for the formation of 2A-induced aSGs.
Because EV71 induced the formation of aSGs that were distinct from tSGs, we sought to determine whether tSG formation could be blocked during EV71 infection. We infected HeLa cells with EV71, treated them with AS or HS for 1 h prior to fixation at the indicated times, and stained them with antibodies against Sam68 (a marker of EV71-infected cells), TIA-1 (a marker of both tSGs and aSGs), G3BP (a marker of tSGs), or HSP27 (a marker of HS-induced tSGs). When cells were mock- or EV71-infected for 2 h, tSGs marked by G3BP and HSP27 were observed in all the cells upon treatment with AS or HS. As infection proceeded, tSGs were observed in only 20% of the cells at 4 hpi and 10% of the cells at 6 hpi, despite treatment with AS or HS (Fig 3A and 3B); similar results were also observed in EV71-infected RD cells (S4A and S4B Fig), suggesting that EV71 infection blocks the formation of AS- or HS-induced tSGs. However, EV71 infection had no effect on the AS- or HS-activated phosphorylation of eIF2α (S4C Fig), indicating that blockage of tSG formation by EV71 infection is not due to inhibition of eIF2α phosphorylation.
Next, we sought to determine how EV71 blocks tSG formation. Previous studies reported that 3C protease of picornaviruses inhibited tSG formation by cleavage of G3BP, and G3BPQ326E (a 3C cleavage-resistant mutant of G3BP) restored SG formation competency in picornavirus-infected cells [21,38]. We confirmed that GFP-G3BPQ326E was indeed cleavage-resistant upon EV71 infection (Fig 3C) or 3C expression (S4D Fig, left panel), and GFP-G3BPQ326E rescued tSG formation in 3C-expressing cells (S4D Fig, right panel). To determine whether EV71 inhibits tSG formation in a manner similar to that of previous reported picornaviruses, we assessed tSG formation in the presence of EV71 infection, GFP-G3BP or GFP-G3BPQ326E expression plus AS. At 2 hpi with EV71, GFP-G3BP and GFP-G3BPQ326E localized to AS-induced tSGs. As infection proceeded, AS-induced tSG formation was blocked in EV71-infected cells despite expression of GFP-G3BPQ326E (Fig 3D and 3E). Similar results were observed in EV71-infected RD cells expressing GFP-G3BP or GFP-G3BPQ326E (S4E Fig), suggesting that 3C cleavage of G3BP is dispensable for the inhibition of tSG formation during EV71 infection.
To confirm whether 3C protease is required for the inhibition of tSG formation during EV71 infection, we used guanidine hydrochloride (GuHCl, an ATPase inhibitor) to suppress viral replication to an extremely low level in HeLa and RD cells [39,40]. Under these conditions, viral proteins were undetectable, and the level of full-length G3BP in EV71-infected cells was comparable with that in uninfected cells, but eIF4G was remarkably cleaved (Fig 3F and S4F Fig), as reported previously [21,41], indicating that 3C cannot work and the cleavage of eIF4G can be used to distinguish the infected cells. However, the AS-induced tSG formation was still inhibited in EV71-infected cells (Fig 3G and 3H and S4F Fig). Taken together, our results demonstrate that the blockage of tSG formation during EV71 infection is not due to 3C protease.
Since 3C was dispensable for the blockage of tSG formation, and 2A-induced aSGs did not contain tSG components such as G3BP and eIF4G, we hypothesized that 2A plays a critical role in inhibiting tSG formation. To validate this possibility, we examined tSG formation induced by AS or HS in the presence of 2A (Sam68 foci indicate 2A-expressing cells) and found that AS- or HS-induced tSGs (marked by G3BP or HSP27) appeared in 96% of empty vector-transfected cells but appeared in only 35% of 2A-transfected cells at 12 h and in less than 20% of 2A-transfected cells at 24 h (Fig 4A–4C). 2A-expressing cells all failed to form tSGs (Fig 4A and 4B, “+” indicates expression of 2A). Similar results were observed in 2A-transfected RD cells (S5A and S5B Fig). Furthermore, 2A expression had no effect on AS- or HS-activated eIF2α phosphorylation (S5C Fig), suggesting that 2A blocks tSG formation without inhibiting eIF2α phosphorylation, which is consistent with those in EV71 infection conditions (S4C Fig).
To confirm that 2A indeed plays a critical role in the inhibition of tSG formation during EV71 infection, we expressed Myc-tagged 2A in HeLa cells stably expressing GFP-G3BPQ326E and then treated cells with AS and found that 2A, but not 3C, blocked the formation of tSGs (S5D and S5E Fig), suggesting that 2A, but not 3C, plays a critical role in the inhibition of tSG formation during EV71 infection. In addition, we found that 2AC110S also lost the ability to block the formation of AS-induced tSGs (Fig 4D and 4E), suggesting that 2A protease activity is essential for its blockage of tSG formation. Having found that eIF4GIG689E blocked aSG formation, we further analyzed whether eIF4GIG689E could restore tSG formation in the presence of AS. To our surprise, we found that 2A still blocked tSG formation in spite of expression of eIF4GIG689E (S5F and S5G Fig), suggesting that the blockage of tSG formation is not due to cleavage of eIF4GI by 2A.
To further elucidate the critical role of 2A protease activity in blocking tSG formation and inducing aSG formation during EV71 infection, we generated recombinant EV71 with the 2AC110S mutation (EV71-2AC110S). Since 2A is important for the viral life cycle, the replication activity of EV71-2AC110S is much lower (50-fold less in viral titer) than that of EV71 and VP1 expression of EV71-2AC110S was also much lower than that of EV71 (S6A Fig). Correspondingly, EV71-2AC110S could not shut off cellular translation as quickly as EV71 (S6B Fig).
Therefore, we infected cells with a higher MOI of EV71-2AC110S and a lower MOI of EV71 and found that when the 3C protein levels and G3BP cleavage levels were comparable, eIF4G was no longer cleaved in EV71-2AC110S-infected cells (Fig 5A). Correspondingly, EV71-2AC110S infection failed to induce aSG formation; instead, EV71-2AC110S induced the formation of tSGs containing G3BP and TIA-1 in about 65% of infected cells, which could be completely dispersed by CHX (Fig 5B and 5C). With additional AS or HS treatment, all the EV71-2AC110S-infected cells formed tSGs (Fig 5B and 5D and S6C and S6D Fig). Taken together, these data demonstrate that EV71 infection induces tSG formation independent of 2A protease activity, but 2A inhibits EV71-induced tSG formation and induces aSG formation.
We next sought to determine how EV71 infection induces tSG formation. Since EV71 is a positive-sense single-stranded RNA virus, it generates significant amounts of viral replication intermediate dsRNAs during replication. And the dsRNAs commonly activate PKR to phosphorylate eIF2α, which results in tSG assembly [42]. To determine whether EV71-2AC110S induces tSG formation via the PKR-eIF2α pathway, we generated HeLa cells with stable knockdown (KD) of PKR (shPKR-HeLa cells) and infected them with EV71 or EV71-2AC110S. We found that KD of PKR decreased the phosphorylation levels of PKR and eIF2α in EV71- or EV71-2AC110S-infected cells (Fig 6A). In shPKR-HeLa cells, the formation of EV71-induced aSGs was not affected, but the formation of EV71-2AC110S-induced tSGs was blocked (Fig 6B and 6C). Furthermore, expression of eIF2αS51A blocked the formation of EV71-2AC110S-induced tSGs but not EV71-induced aSGs (Fig 6D and 6E). Taken together, these results suggest that, unlike aSGs, tSGs are induced via the PKR-eIF2α pathway by viral dsRNAs during EV71 infection.
Next, we sought to determine why 2A blocks tSG formation but induces aSG formation. We hypothesized that it is a strategy by which the virus facilitates its own translation. Thus, we generated a Renilla luciferase mRNA reporter, the translation of which is driven by EV71-UTR to mimic EV71 translation (UTREV71-Rluc) (Fig 7A). We found that eIF4GIG689E did not influence translation of UTREV71-Rluc in EV71-2AC110S-infected cells, but once aSG formation was inhibited by eIF4GIG689E in EV71-infected cells, the translation efficacy of UTREV71-Rluc was dramatically decreased (Fig 7B), suggesting that aSG formation benefits EV71 translation. Conversely, the translation efficacy of UTREV71-Rluc was not influenced by KD of PKR in EV71-infected cells, but increased at least three-fold when tSG formation was inhibited by KD of PKR in EV71-2AC110S-infected cells (Fig 7C), suggesting that tSGs inhibit EV71 translation.
Next, we sought to determine how aSGs benefit viral translation. We hypothesized that different mRNAs are sequestered in aSGs and tSGs to regulate viral translation. First, using poly(A) fluorescence in situ hybridization (FISH) assays, we confirmed that numerous mRNAs were present in the EV71- or 2A-induced aSGs and in the AS-induced tSGs (Fig 7D). Second, using RNA FISH assays, we evaluated the localization of viral mRNAs (positive-strand RNA, +vRNA) and cellular PABPC1 mRNAs (with high TIA-1 affinity) [43] in HeLa cells stably expressing GFP-TIA-1. In EV71-infected cells, PABPC1 mRNAs, but not +vRNA, efficiently localized in the aSGs (Fig 7E, top panel). In EV71-2AC110S-infected cells, both PABPC1 mRNAs and +vRNA efficiently localized in the tSGs (Fig 7E, bottom panel). Furthermore, we generated a Renilla luciferase mRNA reporter, the translation of which is driven by PABPC1-UTR to mimic PABPC1 translation (UTRPABPC1-Rluc) (S7A Fig). We found whether the aSGs was blocked by eIF4GIG689E in EV71-infected cells or the tSGs was blocked by KD of PKR in EV71-2AC110S-infected cells, the translation efficacy of UTRPABPC1-Rluc mRNA increased (S7B and S7C Fig), suggesting that both aSGs and tSGs inhibit PABPC1 mRNA translation. Taken together, our results suggest that tSGs stall both cellular and viral mRNAs to shut down overall translation, resulting in the inhibition of viral translation; however, 2A of EV71 blocked tSG formation but induced aSG formation to facilitate viral translation by stalling only cellular mRNAs.
The function of the 2A protease is highly conserved among Picornaviridae. Thus, we sought to determine whether its role in SG formation regulation is conserved among picornaviruses. We used Sam68 cytoplasmic re-localization as a marker of 2A expression in IF assays. Indeed, expression of 2A of EV71-BrCr, PV, and coxsackievirus A (CVA) also triggered Sam68 and TIA-1 to form aSGs that were devoid of G3BP and resistant to CHX (Fig 8A, upper panel, and Fig 8B). Furthermore, 2A of these picornaviruses also blocked AS- and HS-induced tSG formation (Fig 8A, lower panel, and Fig 8C and 8D). Similarly, 2A of these picornaviruses cleaved eIF4GI but not eIF4GIG689E (S8A Fig), and 2A of these picornaviruses also failed to induce the formation of aSGs in eIF4GIG689E-expressing cells (S8B and S8C Fig). Taken together, these findings suggest that although different picornaviruses may use different mechanisms to alter host cell function, the effect of 2A on the regulation of SGs is common.
In previous studies, TIA-1 foci have been observed during picornavirus infection; however, the essence of these aggregates was ambiguous and controversial. The 2A-induced TIA-1 foci were thought to be tSGs [22]. We also found that 2A-induced TIA-1 foci share some features with tSGs, such as they are dynamic and adjacent to p-bodies (Fig 1C and S2C Fig), but they are different from tSGs and thus named them aSGs. The conclusion of aSGs different from tSGs is supported by multiple lines of evidence. First, the aSGs contained TIA-1, TIAR, and Sam68 but were devoid of G3BP, a series of eIFs and viral mRNA (S1A–S1G and S2B Figs and Fig 7E). Second, the formation of aSGs was independent of PKR activation (Fig 6B and 6C) and the phosphorylation of eIF2α (Fig 1D and 1E). Third, the aSGs could not be disassembled by CHX (Fig 1F and 1G). Fourth, EV71-2AC110S-induced TIA-1 foci were tSGs, which contained TIA-1, G3BP, eIF4G, and viral mRNA, and could be completely dispersed by CHX, and the formation of which was dependent on PKR activation and the phosphorylation of eIF2α (Figs 5B, 6B and 7E).
Although both eIF4G and PABP were cleavage substrates of 2A, the cleavage efficiency of PABP is much lower than that of eIF4G [44]. We found that the only eIF4GIG689E blocked the aSGs formation, indicating that 2A induces aSG formation by cleaving eIF4GI (Fig 2E and 2F and S3A–S3C Fig). PABP is also a cleavage substrate of 3C, but expression of 3C is unable to induce aSG formation (S2A Fig). Therefore, we thought that even if the cleavage of PABP is more robust, cleavage of PABP should not result in aSG formation. We also found that 2A-induced aSGs are a common phenomenon in picornaviruses (Fig 8), and the dynamics of TIA-1 in aSGs were similar to those of TIA-1 in tSGs (Fig 1C). Thus, it is possible that aSGs are also non-membrane-bound cellular compartments and formed via liquid-liquid phase separation (LLPS), as is the case for tSGs. Although the details of LLPS are not clear, previous studies have shown that the high concentration of RBPs, which contain intrinsically disordered regions (IDRs), triggers LLPS [45–47]. When the concentration of IDR-containing RBPs reaches a certain level, the LLPS of these RBPs can be spontaneously initiated by IDR-mediated interaction, thus promoting tSG formation. Furthermore, during molecular crowding, IDR-containing RBPs can initiate LLPS at lower protein levels [46]. The expression of 2A could shuttle many nuclear IDR-containing RBPs, including hnRNPs, TDP-43, HuR, and Sam68, to the cytoplasm [14,48–51], thus resulting in increased concentrations of IDR-containing RBPs in the cytoplasm. Furthermore, the cleavage of eIF4GI by 2A leads to the accumulation of stalled PICs, thus resulting in molecular crowding. In cells expressing eIF4GIG689E and 2A, translation is initiated, molecular crowding is inhibited [35,52], and tSG formation is blocked (S5F and S5G Fig), and aSGs are therefore unable to form (Fig 2E and 2F and S8 Fig).
In principle, the PICs rapidly exchange between tSGs and polyribosomes, the addition of CHX during AS and HS treatment inhibits the dissociation of polyribosomes and results in the disassembly of tSGs [4,5]. However, we found that aSGs could not be dissolved by CHX (Fig 1F and 1G). Furthermore, we also found that the aSGs did not contain eIF1a, eIF3a, eIF4A, eIF4E, eIF4G, and RPS3 (S1A–S1H and S2B Figs), indicating that the mRNAs in aSGs are not equipped with eIFs and 40S ribosomal subunits, and cannot participate in translation. Therefore, the mRNAs within aSGs cannot flow to polyribosomes and aSGs cannot be dissolved by CHX.
Piotrowska et al. found that poliovirus did not block HS-induced tSG formation at 4 hpi [14], but we found that both AS- and HS-induced tSG formation were inhibited at 4 hpi and 6 hpi during EV71 infection (Fig 3A and 3B and S4A and S4B Fig). We speculate that poliovirus could not block HS-induced tSGs completely at 4hpi, but as infection proceeded, poliovirus blocked HS-induced tSGs eventually. To our knowledge, we are the first to find that tSGs are totally abrogated specifically by 2A, but not by 3C. Our conclusion is supported by three lines of evidence. First, EV71 and 2A still blocked tSG formation in cells expressing 3C cleavage-resistant G3BP (G3BPQ326E) (Fig 3D and 3E, S4D, S4E, S5D and S5E Figs), which contradicts previous findings [21,38,49]. We speculate that most of the viruses failed to replicate in cells expressing G3BPQ326E, because G3BP exhibits antiviral activity against several picornaviruses [53], thus resulting in an eight-fold reduction in viral titer and tSG formation rescue [21]. Although tSG formation was inhibited in 3C-expressing cells (S4D Fig), we think this is just a phenomenon accompanying the cleavage of G3BP by 3C to inhibit the antiviral activity of G3BP in the late phase of EV71 infection, because tSG inhibition occurred much earlier than G3BP cleavage (Fig 3C and 3D). Second, the addition of GuHCl in EV71-infected cells suppressed viral replication to an extremely low level, and the protease activity of 3C could not be detected (Fig 3F and S4F Fig), but tSG formation was still blocked in infected cells (Fig 3G and 3H and S4F Fig), suggesting that 3C is not required for blocking tSG formation. Third, by using the EV71-2AC110S mutant virus, in which 2A protease activity was abolished but 3C protease activity was intact (Fig 5A), we found that EV71-2AC110S induced tSG formation instead of inhibiting it (Fig 5B–5D), demonstrating that 2A is indispensable for the EV71-mediated inhibition of tSG formation. We also tried to rescue recombinant EV71 with a catalytically inactive 3C mutation but failed, which may be attributed to the presence of many 3C cleavage sites in the viral polyproteins.
2A of EV71 was previously suggested to induce tSG formation, but the exact mechanism is unknown [21,22]. To our knowledge, we are the first to demonstrate that EV71 induces tSG formation through the PKR-eIF2α pathway via dsRNA, but not 2A (Fig 6); on the contrary, 2A is indispensable for the blockage of tSG formation (Figs 4 and 5 and S5 Fig). Although both induction of aSG formation and blockage of tSG formation require 2A protease activity, the molecular mechanism of 2A contribution to induction of aSG formation and blockage of tSG formation is different. The aSG formation is accompanied by the tSG blockage in 2A-expessing or EV71-infected cells (Figs 3A, 4A and 4B, S4A and S5A Figs), but inhibition of 2A-induced aSGs by expression of eIF4GIG689E is unable to recover tSGs in the presence of AS (Fig 2E and 2F and S5F and S5G Fig); Furthermore, KD of PKR or expression of eIF2αS51A inhibits EV71-2AC110S-induced tSG formation (Fig 6), but has no effect on EV71-induced aSG formation (Figs 1D, 1E, 6B and 6C), suggesting that blockage of tSG formation and induction of aSGs by virus are independent to each other, not due to turning tSGs into aSGs. The molecular mechanism of how 2A prevents tSG formation should be more complicated.
EV71-2AC110S-induced tSGs contained eIFs and both viral and cellular mRNAs (Figs 6B, 7D and 7E) to inhibit viral translation (Fig 7C). But EV71-induced aSGs did not contain a series of eIFs or viral mRNA (S1A–S1H Fig, Figs 6B, 7D and 7E) and instead contained selectively confined host mRNAs (Fig 7E) to facilitate viral translation (Fig 7B). The mechanism of how viral RNA avoids being recruited to the aSGs is unclear, but we thought differences of RNA-binding proteins within aSGs and tSGs result in elimination of viral RNA from aSGs. It has been reported that many cellular mRNAs also contained IRES [54], and if the binding proteins of IRES-containing cellular mRNAs are eliminated from aSGs, the aSGs should not stall these IRES-containing cellular mRNAs either.
In conclusion, our findings reveal the common molecular mechanisms and functions of picornavirus-mediated regulation of SG formation (Fig 9). On one hand, host cells recognize viral dsRNA via PKR during EV71 infection, which activates the PKR-eIF2α signaling cascade and results in tSG formation. Both cellular and viral mRNAs are sequestered in tSGs, which leads to an overall shutdown of translation to inhibit viral translation. On the other hand, 2A protease of EV71 blocks tSG formation drastically to remove the antiviral effects of tSG formation from host cells. Furthermore, 2A induces aSG formation by cleaving eIF4GI to sequester cellular mRNA but releases eIFs and viral mRNA, which benefit viral translation. Thus, the blockage of tSG formation and induction of aSG formation are strategies used by EV71 to survive in host cells.
Further studies are needed to reveal the complete RNA and protein composition of aSGs and the mechanisms of EV71-induced blockage of tSG formation. Likewise, the full effect of the regulation of SG formation on aspects of cell physiology other than translation control of host mRNAs should be further explored. Previous studies have suggested that persistent SG formation is linked to some neurodegenerative diseases [55–57]. Research in these areas may offer fascinating new insights into cellular function and may also yield novel therapeutic strategies in picornavirus-induced disease.
Human RD (Rhabdomyosarcoma cells and were obtained from China Center for Type Culture Collection), HEK293T (Human embryonic kidney 293 cells and were obtained from China Center for Type Culture Collection), HeLa cells (Human cervical cancer epithelial cells and were obtained from China Center for Type Culture Collection), and stably expressing cells (GFP-G3BP-HeLa/RD, GFP-G3BPQ326E-HeLa/RD, GFP-TIA-1-HeLa) derived from HeLa or RD cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco) supplemented with 10% fetal bovine serum (FBS) (Gibco) and 100 U/ml penicillin/streptomycin (Gibco) at 37°C and 5% CO2. Other stably expressing cells (eIF4GI-HA-HeLa, eIF4GIG689E-HA-HeLa), cells with KD of PKR (shPKR-HeLa) or negative control (shNC-HeLa) cells derived from HeLa cells were maintained in DMEM with 10% FBS, 100 U/ml penicillin/streptomycin, and 1 μg/ml puromycin (Sigma-Aldrich) at 37°C and 5% CO2.
For infection, HeLa or RD cells were infected with DMEM containing viruses with a multiplicity of infection (MOI) of 10 plaque-forming units (PFUs) or as indicated in the figure legends. After 1 h incubation, the medium was replaced with fresh DMEM with 4% FBS, and this time point was considered 0 hpi; cells were harvested for further analysis at 6 hpi or as indicated. For transfection, plasmids and RNAs were transfected by using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions, and cells were harvested or subjected to further treatment at 24 h post-transfection (hpt) or as indicated.
For tSG induction, cells were treated with 200 μM AS (Sigma-Aldrich) or incubated at 43°C (HS) for 1 h (or otherwise as indicated) before being harvested for further analysis.
For quantification of the foci of different protein markers in EV71-infected or 2A-expressing cells, EV71- or Myc-tagged cells with Sam68 and TIA-1 foci were counted from cells containing EV71 or Myc tag, and others were counted from cells forming TIA-1 foci. For quantification of tSG formation in EV71-infected or 2A-expressing cells, G3BP was used as an AS- or HS-induced SG marker, and HSP27 was used as an HS-induced SG marker. Cells were considered tSG positive only if they had SGs containing the indicated marker, and the diameter of the biggest SGs in EV71-infected or 2A-expressing cells was at least 1.5 μm (the diameter of the biggest SGs in mock-treated cells was about 3–5 μm).
To distinguish tSGs and aSGs, we treated SG-formed cells with 50 μg/ml CHX (Sigma-Aldrich) for 1 h before fixation. The tSGs would disassemble, and aSGs would remain; as a positive control, cells were treated with AS (200 μM for 0.5 h) to induce tSG formation and then treated with both CHX and AS for another 1 h. In the GuHCl treatments, HeLa or RD cells were infected with 2 mM GuHCl (Sigma-Aldrich) after viral incubation for 1 h; for Western blotting assays, cells were harvested at 6 hpi, and for IF assays, cells were treated with 200 μM AS in the presence of GuHCl for the final 1 h and fixed at 6 hpi.
The coding regions of G3BP (NCBI accession no. NM_005754.2), TIA-1 (NCBI accession no. NM_022037.2, isoform 1), eIF2α (NCBI accession no. NM_004094.4) and PABP (NCBI accession no. NM_002568.3) were obtained from HeLa cells via RNA extraction and subsequent reverse transcription polymerase chain reaction (RT-PCR) and cloned into the PmeI–Bsp119I sites in pWPI vector with a N-terminal GFP or HA tag (the original region of EMCV and GFP in pWPI was removed). To clone full-length eIF4GI (NCBI accession no. NM_001194947.1, isoform 6), the N-terminal coding region (1–203 aa) was generated by chemosynthesis, and the C-terminal coding region (204–1606 aa) was obtained from HeLa cells via RT-PCR; the full-length clone was generated via overlapping PCR and cloned into pHAGE (puro) vector (substituting fluorescent tag in pHAGE-CMV-MCS-IRES-ZsGreen [EvNO00061605] with puromycin) with an HA tag at the C-terminal. The GFP-PABPM490P/Q540N mutant was generated by using overlapping PCR; PABPM490P was cloned first, and then a Q540N mutation was added.
A plasmid, pCDNA3.0-EV71-CDS, containing completed CDS of EV71 (Hubei-Xiangyang-09, genotype C4) was obtained from Dr. K.L. Wu (State Key Laboratory of Virology, Wuhan University, China). The 5’-UTR was obtained from the total RNAs of EV71-infected HeLa cells by using a 5´ RACE System (Invitrogen) for rapid amplification of cDNA ends according to the manufacturer’s instructions (GSP1, AGGGCAGTGCGTTTATGTATGG; GSP2, GGGTGACTGTCTTCCGTTCCT). The 3’-UTR was obtained from the EV71-infected HeLa cells by using RT-PCR (GGAGAGATCCAGTGGGTTAAG and oligo[dT]18). A completed genome of the EV71 clone (pCDNA3.0-EV71) was generated via overlapping PCR and cloned into the pCDNA3.0 vector, and it was taken as a template for creating all other clones that contained the region of the EV71 genome. EV71 proteins were obtained via PCR and cloned into the pCAGGS vector. To generate a vector containing an IRES located in multiple clone sites upstream (pCDNA3.0-IRES), the 5’-UTR of EV71 was cloned into the EcoRI–ClaI sites in the pCDNA3.0 vector. The 2A protease and 2AC110S were obtained via PCR and cloned into the pCDNA3.0-IRES vector. To clone the 2A protease of EV71-BrCr (NCBI accession no. U22521), PV (NCBI accession no. NC_002058.3), and CVA (NCBI accession no. KC117318.1), the coding regions of the three respective viral proteases were generated by chemosynthesis and cloned into the pCDNA3.0-IRES vector.
All the structures were confirmed by DNA sequencing.
The shRNA constructs were designed by using the pLKO.1 vector [58] according to protocols recommended by the manufacturer. For stable KD of target protein expression, HEK293T cells were cotransfected with plasmid psPAX2, pMD2.G, and shRNA constructs for 48 h to generate lentiviral particles. The medium was harvested and filtered by a 0.45 μm filter and then divided and stored at -80°C. When the lentivirus was used for infection, HeLa cells were seeded in 6-well plates (1–2 x 105 cells per well), and medium containing the lentivirus was added. After 24 h incubation, the infection medium was replaced with fresh complete growth media. Then, 24 h later, the infected cells were replated and selected in complete growth media with the addition of 2 μg/ml puromycin. After 48 h, the selected cells were replated and selected again in complete growth media with the addition of 2 μg/ml puromycin for another 48 h. Then, the selected cells were maintained in complete growth media plus 1 μg/ml puromycin or subjected to Western blotting to confirm deletion of target proteins. The stable KD cell lines were used for subsequent experiments. The target sequences for the shRNA constructs were: shPKR, GAGGCGAGAAACTAGACAAAG; shNC, GCGCGATAGCGCTAATAATTT.
Cells were infected with lentiviruses that were generated by cotransfection of plasmid psPAX2, pMD2.G, and target protein expression constructs. The eIF4GI-HA-HeLa and eIF4GIG689E-HA-HeLa cell lines were obtained as previously described for stable KD cell lines. For the other cell lines with overexpression of GFP-tagged targets, GFP-positive cells were sorted from the infected cells via flow cytometry and cultured in complete growth media. The cell lines with stable overexpression were used for subsequent experiments.
The full-length recombinant EV71 infectious clone was constructed into a pBS vector bearing a T7 promoter upstream of the virus genome (pBS-T7-EV71). The 2A protease activity-deficient EV71 infectious clone was constructed using site-directed mutagenesis via overlapping PCR (pBS-T7-EV71-2AC110S). An IRES structure was inserted between the sequence of VP1 and 2A to counteract the 2Apro defect. Constructs were then linearized by restriction enzyme and purified by phenol:chloroform and ethanol precipitation. To package EV71 viruses, viral RNA was transcribed using the TranscriptAid T7 High-Yield Transcription Kit (ThermoFisher Scientific) and purified using the RNeasy Mini Kit (Qiagen). The viral RNA (2 μg) transcripts were transfected into HeLa cells grown in a monolayer on 6-well plates by using Lipofectamine 2000 for 3 days. The supernatants were passaged on fresh RD cells for further amplification. After 3 days, the supernatants were collected and stored at -80°C, and the cells were collected and divided into two groups, one for Western blotting to confirm the viral infection and the other for RT-PCR and DNA sequencing to confirm the mutation. A single recovered recombinant EV71-2AC110S was isolated by removing the agar plug during a plaque assay. The agar plug was dissolved in 500 μl of opti-MEM overnight at 4°C, and half was used for EV71-2AC110S amplification by infecting RD cell monolayers. Finally, the whole genome sequence of EV71-2AC110S was confirmed by RT-PCR and DNA-sequencing, and the titer of EV71-2AC110S was measured via plaque assay.
For IF, cells were fixed with 4% (wt/vol) paraformaldehyde/phosphate-buffered saline (PBS) and permeated with 0.2% (wt/vol) Triton X-100/PBS solution at room temperature (RT) for 20 min, respectively, and then blocked with 3% (wt/vol) bovine serum albumin (BSA) in PBS at RT for 30 min. Primary antibodies were diluted in 1% (wt/vol) BSA/PBS and incubated overnight at 4°C, followed by incubation of secondary antibodies at RT for 2 h. The following dye-conjugated secondary antibodies were used for this analysis: Alexa Fluor 647 donkey anti-goat immunoglobulin (IgG) H+L, Alexa Fluor 488 donkey anti-rabbit IgG H+L, and Alexa Fluor 594 donkey anti-mouse IgG H+L (Life Technologies). After being stained with 1 μg/ml DAPI (Roche) in PBS for 5 min, cells were mounted with Prolong Diamond Antifade Mountant (Life Technology) and examined on a Leica confocal microscope.
For Western blotting, cells were harvested and lysed in lysis buffer (150 nM NaCl, 50 nM Tris-HCl [pH 7.4],1% Triton X-100, 1 mM EDTA [pH 8.0], and 0.1% sodium dodecyl sulfate [SDS]) with a protease inhibitor cocktail, incubated on ice for 30 min, and centrifuged at 4°C for 30 min at 12,000 g. The supernatants were boiled in SDS-polyacrylamide gel electrophoresis (PAGE) loading buffer at 100°C for 10 min and then resolved on SDS-PAGE and detected on a Fujifilm LAS-4000 imaging system. The indicated primary and horseradish peroxidase-conjugated secondary antibodies (ThermoFisher Scientific) were used.
The following primary antibodies were used: mouse monoclonal anti-c-Myc (Cat #sc-40), rabbit polyclonal anti-c-Myc (Cat #sc-789), rabbit monoclonal anti-Sam68 (Cat #sc-333), goat polyclonal anti-TIA-1 (Cat #sc-1751), mouse monoclonal anti-GAPDH (Cat #sc-32233) and rabbit monoclonal anti-GFP (Cat #sc-8334) were purchased from Santa Cruz Biotechnology. Mouse monoclonal anti-G3BP (Cat #611127) and mouse monoclonal anti-TIAR (Cat #610352) were purchased from BD Transduction Laboratories. Mouse monoclonal anti-β-actin (Cat #AC004), rabbit polyclonal anti-eIF4A (Cat #A5294), rabbit polyclonal anti-eIF4E (Cat #A2162), rabbit polyclonal anti-eIF1a (Cat #A5917), rabbit polyclonal anti-eIF3a (Cat #A0573), rabbit polyclonal anti-RPS3 (Cat #A11131), and rabbit polyclonal anti-3C (Cat #A10003) were purchased from ABclonal. Rabbit monoclonal anti-eIF4G (Cat #2469S), rabbit monoclonal anti-p-eIF2α (Cat #9721), and rabbit monoclonal anti-eIF2α (Cat #9722) were purchased from Cell Signaling Technology. Mouse monoclonal anti-Flag (Cat #F1804), mouse monoclonal anti-HA (Cat #H9658), and rabbit monoclonal anti-HA (Cat #H6908) were purchased from Sigma-Aldrich. Mouse monoclonal anti-PABP (Cat #ab6125) was purchased from Abcam. Mouse monoclonal anti-HSP27 (Cat #ADI-SPA-800D) was purchased from StressGen. Mouse monoclonal anti-EV71 (Cat #MAB979) and Mouse monoclonal anti-puromycin, clone 12D10 (Cat #MABE343) were purchased from Millipore. Mouse monoclonal anti-VP1 was purchased from Abmax (Clone 22A14) [59].
For detection of total polyadenylated mRNA (polyA+ mRNA), cells were plated on coverslips and incubated overnight before treatment with AS, EV71, or 2A. Cells were fixed with 2% formaldehyde for 10 min and processed as previously described [60] by using a 3’-biotinylated oligo(dT)40 probe. Cells were then processed as those described for the aforementioned IF assays. The oligo(dT)40 probe was visualized by streptavidin conjugated to cyanin 3 (Cy3).
The ViewRNA ISH Cell Assay Kit and probes for EV71 positive-strand RNA (+vRNA) and PABPC1 mRNA were purchased from Affymetrix and used to detect target mRNAs according to protocols recommended by the manufacturer.
A Renilla luciferase gene sequence was constructed into the aforementioned EV71 infectious clone backbone (pBS-T7-EV71) by replacing the whole viral protein-coding sequence (UTREV71-Rluc). The UTRPABPC1-Rluc reporter was generated by replacing the EV71 UTR region by PABPC1 UTR region. The RNAs were transcribed using the TranscriptAid T7 High-Yield Transcription Kit (ThermoFisher Scientific) and purified using the RNeasy Mini Kit (Qiagen). To evaluate the influence of aSGs on EV71 or PABPC1 translation, eIF4GI-HA- and eIF4GIG689E-HA-HeLa cells were seeded in 24-well plates, and the RNA (0.4 μg/well) transcripts were transfected into the cells after EV71 (MOI = 10) infection for 3 hours or EV71-2AC110S (MOI = 10) infection for 6 hours, then analyzed the reporter expression at 1.5 hpt and 3 hpt. Cells infected with EV71-2AC110S served as negative control and analyzed the reporter expression at 3 hpt. To evaluate the influence of tSGs on EV71 or PABPC1 translation, shNC- and shPKR-HeLa cells were seeded in 24-well plates and the RNA (0.4 μg/well) transcripts were transfected into the cells after EV71-2AC110S (MOI = 10) infection for 6 h or EV71 (MOI = 10) infection for 3 h. The reporter expression in EV71-2AC110S-infected cells were analyzed at 3 hpt and 6 hpt and the reporter expression in EV71-infected cells were analyzed at 3 hpt as control. Renilla luciferase activity was assessed by using a Renilla Luciferase Assay Kit (Promega) according to the manufacturer’s instructions. All experiments were performed in triplicate, and assays were repeated at least three times.
Statistical analysis was performed using GraphPad Prism v6.01. All results are expressed as means ± SD of at least three independent experiments (n≥3). The p value was calculated using an unpaired Student's t-test. In all tests, p>0.05 was considered non-statistically significant (n.s.), and p<0.05 was considered statistically significant, marked as follows: *, p<0.05; **, p<0.01; ***, p<0.001.
|
10.1371/journal.pcbi.1005268 | Could a Neuroscientist Understand a Microprocessor? | There is a popular belief in neuroscience that we are primarily data limited, and that producing large, multimodal, and complex datasets will, with the help of advanced data analysis algorithms, lead to fundamental insights into the way the brain processes information. These datasets do not yet exist, and if they did we would have no way of evaluating whether or not the algorithmically-generated insights were sufficient or even correct. To address this, here we take a classical microprocessor as a model organism, and use our ability to perform arbitrary experiments on it to see if popular data analysis methods from neuroscience can elucidate the way it processes information. Microprocessors are among those artificial information processing systems that are both complex and that we understand at all levels, from the overall logical flow, via logical gates, to the dynamics of transistors. We show that the approaches reveal interesting structure in the data but do not meaningfully describe the hierarchy of information processing in the microprocessor. This suggests current analytic approaches in neuroscience may fall short of producing meaningful understanding of neural systems, regardless of the amount of data. Additionally, we argue for scientists using complex non-linear dynamical systems with known ground truth, such as the microprocessor as a validation platform for time-series and structure discovery methods.
| Neuroscience is held back by the fact that it is hard to evaluate if a conclusion is correct; the complexity of the systems under study and their experimental inaccessability make the assessment of algorithmic and data analytic technqiues challenging at best. We thus argue for testing approaches using known artifacts, where the correct interpretation is known. Here we present a microprocessor platform as one such test case. We find that many approaches in neuroscience, when used naïvely, fall short of producing a meaningful understanding.
| The development of high-throughput techniques for studying neural systems is bringing about an era of big-data neuroscience [1, 2]. Scientists are beginning to reconstruct connectivity [3], record activity [4], and simulate computation [5] at unprecedented scales. However, even state-of-the-art neuroscientific studies are still quite limited in organism complexity and spatiotemporal resolution [6–8]. It is hard to evaluate how much scaling these techniques will help us understand the brain.
In neuroscience it can be difficult to evaluate the quality of a particular model or analysis method, especially in the absence of known truth. However, there are other systems, in particular man made ones that we do understand. As such, one can take a human-engineered system and ask if the methods used for studying biological systems would allow understanding the artificial system. In this way, we take as inspiration Yuri Lazbnick’s well-known 2002 critique of modeling in molecular biology, “Could a biologist fix a radio?” [9]. However, a radio is clearly much simpler than the nervous system, leading us to seek out a more complex, yet still well-understood engineered system. The microprocessors in early computing systems can serve this function.
Here we will try to understand a known artificial system, a classical microprocessor by applying data analysis methods from neuroscience. We want to see what kind of an understanding would emerge from using a broad range of currently popular data analysis methods. To do so, we will analyze the connections on the chip, the effects of destroying individual transistors, single-unit tuning curves, the joint statistics across transistors, local activities, estimated connections, and whole-device recordings. For each of these, we will use standard techniques that are popular in the field of neuroscience. We find that many measures are surprisingly similar between the brain and the processor but that our results do not lead to a meaningful understanding of the processor. The analysis can not produce the hierarchical understanding of information processing that most students of electrical engineering obtain. It suggests that the availability of unlimited data, as we have for the processor, is in no way sufficient to allow a real understanding of the brain. We argue that when studying a complex system like the brain, methods and approaches should first be sanity checked on complex man-made systems that share many of the violations of modeling assumptions of the real system.
The MOS 6502 (and the virtually identical MOS 6507) were the processors in the Apple I, the Commodore 64, and the Atari Video Game System (VCS) (see [10] for a comprehensive review). The Visual6502 team reverse-engineered the 6507 from physical integrated circuits [11] by chemically removing the epoxy layer and imaging the silicon die with a light microscope. Much like with current connectomics work [12, 13], a combination of algorithmic and human-based approaches were used to label regions, identify circuit structures, and ultimately produce a transistor-accurate netlist (a full connectome) for this processor consisting of 3510 enhancement-mode transistors. Several other support chips, including the Television Interface Adaptor (TIA) were also reverse-engineered and a cycle-accurate simulator was written that can simulate the voltage on every wire and the state of every transistor. The reconstruction has sufficient fidelity to run a variety of classic video games, which we will detail below. The simulation generates roughly 1.5GB/sec of state information, allowing a real big-data analysis of the processor.
The simplicity of early video games has led to their use as model systems for reinforcement learning [14] and computational complexity research [15]. The video game system (“whole animal”) has a well defined output in each of the three behavioral conditions (games). It produces an input-dependent output that is dynamic, and, in the opinion of the authors, quite exciting. It can be seen as a more complex version of the Mus Silicium project [16]. It is also a concrete implementation of a thought experiment that has been mentioned on and off in the literature [17–20]. The richness of the dynamics and our knowledge about its inner workings makes it an attractive test case for approaches in neuroscience.
Here we will examine three different “behaviors”, that is, three different games: Donkey Kong (1981), Space Invaders (1978), and Pitfall (1981). Obviously these “behaviors” are qualitatively different from those of animals and may seem more complicated. However, even the simple behaviors that are studied in neuroscience still involve a plethora of components, typically including the allocation of attention, cognitive processing, and multiple modalities of inputs and outputs. As such, the breadth of ongoing computation in the processor may actually be simpler than those in the brain.
The objective of clever experimental design in neuroscience often is to find behaviors that only engage one kind of computation in the brain. In the same way, all our experiments on the chip will be limited by us only using these games to probe it. As much as more neuroscience is interested in naturalistic behaviors [21], here we analyze a naturalistic behavior of the chip. In the future it may be possible to excute simpler, custom code on the processor to tease apart aspects of computation, but we currently lack such capability in biological organisms.
Much has been written about the differences between computation in silico and computation in vivo [22, 23]—the stochasticity, redundancy, and robustness [24] present in biological systems seems dramatically different from that of a microprocessor. But there are many parallels we can draw between the two types of systems. Both systems consist of interconnections of a large number of simpler, stereotyped computing units. They operate on multiple timescales. They consist of somewhat specialized modules organized hierarchically. They can flexibly route information and retain memory over time. Despite many differences there are also many similarities. We do not wish to overstate this case—in many ways, the functional specialization present in a large mammalian brain far eclipses that present in the processor. Indeed, the processor’s scale and specialization share more in common with C. elegans than a mouse.
Yet many of the differences should make analysing the chip easier than analyzing the brain. For example, it has a clearer architecture and far fewer modules. The human brain has hundreds of different types of neurons and a similar diversity of proteins at each individual synapse [25], whereas our model microprocessor has only one type of transistor (which has only three terminals). The processor is deterministic while neurons exhibit various sources of randomness. With just a couple thousand transistors it is also far smaller. And, above all, in the simulation it is fully accessible to any and all experimental manipulations that we might want to do on it.
Importantly, the processor allows us to ask “do we really understand this system?” Most scientists have at least behavioral-level experience with these classical video game systems, and many in our community, including some electrophysiologists and computational neuroscientists, have formal training in computer science, electrical engineering, computer architecture, and software engineering. As such, we believe that most neuroscientists may have better intuitions about the workings of a processor than about the workings of the brain.
What constitutes an understanding of a system? Lazbnick’s original paper argued that understanding was achieved when one could “fix” a broken implementation. Understanding of a particular region or part of a system would occur when one could describe so accurately the inputs, the transformation, and the outputs that one brain region could be replaced with an entirely synthetic component. Indeed, some neuroengineers are following this path for sensory [26] and memory [27] systems. Alternatively, we could seek to understand a system at differing, complementary levels of analysis, as David Marr and Tomaso Poggio outlined in 1982 [28]. First, we can ask if we understand what the system does at the computational level: what is the problem it is seeking to solve via computation? We can ask how the system performs this task algorithmically: what processes does it employ to manipulate internal representations? Finally, we can seek to understand how the system implements the above algorithms at a physical level. What are the characteristics of the underlying implementation (in the case of neurons, ion channels, synaptic conductances, neural connectivity, and so on) that give rise to the execution of the algorithm? Ultimately, we want to understand the brain at all these levels.
In this paper, much as in systems neuroscience, we consider the quest to gain an understanding of how circuit elements give rise to computation. Computer architecture studies how small circuit elements, like registers and adders, give rise to a system capable of performing general-purpose computation. When it comes to the processor, we understand this level extremely well, as it is taught to most computer science undergraduates. Knowing what a satisfying answer to “how does a processor compute?” looks like makes it easy to evaluate how much we learn from an experiment or an analysis.
We can draw from our understanding of computer architecture to firmly ground what a full understanding of a processor would look like (Fig 1). The processor is used to implement a computing machine. It implements a finite state machine which sequentially reads in an instruction from memory (Fig 1a, green) and then either modifies its internal state or interacts with the world. The internal state is stored in a collection of byte-wide registers (Fig 1a, red). As an example, the processor might read an instruction from memory telling it to add the contents of register A to the contents of register B. It then decodes this instruction, enabling the arithmetic logic unit (ALU, Fig 1a, blue) to add those registers, storing the output. Optionally, the next instruction might save the result back out to RAM (Fig 1a, yellow). It is this repeated cycle that gives rise to the complex series of behaviors we can observe in this system. Note that this description in many ways ignores the functions of the individual transistors, focusing instead on circuits modules like “registers” which are composed of many transistors, much as a systems neuroscientist might focus on a cytoarchitecturally-distinct area like hipppocampus as opposed to individual neurons.
Each of the functions within the processor contains algorithms and a specific implementation. Within the arithmetic logic unit, there is a byte wide adder, which is in part made of binary adders (Fig 1b), which are made out of AND/NAND gates, which are made of transistors. This is in a similar way as the brain consists of regions, circuits, microcircuits, neurons, and synapses.
If we were to analyze a processor using techniques from systems neuroscience we would hope that it helps guide us towards the descriptions that we used above. In the rest of the paper we will apply neuroscience techniques to data from the processor. We will finally discuss how neuroscience can work towards techniques that will make real progress at moving us closer to a satisfying understanding of computation, in the chip, and in our brains.
Validating our understanding of complex systems is incredibly difficult when we do not know the actual ground truth. Thus we use an engineered system, the MOS6502, where we understand every aspect of its behavior at many levels. We will examine the processor at increasingly-fine spatial and temporal resolutions, eventually achieving true “big-data” scale: a “processor activity map”, with every transistor state and every wire voltage. As we apply the various techniques that are currently used in neuroscience we will ask how the analyses bring us closer to an understanding of the microprocessor (Fig 2). We will use this well defined comparison to ask questions about the validity of current approaches to studying information processing in the brain.
The earliest investigations of neural systems were in-depth anatomical inquiries [29]. Fortunately, through large scale microscopy (Fig 2a) we have available the full 3d connectome of the system. In other words, we know how each transistor is connected to all the others. The reconstruction is so good, that we can now simulate this processor perfectly—indeed, were it not for the presence of the processor’s connectome, this paper would not have been possible. This process is aided by the fact that we know a transistor’s deterministic input-output function, whereas neurons are both stochastic and vastly more complex.
Recently several graph analysis methods ranging from classic [30] to modern [31, 32] approaches have been applied to neural connectomes. The approach in [31] was also applied to a region of this processor, attempting to identify both circuit motifs as well as transistor “types” (analogous to cell types) in the transistor wiring diagram. Fig 3 (adapted from [31]) shows the results of the analysis. We see that one identified transistor type contains the “clocked” transistors, which retain digital state. Two other types contain transistors with pins C1 or C2 connected to ground, mostly serving as inverters. An additional identified type controls the behavior of the three registers of interest (X, Y, and S) with respect to the SB data bus, either allowing them to latch or drive data from the bus. The repeat patterns of spatial connectivity are visible in Fig 3a, showing the man-made horizontal and vertical layout of the same types of transistors.
While superficially impressive, based on the results of these algorithms we still can not get anywhere near an understanding of the way the processor really works. Indeed, we know that for this processor there is only one physical “type” of transistor, and that the structure we recover is a complex combination of local and global circuitry.
In neuroscience, reconstructing all neurons and their connections perfectly is the dream of a large community studying connectomics [33, 34]. Current connectomics approaches are limited in their accuracy and ability to definitively identify synapses [13], Unfortunately, we do not yet have the techniques to also reconstruct the i/o function–neurotransmitter type, ion channel type, I/V curve of each synapse, etc.—of each neuron. But even if we did, just as in the case of the processor, we would face the problem of understanding the brain based on its connectome. As we do not have algorithms that go from anatomy to function at the moment that go considerably beyond cell-type clustering [31, 35, 36] it is far from obvious how a connectome would allow an understanding of the brain.
Note we are not suggesting connectomics is useless, quite the contrary–in the case of the processor the connectome was the first crucial step in enabling reliable, whole-brain-scale simulation. But even with the whole-brain connectome, extracting hierarchical organization and understanding the nature of the underlying computation is incredibly difficult.
Lesions studies allow us to study the causal effect of removing a part of the system. We thus chose a number of transistors and asked if they are necessary for each of the behaviors of the processor (Fig 4. In other words, we asked if removed each transistor, if the processor would then still boot the game. Indeed, we found a subset of transistors that makes one of the behaviors (games) impossible. We can thus conclude they are uniquely necessary for the game—perhaps there is a Donkey Kong transistor or a Space Invaders transistor. Even if we can lesion each individual transistor, we do not get much closer to an understanding of how the processor really works.
This finding of course is grossly misleading. The transistors are not specific to any one behavior or game but rather implement simple functions, like full adders. The finding that some of them are important while others are not for a given game is only indirectly indicative of the transistor’s role and is unlikely to generalize to other games. Lazebnik [9] made similar observations about this approach in molecular biology, suggesting biologists would obtain a large number of identical radios and shoot them with metal particles at short range, attempting to identify which damaged components gave rise to which broken phenotype.
This example nicely highlights the importance of isolating individual behaviors to understand the contribution of parts to the overall function. If we had been able to isolate a single function, maybe by having the processor produce the same math operation every single step, then the lesioning experiments could have produced more meaningful results. However, the same problem exists in neuroscience. It is extremely difficult or technically impossible to produce behaviors that only require a single aspect of the brain.
Beyond behavioral choices, we have equivalent problems in neuroscience that make the interpretation of lesioning data complicated [37]. In many ways the chip can be lesioned in a cleaner way than the brain: we can individually abolish every single transistor (this is only now becoming possible with neurons in simple systems [38, 39]). Even without this problem, finding that a lesion in a given area abolishes a function is hard to interpret in terms of the role of the area for general computation. And this ignores the tremendous plasticity in neural systems which can allow regions to take over for damaged areas. In addition to the statistical problems that arise from multiple hypothesis testing, it is obvious that the “causal relationship” we are learning is incredibly superficial: a given transistor is obviously not specialized for Donkey Kong or Space Invaders.
While in most organisms individual transistors are not vital, for many less-complex systems they are. Lesion individual interneurons in C. elegans or the H1 neuron in the fly can have marked behavioral impacts. And while lesioning larger pieces of circuitry, such as the entire TIA graphics chip, might allow for gross segregation of function, we take issue with this constituting “understanding”. Simply knowing functional localization, at any spatial scale, is only the most nacent step to the sorts of understanding we have outlined above.
We may want to try to understand the processor by understanding the activity of each individual transistor. We study the “off-to-on” transition, or “spike”, produced by each individual transistor. Each transistor will be activated at multiple points in time. Indeed, these transitions look surprisingly similar to the spike trains of neurons (Fig 5). Following the standards in neuroscience we may then quantify the tuning selectivity of each transistor. For each of our transistors we can plot the spike rate as a function of the luminance of the most recently displayed pixel (Fig 6). For a small number of transistors we find a strong tuning to the luminance of the most recently displayed pixel, which we can classify into simple (Fig 6a) and (Fig 6b) complex curves. Interestingly, however, we know for each of the five displayed transistors that they are not directly related to the luminance of the pixel to be written, despite their strong tuning. The transistors relate in a highly nonlinear way to the ultimate brightness of the screen. As such their apparent tuning is not really insightful about their role. In our case, it probably is related to differences across game stages. In the brain a neuron can calculate something, or be upstream or downstream of the calculation and still show apparent tuning making the inference of a neurons role from observational data very difficult [40]. This shows how obtaining an understanding of the processor from tuning curves is difficult.
Much of neuroscience is focused on understanding tuning properties of neurons, circuits, and brain areas [41–44]. Arguably this approach is more justified for the nervous system because brain areas are more strongly modular. However, this may well be an illusion and many studies that have looked carefully at brain areas have revealed a dazzling heterogeneity of responses [45–47]. Even if brain areas are grouped by function, examining the individual units within may not allow for conclusive insight into the nature of computation.
Moving beyond correlating single units with behavior, we can examine the correlations present between individual transistors. We thus perform a spike-word analysis [48] by looking at “spike words” across 64 transistors in the processor. We find little to very weak correlation among most pairs of transistors (Fig 7a). This weak correlation suggests modeling the transistors’ activities as independent, but as we see from shuffle analysis (Fig 7b), this assumption fails disastrously at predicting correlations across many transistors.
In neuroscience, it is known that pairwise correlations in neural systems can be incredibly weak, while still reflecting strong underlying coordinated activity. This is often assumed to lead to insights into the nature of interactions between neurons [48]. However, the processor has a very simple nature of interactions and yet produces remarkably similar spike word statistics. This again highlights how hard it is to derive functional insights from activity data using standard measures.
The activity of the entire chip may be high dimensional, yet we know that the chip, just like the brain, has some functional modularity. As such, we may be able to understand aspects of its function by analyzing the average activity within localized regions, in a way analogous to the local field potentials or the BOLD signals from functional magnetic imaging that are used in neuroscience. We thus analyzed data in spatially localized areas (Fig 8a). Interestingly, these average activities look quite a bit like real brain signals (Fig 8b). Indeed, they show a rather similar frequency power relation of roughly power-law behavior. This is often seen as a strong sign of self-organized criticality [49]. Spectral analysis of the time-series reveals region-specific oscillations or “rhythms” that have been suggested to provide a clue to both local computation and overall inter-region communication. In the chip we know that while the oscillations may reflect underlying periodicity of activity, the specific frequencies and locations are epiphenomena. They arise as an artifact of the computation and tell us little about the underlying flow of information. And it is very hard to attribute (self-organized) criticality to the processor.
In neuroscience there is a rich tradition of analyzing the rhythms in brain regions, the distribution of power across frequencies as a function of the task, and the relation of oscillatory activity across space and time. However, the example of the processor shows that the relation of such measures to underlying function can be extremely complicated. In fact, the authors of this paper would have expected far more peaked frequency distributions for the chip. Moreover, the distribution of frequencies in the brain is often seen as indicative about the underlying biophysics. In our case, there is only one element, the transistor, and not multiple neurotransmitters. And yet, we see a similarly rich distribution of power in the frequency domain. This shows that complex multi-frequency behavior can emerge from the combination of many simple elements. Analyzing the frequency spectra of artifacts thus leads us to be careful about the interpretation of those occurring in the brain. Modeling the processor as a bunch of coupled oscillators, as is common in neuroscience, would make little sense.
Granger causality [50] has emerged as a method of assessing putative causal relationships between brain regions based on LFP data. Granger causality assesses the relationship between two timeseries X and Y by comparing the predictive power of two different time-series models to predict future values of Y. The first model uses only past values of Y, whereas the second uses the history of X and Y. The additon of X allows one to assess the putative “causality” (really, the predictive power) of X.
To see if we can understand information transmission pathways in the chip based on such techniques, we perform conditional Granger causality analysis on the above-indicated LFP regions for all three behavioral tasks, and plot the resulting inferences of causal interactions (Fig 9). We find that the decoders affect the status bits. We also find that the registers are affected by the decoder, and that the accumulator is affected by the registers. We also find communication between the two parts of the decoder for Donkey Kong, and a lack of communication from the accumulator to the registers in Pitfall. Some of these findings are true, registers really affect the accumulator and decoders really affect the status bits. Other insights are less true, e.g. decoding is independent and the accumulator obviously affects the registers. While some high level insights may be possible, the insight into the actual function of the processor is limited.
The analysis that we did is very similar to the situation in neuroscience. In neuroscience as well, the signals come from a number of local sources. Moreover, there are also lots of connections but we hope that the methods will inform us about the relevant ones. It is hard to interpret the results—what exactly does the Granger causality model tell us about. Granger causality tells us how activity in the past are predictive of activity in the future, and the link from there to causal interactions is tentative at best [51] and yet such methods are extensively used across large subfields of neuroscience. Even if such methods would reliably tell us about large scale influences, it is very hard to get from a coarse resolution network to the microscopic computations.
In line with recent advances in whole-animal recordings [2, 6–8], we measure the activity across all 3510 transistors simultaneously for all three behavioral states (Fig 10) and plot normalized activity for each transistor versus time. Much as in neural systems, some transistors are relatively quiet and some are quite active, with a clear behaviorally-specific periodicity visible in overall activity.
While whole-brain recording may facilitate identification of putative areas involved in particular behaviors [52], ultimately the spike-level activity at this scale is difficult to interpret. Thus scientists turn to dimensionality reduction techniques [2, 53, 54], which seek to explain high-dimensional data in terms of a low-dimensional representation of state. We use non-negative matrix factorization [55] to identify constituent signal parts across all time-varying transistor activity. We are thus, for the first time in the paper, taking advantage of all transistors simultaneously.
Non-negative matrix factorization assumes each recovered timeseries of transistor activity is a linear combination of a small number of underlying nonnegative time-varying signals (dimensions). Analogous with [2] we plot the recovered dimensions as a function of time (Fig 11a) and the transistor activity profile of each component (Fig 11b). We can also examine a map of transistor-component activity both statically (Fig 11c) and dynamically (S1–S3 Videos available in online supplementary materials). Clearly there is a lot of structure in this spatiotemporal dataset.
To derive insight into recovered dimensions, we can try and relate parts of the low-dimensional time series to known signals or variables we know are important (Fig 12a). Indeed, we find that some components relate to both the onset and offset (rise and fall) of the clock signal(Fig 12b and 12c). This is quite interesting as we know that the processor uses a two-phase clock. We also find that a component relates strongly to the processors read-write signal (Fig 12d). Thus, we find that variables of interest are indeed encoded by the population activity in the processor.
In neuroscience, it is also frequently found that components from dimensionality reduction relate to variables of interest [56, 57]. This is usually then seen as an indication that the brain cares about these variables. However, clearly, the link to the read-write signal and the clock does not lead to an overly important insight into the way the processor actually processes information. Similar questions arise in neuroscience where scientists ask if signals, such as synchrony, are a central part of information processing or if they are an irrelevant byproduct [58]. We should be careful at evaluating how much we understand and how much we are aided by more data.
Pondering the results of the processor analysis we can obtain some insights into the developments needed to better utilize dimensionality reduction towards an understanding. The narrow range of games that we considered and the narrow range of their internal states (we just simulated booting), means that many aspects of computation will not be reflected by the activities and hence not in the dimensionality reduction results. Moreover, the fact that we used linear reduction only allows for linear dependencies and transistors, just like neurons, have important nonlinear dependencies. Lastly, there is clearly a hierarchy in function in the processor and we would need to do it justice using hierarchical analysis approaches. The results of dimensionality reduction should be meaningful for guiding new experiments, necessitating transfer across chips in the same way as neuroscience experiments should transfer across animals. Importantly, the chip can work as a test case while we develop such methods.
Here we have taken a reconstructed and simulated processor and treated the data “recorded” from it in the same way we have been trained to analyze brain data. We have used it as a test case to check the naïve use of various approaches used in neuroscience. We have found that the standard data analysis techniques produce results that are surprisingly similar to the results found about real brains. However, in the case of the processor we know its function and structure and our results stayed well short of what we would call a satisfying understanding.
Obviously the brain is not a processor, and a tremendous amount of effort and time have been spent characterizing these differences over the past century [22, 23, 59]. Neural systems are analog and and biophysically complex, they operate at temporal scales vastly slower than this classical processor but with far greater parallelism than is available in state of the art processors. Typical neurons also have several orders of magnitude more inputs than a transistor. Moreover, the design process for the brain (evolution) is dramatically different from that of the processor (the MOS6502 was designed by a small team of people over a few years). As such, we should be skeptical about generalizing from processors to the brain.
However, we cannot write off the failure of the methods we used on the processor simply because processors are different from neural systems. After all, the brain also consists of a large number of modules that can equally switch their input and output properties. It also has prominent oscillations, which may act as clock signals as well [60]. Similarly, a small number of relevant connections can produce drivers that are more important than those of the bulk of the activity. Also, the localization of function that is often assumed to simplify models of the brain is only a very rough approximation. This is true even in an area like V1 where a great diversity of co-localized cells can be found [61]. Altogether, there seems to be little reason to assume that any of the methods we used should be more meaningful on brains than on the processor.
To analyze our simulations we needed to convert the binary transistor state of the processor into spike trains so that we could apply methods from neuroscience to (see Methods). While this may be artefactual, we want to remind the reader that in neuroscience the idea of an action potential is also only an approximate description of the effects of a cell’s activity. For example, there are known effects based on the extrasynaptic diffusion of neurotransmitters [62] and it is believed that active conductances in dendrites may be crucial to computation [63].
Our behavioral mechanisms are entirely passive as both the transistor based simulator is too slow to play the game for any reasonable duration and the hardware for game input/output has yet to be reconstructed. Even if we could “play” the game, the dimensionality of the input space would consist at best of a few digital switches and a simple joystick. One is reminded of the reaching tasks which dominate a large fraction of movement research. Tasks that isolate one kind of computation would be needed so that interference studies would be really interpretable.
If we had a way of hypothesizing the right structure, then it would be reasonably easy to test. Indeed, there are a number of large scale theories of the brain [5, 64, 65]. However, the set of potential models of the brain is unbelievably large. Our data about the brain from all the experiments so far, is very limited and based on the techniques that we reviewed above. As such, it would be quite impressive if any of these high level models would actually match the human brain to a reasonable degree. Still, they provide beautiful inspiration for a lot of ongoing neuroscience research and are starting to exhibit some human-like behaviors [64]. If the brain is actually simple, then a human can guess a model, and through hypothesis generation and falsification we may eventually obtain that model. If the brain is not actually simple, then this approach may not ever converge. Simpler models might yield more insight—specifically seeking out an “adder” circuit might be possible, if we had a strong understanding of binary encoding and could tease apart the system to specifically control inputs and outputs of a subregion—examine it in slice, if you will.
The analytic tools we have adopted are in many ways “classic”, and are taught to graduate students in neuroinformatics courses. Recent progress in methods for dimensionality reduction, subspace identification, time-series analysis, and tools for building rich probabilistic models may provide some additional insight, assuming the challenges of scale can be overcome. Culturally, applying these methods to real data, and rewarding those who innovate methodologically, may become more important. We can look at the rise of bioinformatics as an independent field with its own funding streams. Neuroscience needs strong neuroinformatics to make sense of the emerging datasets and known artificial systems can serve as a sanity check and a way of understanding failure modes.
We also want to suggest that it may be an important intermediate step for neuroscience to develop methods that allow understanding a processor. Because they can be simulated in any computer and arbitrarily perturbed, they are a great testbed to ask how useful the methods are that we are using in neuroscience on a daily basis. Scientific fields often work well in situations where we can measure how well a project is doing. In the case of processors we know their function and we can know if our algorithms discover it. Unless our methods can deal with a simple processor, how could we expect it to work on our own brain? Machine learning and statistics currently lack good high-dimensional datasets with complex underlying dynamics and known ground truth. While not a perfect match, the dynamics of a processor may provide a compelling intermediate step. Additionally, most neural datasets are still “small data”—hundreds of cells over tens of minutes. The processor enables the generation of arbitrary complexity and arbitrarially-long timeseries, enabling a focus on scalable algorithms. We must be careful to not over-fit, but neuroscience is rife with examples of adopting analytic tools from vary different domains (linear system theory, stochastic process theory, kalman filtering) to understand neural systems.
In the case of the processor, we really understand how it works. We have a name for each of the modules on the chip and we know which area is covered by each of them (Fig 13a). Moreover, for each of these modules we know how its outputs depend on its inputs and many students of electrical engineering would know multiple ways of implementing the same function. In the case of the brain, we also have a way of dividing it into regions (Fig 13b, adopted from [66]). However, we only use anatomy to divide into modules and even among specialists there is a lot of disagreement about the division. Most importantly though, we do not generally know how the output relates to the inputs. As we reviewed in this paper, we may even want to be careful about the conclusions about the modules that neuroscience has drawn so far, after all, much of our insights come from small datasets, with analysis methods that make questionable assumptions.
There are other computing systems that scientists are trying to reverse engineer. One particularly relevant one are artificial neural networks. A plethora of methods are being developed to ask how they work. This includes ways of letting the networks paint images [67] and ways of plotting the optimal stimuli for various areas [68]. While progress has been made on understanding the mechanisms and architecture for networks performing image classification, more complex systems are still completely opaque [69]. Thus a true understanding even for these comparatively simple, human-engineered systems remains elusive, and sometimes they can even surprise us by having truly surprising properties [70]. The brain is clearly far more complicated and our difficulty at understanding deep learning may suggest that the brain is hard to understand if it uses anything like gradient descent on a cost function.
What kind of developments would make understanding the processor, and ultimately the brain, more tractable? While we can offer no definitive conclusion, we see multiple ways in which we could have better understood the processor. If we had experiments that would more cleanly separate one computation then results would be more meaningful. For example, lesion studies would be far more meaningful if we could also simultaneously control the exact code the processor was executing at a given moment. Better theories could most obviously have helped; if we had known that the microprocessor has adders we could have searched for them. Lastly, better data analysis methods, e.g. those that can explicitly search for hierarchical structure or utilize information across multiple processors. Development in these areas seems particularly promising. The microprocessor may help us by being a sieve for ideas: good ideas for understanding the brain should also help us understand the processor. Ultimately, the problem is not that neuroscientists could not understand a microprocessor, the problem is that they would not understand it given the approaches they are currently taking.
All acquisition and development of the initial simulation was performed in James [11]. 200°F sulfuric acid was used to decap multiple 6502D ICs. Nikon LV150n and Nikon Optiphot 220 light microscopes were used to capture 72 tiled visible-light images of the die, resulting in 342 Mpix of data. Computational methods and human manual annotation used developed to reconstruct the metal, polysilicon, via, and interconnect layers. 3510 active enhancement-mode transistors were captured this way. The authors inferred 1018 depletion-mode transistors (serving as pullups) from the circuit topology as they were unable to capture the depletion mask layer.
An optimized C++ simulator was constructed to enable simulation at the rate of 1000 processor clock cycles per wallclock second. We evaluated the four provided ROMs (Donkey Kong, Space Invaders, Pitfall, and Asteroids) ultimately choosing the first three as they reliably drove the TIA and subsequently produced image frames. 10 seconds of behavior were simulated for each game, resulting in over 250 frames per game.
Whole-circuit simulation enables high-throughput targeted manipulation of the underlying circuit. We systematically perturb each transistor in the processor by forcing its input high, thus leaving it in an “on” state. We measure the impact of a lesion by whether or not the system advances far enough to draw the first frame of the game. Failure to produce the first frame constitutes as a loss of function. We identified 1560 transistors which resulted in loss of function across all games, 200 transistors which resulted in loss of function across two games, and 186 transistors which resulted in loss of function for a single game. We plot those single-behavior lesion transistors by game in Fig 4.
Using the acquired netlist, we implement the authors method from [31] on the region of the processor consisting of the X, Y, and S registers. A nonparametric distance-dependent stochastic block model is jointly fit to six connectivitiy matrices: G → C1, G → C2, C1 → C2 C2 → C1, C1 → G, C2 → G, and via Markov-chain Monte Carlo, seeks the maximum a posteriori estmate for the observed connectivity.
We chose to focus on transistor switching as this is the closest in spirit to discrete action potentials of the sort readily available to neuroscientific analysis. The alternative, performing analysis with the signals on internal wires, would be analogous to measuring transmembrane voltage. Rasters were plotted from 10 example transistors which showed sufficient variance in spiking rate.
We compute luminance from the RGB output value of the simulator for each output pixel to the TIA. We then look at the transistor rasters and sum activity for 100 previous timesteps and call this the “mean rate”. For each transistor we then compute a tuning curve of mean rate versus luminance, normalized by the frequency of occurrence of that luminance value. Note that each game outputs only a small number of discrete colors and thus discrete luminance values. We used SI as it gave the most equal sampling of luminance space. We then evaluate the degree of fit to a unimodial Gaussian for each resulting tuning curve and classify tuning curves by eye into simple and complex responses, of which Fig 4 contains representative examples.
For the SI behavior we took spiking activity from the first 100ms of SI and performed spike word analysis on a random subset of 64 transistors close to the mean firing rate of all 3510.
To derive “local field potentials” we spatially integrate transistor switching over a region with a Gaussian weighting of σ = 500μm and low-pass filter the result using a window with a width of 4 timesteps.
We compute periodograms using Welch’s method with 256-sample long windows with no overlap and a Hanning window.
We adopt methods for assessing conditional Granger causality as outlined in [71]. We take the LFP generated using methods in section and create 100 1ms-long trials for each behavioral experiment. We then compute the conditional Granger causality for model orders ranging from 1 to 31. We compute BIC for all behaviors and select a model order of 20 as this is where BIC plateaus.
The transistor switching state for the first 106 timestamps for each behavioral state is acquired, and binned in 100-timestep increments. The activity of each transistor is converted into a z-score by subtracting mean and normalizing to unit variance.
We perform dimensionality reduction on the first 100,000 timesteps of the 3510-element transistor state vectors for each behavioral condition. We use non-negative matrix factorization, which attempts to find two matrices, W and H, whose product WH approximates the observed data matrix X. This is equivalent to minimizing the objective | | W H - X | | 2 2.
The Scikit-Learn [72] implementation initialized via nonnegative double singular value decomposition solved via coordinate descent, as is the default. We use a latent dimensionality of 6 as it was found by hand to provide the most interpretable results. When plotting, the intensity of each transistor in a latent dimension is indicated by the saturation and size of point.
To interpret the latent structure we first compute the signed correlation between the latent dimension and each of the 25 known signals. We show particularly interpretable results.
|
10.1371/journal.pgen.1000452 | Positive Feedback between Transcriptional and Kinase Suppression in Nematodes with Extraordinary Longevity and Stress Resistance | Insulin/IGF-1 signaling (IIS) regulates development and metabolism, and modulates aging, of Caenorhabditis elegans. In nematodes, as in mammals, IIS is understood to operate through a kinase-phosphorylation cascade that inactivates the DAF-16/FOXO transcription factor. Situated at the center of this pathway, phosphatidylinositol 3-kinase (PI3K) phosphorylates PIP2 to form PIP3, a phospholipid required for membrane tethering and activation of many signaling molecules. Nonsense mutants of age-1, the nematode gene encoding the class-I catalytic subunit of PI3K, produce only a truncated protein lacking the kinase domain, and yet confer 10-fold greater longevity on second-generation (F2) homozygotes, and comparable gains in stress resistance. Their F1 parents, like weaker age-1 mutants, are far less robust—implying that maternally contributed trace amounts of PI3K activity or of PIP3 block the extreme age-1 phenotypes. We find that F2-mutant adults have <10% of wild-type kinase activity in vitro and <60% of normal phosphoprotein levels in vivo. Inactivation of PI3K not only disrupts PIP3-dependent kinase signaling, but surprisingly also attenuates transcripts of numerous IIS components, even upstream of PI3K, and those of signaling molecules that cross-talk with IIS. The age-1(mg44) nonsense mutation results, in F2 adults, in changes to kinase profiles and to expression levels of multiple transcripts that distinguish this mutant from F1 age-1 homozygotes, a weaker age-1 mutant, or wild-type adults. Most but not all of those changes are reversed by a second mutation to daf-16, implicating both DAF-16/ FOXO–dependent and –independent mechanisms. RNAi, silencing genes that are downregulated in long-lived worms, improves oxidative-stress resistance of wild-type adults. It is therefore plausible that attenuation of those genes in age-1(mg44)-F2 adults contributes to their exceptional survival. IIS in nematodes (and presumably in other species) thus involves transcriptional as well as kinase regulation in a positive-feedback circuit, favoring either survival or reproduction. Hyperlongevity of strong age-1(mg44) mutants may result from their inability to reset this molecular switch to the reproductive mode.
| Insulin/IGF-1 signaling (IIS) impacts development, metabolism, and longevity in Caenorhabditis elegans. It has been viewed as a cascade of kinase reactions, chiefly phosphorylation of other kinases, leading to inactivation of the DAF-16/FOXO transcription factor. PI3K, a phosphatidylinositol kinase at the center of this pathway, converts PIP2 to PIP3, instrumental to kinase docking and activation. Here we show that PI3K deficiency elicits transcriptional inhibition of many kinases, including those of IIS itself. This creates a positive-feedback loop, wherein DAF-16/FOXO silences expression of the very kinases that would have inactivated it. In the resulting “flip-flop” genetic switch, either kinase signaling or transcriptional silencing may predominate. We discovered the transcriptional arm of this switch in infertile age-1(mg44) mutants, defective for PI3K activity. The absence of PIP3 and PIP3-dependent kinase activity gives free rein to gene silencing by DAF-16/FOXO. This two-tiered response could scarcely have evolved for the benefit of a sterile mutant; some components presumably serve regulatory functions in normal animals, reinforcing a switch responsive to environmental and internal signals. In age-1(mg44) mutants, complete inactivation of PI3K “fuses” the switch, locking worms into longevity mode. With signaling profoundly silenced, they cannot resume reproduction, but instead acquire a remarkable capacity for individual survival.
| The IIS pathway, governing developmental arrest, metabolism and life span in Caenorhabditis elegans [1]–[3], is highly conserved from invertebrates to mammals. The single IIS pathway of nematodes corresponds in structure and function to two distinct pathways of mammals that signal metabolic responses to insulin, and growth response to insulin-like growth factor 1 (IGF-1), respectively [4]. IIS disruption was first discovered to enhance longevity in C. elegans [5]–[8], but it was subsequently shown to also extend life in D. melanogaster and mice [9]–[11]. Binding of insulin-like peptides to DAF-2, the insulin/IGF-1 receptor of nematodes, modulates receptor autophosphorylation and activation [12]. Active DAF-2 recruits and phosphorylates the AGE-1 catalytic subunit of phosphatidylinositol 3-kinase (PI3K), which in turn phosphorylates the regulatory subunit. Activated AGE-1 then adds a phosphate to phosphatidylinositol 4,5-diphosphate [PI(4,5)P2] at the inositol-ring 3-position, converting it to phosphatidylinositol 3,4,5-triphosphate [PI(3,4,5)P3 or PIP3].
PIP3 plays a dual role in the canonical insulin/IGF-1 pathway. The first pivotal role is membrane tethering of many signaling molecules including AKT-1 and -2, PDK-1, GSK-3 and protein kinase C [13]–[16]. PIP3-binding recruits or retains many kinases at the cytoplasmic surface of the cell membrane, where these enzymes and their substrates (largely other kinases) are concentrated and, by mass action, interact more efficiently. Because PIP3 quantitatively affects multiple components of the IIS cascade, the influence of its concentration is compounded. In addition, PIP3 binding to AKT-1 allosterically exposes a cryptic site recognized by PDK-1 (phosphatidylinositol-dependent kinase 1), allowing AKT phosphorylation and activation [17]. In this second role, PIP3 may act catalytically, in that a single molecule of PIP3 has the potential to bind successively to many AKT-1 molecules, enabling their activation. Although AKT-1 is the only target for which this allosteric role has been documented [17], it is possible that other signaling molecules that also possess high-affinity PIP3 binding sites (termed “Pleckstrin homology domains”) may be similarly controlled. In any event, we infer that insulinlike signaling should be exquisitely sensitive to PIP3 depletion, and that AKT-1 action (which extends far beyond IIS [18],[19]) may be absolutely dependent on the presence of at least trace amounts of PIP3.
The AKT-1/AKT-2/SGK-1 complex, once all of its constituent kinases have been activated by PDK-1 [20], phosphorylates the DAF-16/FOXO transcription factor at sites that block its entry into the nucleus, where it would activate or repress transcription of hundreds of target genes, including many that modulate metabolism, reproduction, life span, and resistance to oxidative stresses [21]–[24].
Reduction-of-function mutations impairing the C. elegans IIS pathway (e.g., daf-2 and age-1 mutations) cause these worms to arrest development as dauer (alternative stage-3) larvae [1]–[3]. If allowed to mature at a permissive temperature, temperature-sensitive (ts) daf-2 mutant adults can attain twice the normal longevity [6]; life extension ranges from 1.1- to 2.5-fold for different daf-2 alleles [25]. A ts mutant allele of age-1, hx546, was discovered by Klass [26] and reported to confer 40% and 65% life extension at 20° and 25°C respectively [5],[27],[28]. Two constitutive age-1 alleles, m333 and mg44, were initially reported to extend C. elegans life span by 2- to 2.6-fold [2],[8],[29]; these survivals were conducted only for first-generation (“F1”) homozygotes. We recently observed that second-generation age-1(mg44) and (m333) larvae slowly mature at 15–20°C into adults that live close to ten times as long as near-isogenic wild-type controls, and are highly resistant to oxidative and electrophilic stresses [30]. These exceptional worms have mean and maximal adult life spans at least three times those conferred by any other longevity-extending mutation, and throughout their adult lives they appear and behave very much like wild-type worms of a tenth their age. Addition of a second mutation in the daf-16 gene largely or entirely reverses life-span extension and other phenotypes of all daf-2 or age-1 mutations examined to date [3],[6],[29],[30].
Studies of IIS-pathway mutants in C. elegans and other taxa have provided valuable insights into genetic mechanisms regulating life span [4]. The molecular basis for the extreme survival phenotypes of age-1(mg44) F2 homozygotes remains unknown, and cannot be assumed to differ only in degree from molecular mechanisms that underlie 4- to 5-fold lesser life extensions seen in other IIS mutants. The key may be PIP3, which plays both structural and catalytic roles in signal transduction [17],[31], and is thought to mediate both DAF-16-dependent and -independent signaling [32]. Strong age-1 mutants, lacking all class-I PI3K activity, have no direct route to generate PIP3 [31]. As a result, they are expected to be deficient in all enzyme activities that require PIP3, either for activation by regulatory kinases, or for membrane tethering which ensures proximity of kinases to their targets [17],[31].
In the present study, we sought evidence to support such a broad role of PIP3 in the unique properties of age-1(mg44)-F2 adults. This role is an inferred one, since even normal PIP3 levels (in unstressed N2 worms) are too low for detection by existing methods; detectable levels are attained in starved, peroxide-stressed wild-type worms but not in similarly stressed age-1-mutants [33]. We were able to document the expected widespread disruption of protein kinase activity in age-1(mg44)-F2 worms, while making the unexpected observation that the same kinases are chiefly inhibited at the transcriptional level. Direct measurement of transcripts confirms silencing of kinase gene expression, leading us to propose a novel “hybrid” positive-feedback loop in which the IIS kinase cascade that inhibits the DAF-16/FOXO transcription factor, is itself attenuated by DAF-16-mediated transcriptional silencing of upstream kinases.
The age-1(mg44) kinase-null mutants should be deficient in phosphatidylinositol 3,4,5-triphosphate production. Given the importance of the PIP3 molecule in signal transduction events originating from many membrane-receptor kinases, we anticipated that phosphorylation of numerous proteins may be impaired in those mutants. To initially assess the breadth of this impairment, we compared in vitro kinase activities with respect to endogenous substrates for five age-1 mutant strains, each normalized to a wild-type N2DRM stock (Figure 1A–1C).
Panels A and B illustrate a typical experiment, and panel C summarizes results for replicate experiments with independent expansions of each group. The first-discovered and most widely used age-1 allele, hx546 [5],[27],[28], showed 32% less kinase activity than N2DRM (Figure 1C). However, worms bearing the age-1(mg44) allele had less than 10% of wild-type kinase activity, whether maternally protected first-generation (F1) or very long-lived second-generation (F2) homozygotes. The F2 worms had somewhat lower kinase activity than F1 (7.3 vs. 8.6% of N2DRM, P = 0.05), although the difference was consistently much greater for specific bands (see Figure 1B). Staining of total protein showed similar loads for all samples, although banding patterns differed (Figure 1A). One obvious difference between age-1(mg44) and the other strains is that these mutants are totally infertile in the F2 generation, despite the presence of syncytial nuclei [30]. Several controls exclude this as an explanation for the mutants' lack of kinase activity. age-1(mg44) F1's have similar kinase levels when gravid (day 2 of adulthood) or post-gravid (adult day 6; see Figure S1). Moreover, N2DRM eggs contain about half as much kinase activity as their parents, per weight of protein (Figure S1), so their absence would not reduce kinase activity in any case. Deficiency of kinase activity is not a characteristic of dauer larvae, which exhibit a level comparable to that of N2DRM adults (Figure S1).
The consequences of adding a daf-16 mutation are quite different for the two age-1 alleles: in age-1(hx546) worms, the daf-16(m26) mutation more than doubled the in vitro kinase activity, from 68% of wild-type to 160%, whereas this mutation restored less than half of the kinase deficiency due to the age-1(mg44) allele. Insofar as kinase suppression is reversed in daf-16; age-1(mg44) double mutants, we infer that activity is inhibited in part through the DAF-16/ FOXO transcription factor. However, reversion is far from complete, by either the m26 (point-mutant) or the mu86 (large-deletion) allele of daf-16 (see Figure 1C and Figure S1). This implies that a large proportion of observed kinase silencing is DAF-16-independent—perhaps reflecting direct effects of PIP3 depletion on kinases other than AKT, or AKT targets other than DAF-16/FOXO.
To corroborate low protein-kinase activity of age-1(mg44) adults, and to distinguish whether they are deficient for many protein kinases or a few very active ones, we constructed arrays of 70 synthetic peptides comprising 50 near-consensus kinase sites from the C. elegans proteome and 20 from mouse or human proteins. Phosphorylation in vitro was observed on 29 peptides, representing potential substrates for at least 18 distinct kinases (Figure S2 and Table S1). Protein kinase activity in extracts from age-1(mg44) F2 adults was reduced by 1.8- to >8-fold, relative to isogenic N2DRM postgravid worms (each at nominal P<0.05), for 22 of the 29 kinase targets that were phosphorylated in vitro. Addition of the daf-16(mu86) mutation produced essentially complete reversion, or hyper-reversion (activity>N2DRM), for 17 of those 22 peptides.
In view of the reduced protein-kinase activity of age-1(mg44) worms, we anticipated that their steady-state level of protein phosphorylation would also be depressed. To assess this, phosphoproteins were separated by acrylamide gel electrophoresis and compared among wild-type and age-1-mutant strains of C. elegans (Figure 1D–1F). Total protein staining (panel D) demonstrated even loading, while panel E shows the same gel stained with Pro-Q Diamond to detect and quantify phosphoproteins. Results for three replicates (independent expansions of each strain) are summarized in panel F. Relative to wild-type N2DRM, age-1(hx546) worms had 16% less phosphoprotein staining (marginally significant at P<0.05), while age-1(mg44) homozygous F2 adults showed a 41% reduction in steady-state phosphoprotein level (P<0.001). The daf-16(m26) mutation restores either allele to ∼92% of the N2DRM level. This finding is also supported by 2-D dual-fluor phosphoprotein gels (Figure 2), in which 72% of the phosphoprotein spots resolved (1199/1669) were reduced at least twofold in F2 age-1(mg44) adults relative to N2DRM. The deficiency of total phosphoprotein content is less pronounced than that of protein kinase activity, in age-1(mg44)-homozygous F2 adults, which is not surprising given that phosphoprotein levels reflect the steady state, i.e., a balance between kinase and phosphatase activities. We present evidence (next section) that the PTEN phosphatase is indeed downregulated in age-1(mg44).
F2 homozygotes for age-1(mg44) are expected to produce only truncated class-I PI3KCS, lacking the kinase domain and C-terminus of the protein. These worms indeed lack the main bands recognized by antibodies to the AGE-1 C-terminal region (Figure S3); residual bands may represent class-II and -III homologs of AGE-1. PIP3, formed exclusively by class-I PI3K, should thus be greatly reduced or absent. PIP3 is strictly required for PDK-1 activation of AKT kinase, which then phosphorylates and inactivates DAF-16/FOXO. Kinases that require PIP3 binding for membrane tethering or kinase activation [17],[31], such as AKT, PDK-1, and SGK-1, are expected to show marked suppression of activity, which cannot be directly reverted by a daf-16 mutation. The surprising observation that mutations to daf-16 restore nearly half of the age-1 (mg44)-F2 kinase deficiency, and >70% of its phosphoprotein deficit, implies that their inhibition must be mediated in part by DAF-16/FOXO. Such regulation could be direct (DAF-16 suppresses transcription of many kinase genes) or indirect (DAF-16 suppresses one or a few kinases, or stimulates one or a few phosphatases, which then suppress other kinases by impeding or opposing their phosphorylation). To test direct effects of DAF-16/FOXO, we used real-time polymerase chain reaction (RT-PCR) to quantify the effects of age-1 alleles, with or without added inactivation of daf-16, on transcript levels for IIS genes and a panel of other signaling components, representing a wide range of transduction pathways.
F2 age-1(mg44) adults, which are 4- to 5-fold longer-lived than F1 adults (comparing data of [8],[29] to [30]), also outperform their parents with respect to resistance to oxidative and electrophilic stresses (Figure 5A and 5B). Relative to N2DRM controls, survival in 5% hydrogen peroxide is extended 2-fold in F1 adults, but 10-fold in their F2 progeny. Because age-1(mg44)-F2 adults barely reached 20% mortality after 24 h, by which time 100% of worms had died in all other groups, survival time is here compared at a threshold of 20% mortality. Survival of an electrophilic stress, 4-HNE (similarly defined as time to 20% mortality) increased 1.6-fold in age-1(mg44) F1 worms but 5-fold at F2, with reference to N2DRM. Although resistance to these stresses is restored almost to wild-type levels in double mutants with daf-16, we note that reversion is not quite complete, whether using the weaker daf-16(m26) allele [30] or the daf-16(mu86) deletion allele (Figure 5A), indicating that such stress-resistance traits are mediated in part by a DAF-16-independent pathway. These results parallel the incomplete reversion, in daf-16; age-1(mg44) double mutants, seen for in vitro kinase activity and phosphoprotein levels (Figure 1) and for transcript levels of several genes (Table 1).
Most or all of the age-1(mg44)-downregulated genes are essential for nematode growth and development. That is, double-stranded RNAs (dsRNAs) targeting them, administered to developing C. elegans, produce embryonic lethality or larval arrest [54]–[56]. The impact of such knockdown, however, has thus far remained largely untested in adults. Because resistance to oxidative stresses is a common feature of many long-lived C. elegans mutants [57], and in particular parallels longevity in the age-1 allele set studied here (Figure 5 and [30]), we employed it as a short-term assay to evaluate the contribution of individual-gene downregulation, to the exceptional survival of age-1(mg44)-F2 adults in both benign and toxic environments.
Hydrogen peroxide resistance was measured in duplicate experiments, for wild-type N2DRM worms that had been exposed to dsRNA-expressing bacteria targeting 10 genes for which transcript levels are markedly reduced in age-1(mg44)-F2 adults. Genes were selected from among those not directly involved in the IIS pathway, but representing a variety of other signaling pathways, and for which RNAi constructs were available from the Ahringer library [56]. E. coli, either harboring an empty-vector control or expressing one of 10 gene-targeted dsRNA species, were fed to mature adults (days 3 through 6 after the L4/adult molt) so as to preclude effects on development. Survival curves, during subsequent exposure to 5-mM H2O2, are shown in Figure 5C. RNAi for four of the ten genes (encoding a transcription factor and three components of distinct protein-kinase signaling cascades) produced highly significant gains in peroxide survival (each P<0.001), and a fifth dsRNA exposure offered marginally significant protection (vps-34, P<0.03). The remaining five dsRNA treatments had no discernible effect on survival, compared to worms exposed only to the empty expression vector. The above results were reproduced in an independent experiment, with the same four genes attaining P<0.001, while vps-34 achieved P<0.08.
Genes (and encoded proteins) for which RNAi knock-down conferred a protective effect were daf-3 (SMAD transcription factor) and daf-4 (TGF-β receptor, a Ser/Thr kinase), both involved in TGF-β signaling; aak-1 (AMP-dependent protein kinase 1), part of the AMPK/TOR pathway; let-60 (RAS-family GTPase activating MAPK), part of the ERK-MAPK pathway, and vps-34 (class-III PI3KCS), involved in vesicular trafficking and autophagy. None of these individual RNAi effects matched the peroxide survival of untreated age-1(mg44) F2 adults at 62 days of adult age (large diamond symbols, Figure 5C). These data demonstrate that transcript-level changes seen in age-1(mg44) F2 homozygotes favor oxidative-stress survival. They may also contribute incrementally to the greatly enhanced longevity of F2 homozygotes, but we have not been able to confirm such effects. When begun at the end of larval development, RNAi to aak-1 and daf-4 extended survival by 7–11%, while let-60 dsRNA reduced it by ∼12% (data not shown). Such small effects on life span require large groups to reach significance; moreover, significance in one experiment provides little assurance that independent replicates will attain significance. This may reflect the low statistical power inherent to survivals of modest size, and/or inability to control environmental variance among experiments.
First-generation homozygotes for the age-1(mg44) mutation develop normally into fertile adults at 15–25°C, and display stress resistance and life extension typical of many IIS mutants [8],[14],[29]. In contrast, their progeny—second generation homozygotes—develop slowly at 15–20°C, to form infertile adults that are far more stress resistant and at least four-fold longer lived than other IIS mutants [30]. Maternal protection (oocyte carryover of age-1 mRNA, AGE-1 kinase or its PIP3 product, synthesized by the heterozygous parent) is thought to blunt both the stress-resistance and longevity traits to approximately those of a weaker age-1 or daf-2 allele [30]. We here show that total kinase activity is also attenuated more severely in age-1(mg44) adults at the F2 than the F1 generation (see Figure 1C and Figure S2), although several kinase substrates remain unaffected by this mutation (Table S1 and Figure S2). At the expression level, the situation is more complex: the majority of tested genes are most strongly affected in the F2 generation of age-1(mg44), whereas others show transcript effects in F1 adults that equal, exceed, or even oppose those observed in F2 adults (Table 1).
Transcriptional effects within the IIS pathway seem fully consistent with impaired insulinlike signaling, which might be expected to further augment survival through the same mechanisms employed by weaker IIS mutations. Moreover, repression of class-II and class-III PI3K catalytic-subunit genes (Table 1) would impede formation of PI(3)P and PI(3,4)P2, suppressing alternative routes to PI(3,4,5)P3. Increased expression of aak-2 contributes to activation of DAF-16/FOXO, further opposing IIS (which inhibits this transcription factor) and increasing life span [43],[44].
In addition to effects on IIS genes, however, age-1(mg44)-F2 adults also show striking transcriptional attenuation of several other signal transduction pathways that interact with IIS and with one another. TGF-β endocrine/paracrine signaling is active in development, and modulates several other signaling pathways including p38/MAPK and ERK/MAPK [46]. Both daf-1 and daf-4, encoding type-I and -II TGF-β receptors, respectively, are downregulated 4- to 5-fold in age-1(mg44)-F2 adults. Expression is also reduced 3-fold for daf-3, encoding a co-SMAD transcription factor deployed by several pathways including TGF-β. Perhaps in partial compensation for this signaling downregulation, the daf-7 gene encoding a TGF-β-family ligand/agonist is 5-fold upregulated. Silencing of TGF-β signaling, by RNAi directed at daf-3 or daf-4, improves survival in the presence of hydrogen peroxide, consistent with a prior observation that daf-1, -4 and -7 mutants are long-lived [46].
AMPK/TOR signaling has been implicated in innate immunity and stress responses. Although it remains controversial whether the primary response is to the microbe or to the stress it causes [47], both could be secondary to its role in nutrient sensing [37],[43]. AAK-1 and AAK-2 are regulated by the PAR-4 transcription factor [37],[58],[59], and aak-1 knockdown by RNAi confers resistance to oxidative stress (Figure 5C). Inhibition of the C. elegans TOR pathway confers stress resistance and extends life span [60],[61].
ERK/MAPK signal-transduction is essential for many developmental processes; because the constituent genes are also expressed in adult nematode tissues, they are presumed to have post-developmental functions not yet defined [62]–[64]. All six genes tested in this pathway are markedly downregulated, by 3- to >6-fold (Table 1), and RNAi inhibition of let-60 (encoding a RAS membrane co-receptor that initiates ERK/MAPK signaling) significantly improves survival of oxidative stress (Figure 5C).
The most dramatic effects of gene mutations on life span have involved hypomorphic (loss-of-function) mutations, and the genes affected have been termed aging “master genes”. The genes encoding IIS components provide the best-studied example. IIS, in common with many “master genes” and essentially all signaling pathways, regulates numerous other genes. In the case of IIS, a number of these are modulated in ways that are protective, or otherwise conducive to long life, such as upregulation of GSTs and other detoxification genes, which are among the “foot soldiers” of longevity assurance [65],[66]. However, we should not expect all such downstream consequences to confer uniformly pro-longevity effects; each gene is likely to serve several “masters”, and its level of expression will depend on the genetic, environmental, and signaling context. In keeping with this perspective, the downstream manifestations of longevity assurance genes are far less conserved, both in evolution and between distinct physiological states of a given species, than are the over-arching pathways and the functions they serve [67].
Improved stress resistance and survival of age-1(mg44) F2 worms, apparently arising from transcriptional attenuation of signaling pathways presumed to be protective, poses an intriguing paradox. These pathways, activated by nutrient deficiency, pathogens, or growth factors, have been reported to cross-talk with IIS at diverse levels [19], [43]–[47],[68],[69]. This suggests a complex fabric of signaling interactions, for which the impact of silencing multiple components cannot be predicted. Moreover, signaling that promotes survival in a variable or hostile setting may entail energy costs and harmful side-effects that would be unwarranted in a constant, pathogen-free environment with abundant food. An organism that avoids the deleterious aspects of these surveillance systems may thus reap survival benefits under benign conditions.
In several instances, the expression changes seen in strong-age-1 mutants appear to oppose their longevity or stress-resistance, based on the effects of down- or upregulation previously reported for the same genes. For example, age-1(mg44) F2 adults downregulate daf-18, which encodes the PIP3 3-phosphatase, PTEN. This would be expected to elevate the steady-state level of PIP3, thereby enhancing IIS and reducing longevity of normal worms. However, in the absence of AGE-1/PI3KCS kinase activity, there may be little or no PIP3 substrate on which PTEN could act. A second example is downregulation in age-1(mg44) F2 adults of skn-1, encoding a transcription factor responsive to oxidative damage and regulated via IIS [70]–[72]. Reduced expression of skn-1 seems at odds with increased oxidative-stress resistance and longevity; however, these very long-lived worms may generate lower levels of reactive oxygen species, thereby reducing skn-1 induction. RNAi to vps-34 (encoding a class-III PI3KCS required for vesicular trafficking and autophagy [73]) was recently shown to block life extension of eat-2(ad1116) and daf-2(mu150) mutants, although not of wild-type worms [58]. Autophagy is induced by TOR deficiency [58], and several TOR signaling components are downregulated in F2 worms (Table 1). Considering this, autophagy should be at least moderately induced in those worms, and its absence would not account for low expression of vps-34. These results argue against any direct role of vps-34 attenuation in the exceptional longevity of age-1(mg44) F2 worms. The possibility remains, however, that vps-34 downregulation could reinforce PIP3-depleting effects of a strong age-1 mutation. Downregulation in age-1(mg44) worms, of transcripts for let-60/RAS and five other members of the ERK-MAPK cascade, might be expected to oppose additional life extension beyond that typical of IIS mutants, because a let-60 gain-of-function mutation enhances daf-2 life extension [74]. RNAi targeting smk-1, encoding a transcriptional coactivator shared by DAF-16 and PHA-4 [75], reduces stress-resistance and lifespan of daf-2(e1370) worms [76], whereas the effect on wild-type worms is controversial [34],[76]. Although smk-1 knockdown impairs sod-3 expression in daf-2 worms [34], we found 9-fold elevation of sod-3 transcripts in the face of a 72% drop in smk-1 expression in age-1(mg44) (Table 1). Finally, sir-2.1 overexpression was reported to extend lifespan, and knock-down to shorten it [77],[78], whereas we found almost 4-fold downregulation of sir-2.1 in age-1(mg44)-F2 adults.
Although contradictory in the context of extreme stress resistance and longevity, all six of these “exceptions” are also mirrored, in most cases to a lesser degree, in dauer larvae (Table 1), a robust state of developmental arrest that can endure for months without reducing adult life span [1],[79],[80]. This raises the possibility that for these genes, the effects of downregulation are context-dependent, and may be beneficial in worms that are already highly protected from stress and aging. Alternatively, these expression changes may follow from regulatory mechanisms shared by age-1(mg44) adults and N2 dauers, and yet work in opposition to their robustness. This is plausible in the case of a severe loss-of-function mutant, effects of which are not orchestrated, but is difficult to reconcile with a highly-evolved alternative developmental state such as the dauer larva. Perhaps, rather than a single coherent program, the patterns we observe reflect aberrant triggering in the adult of one or more regulatory mechanisms that are normally utilized in developmental or metabolic regulation. This particular combination of mechanisms could be the serendipitous result of a profound alteration in PIP3 levels which in turn impacts multiple pathways.
The expression profile of age-1(mg44) worms depends largely on DAF-16/FOXO, consistent with prior evidence that C. elegans IIS operates mainly through this transcription factor, impacting several hundred target genes [2],[21],[22],[29],[52],[81]. Although DAF-16/FOXO has been regarded largely as a transcriptional activator [21],[82], it also effects negative regulation of many genes [24]. In our selected panel of genes, two-thirds (22/33) of the DAF-16-mediated effects of age-1(mg44) mutation involve reduced transcript levels, indicating that silencing prevails for gene transcripts that encode kinases and other mediators of intracellular signaling. Twenty-eight genes (of the 33 for which transcripts appear to be primarily regulated via DAF-16/FOXO) were mapped for DAF-16 binding sites within 5 kb upstream of the initiation codon (Table 1). Of these, 21 (75%) have exact matches to one or both of the two known consensus sites, GTAAA(C/A)AA and CTTATCA. Genes lacking such sites may be indirect targets of DAF-16/FOXO, but considering that those motifs occur at almost the same frequency in the genome at large [81], as in DNA immunoprecipitated with antibody to DAF-16/FOXO [52], it is possible that precise motif matches are neither necessary nor sufficient for DAF-16 binding. In other words, near-match sequences might be able to bind DAF-16, while even perfect-match motifs may require additional features in nearby DNA.
It is surprising that hyperactivation of DAF-16/FOXO in age-1(mg44) F2 adults silences essentially the entire IIS pathway. This implies a positive feedback loop, in which DAF-16/FOXO imposes transcriptional silencing on the very kinases that would inhibit its own nuclear localization and hence access to target genes (Figure 6). We propose that second-generation age-1(mg44) homozygotes are trapped in a nonadaptive state, incapable of responding to diverse environmental and internal signals. This apparent paradox, that failure of adaptive mechanisms greatly extends lifespan, is easily explained because those mechanisms maximize Darwinian fitness – transmission of genetic alleles to ensuing generations – rather than individual survival [83],[84].
When IIS kinase signaling predominates (the reproductive state), it suppresses DAF-16/FOXO activity. Activation of PI3K favors PIP3 production and AKT activation, both of which promote cell proliferation [18],[31],[85]. However, IIS can switch to a second, functionally distinct, state: when kinase signaling is weak, DAF-16/FOXO becomes activated. As we have demonstrated, active DAF-16/FOXO transcriptionally silences its own upstream regulatory kinases, which otherwise would have impeded DAF-16/FOXO action by preventing its nuclear localization. Therefore, the low-signaling, longevity state of IIS is self-sustaining. Biologically, this state promotes dauer formation during development, or life-extension and delayed reproduction in the adult (reviewed in [4]). Signals that inhibit IIS kinases or augment DAF-16/FOXO action, if sufficient, trigger a switch from reproductive to longevity state in which DAF-16/FOXO promotes somatic protective mechanisms (Figure 6). However, exiting the stable longevity mode requires a shift in the balance of inputs that govern the positive feedback loop. Such inputs may include insulin-like peptide agonists and antagonists, hormones, pheromones, transcriptional co-activators and co-repressors of DAF-16/FOXO such as SIR-2 and 14-3-3 proteins, and nutrient- and stress-sensors signaled though other kinase pathways (e.g., MAPK, JNK and AMPK) that cross-talk with IIS. Combined, these two normal states of the IIS pathway (reproductive and longevity) constitute a “flip-flop” circuit with opposing kinase-cascade and transcriptional signals (Figure 6).
The concept of a “genetic switch” for dauer formation is not new [1],[86],[87], and has even been demonstrated to constitute a bistable feedback loop [21],[87]. Nevertheless, a dual-level (kinase/transcriptional) feedback mechanism had not previously been proposed or described. Any such “flip-flop” circuitry allows the organism a simple binary choice in response to its environment: early reproduction under benign conditions, or postponed reproduction and extended survival in harsher conditions. Mutational disruption of IIS forces dauer formation, irrespective of environment, although larvae with temperature-sensitive mutations can mature at lower temperatures into long-lived adults. Recovery from the dauer state requires that pro-reproductive-state kinases retain partial function, so that favorable signals (restoration of food, absence of stress and crowding) can reset the switch to the reproductive mode; this requirement is demonstrated by the impaired post-dauer recovery of IIS-defective mutants [86]. Nonsense mutations truncating AGE-1 produce an extreme phenotype that forfeits this option, while acquiring a distinctive transcriptional profile and greatly enhanced survival. Details of the mechanism or mechanisms, by which elimination of PI3K activity blocks exit from the longevity mode and promotes extreme longevity, remain to be elaborated. Features described in this report, which may contribute, include transcriptional silencing of upstream and collateral signaling components, and accompanying loss of multiple kinase activities. Infertile mutants may thus reveal new strategies to extend life well beyond the limits imposed by natural selection, which of course requires reproduction. In view of the striking evolutionary conservation of the IIS pathway, and the emerging parallels between inter-pathway cross-talk in nematodes and mammals [16],[19],[47],[68],[69], the mechanistic insights afforded by very long-lived worms are likely to also apply to insulin and IGF-1 pathways of mammals.
Nematode strains, supplied by the Caenorhabditis Genetics Center (CGC, Minneapolis), or derived in our laboratory from CGC strains, were maintained at 20°C on 0.6% peptone NGM-agar plates seeded with E. coli strain OP50, as described [88]–[90].
Assays of survival in the presence of 5-mM hydrogen peroxide or 10-mM 4-hydroxynonenal were modified from Ayyadevara et al. [91]. Wild-type (N2DRM) worms were assayed at day 3–4 of adulthood (∼6 d post-hatch). For RNAi experiments, day-1 adults were washed in S-buffer [92] and transferred to nutrient-agar plates seeded with dsRNA-expressing E. coli [56]. After 3 days at 20°C, 20 worms from each RNAi treatment were transferred to 24-well plates containing 300 µl of S Buffer plus 5 µg/ml cholesterol, supplemented, as indicated, either with 5-mM H2O2 (freshly diluted from 30% H2O2, Sigma) or with 10-mM 4-HNE (freshly obtained by acid hydrolysis of 4-HNE dimethylacetal which was synthesized according to [93]. Survival was scored as described [30],[89].
Worms grown at 20°C were quickly frozen in liquid nitrogen to preserve endogenous kinase activity. Worms suspended in 50-mM Tris pH 7.5, 80-mM β-mercaptoethanol, 2-mM EDTA, 1-mM PMSF, and Protease Inhibitor Cocktail I (CalBiochem), were ground at −78°C and sonicated (VIRTIS Virsonic 475, setting 2.5, 0°C) in six 10-s bursts interspersed with 2-min cooling periods. Kinase activity toward endogenous substrates was assessed in cleared supernatants after centrifugation (10 min, 11,000 g), representing 20 µg protein in 100 µl of buffer containing 50-mM Tris pH 7.5, 12.5-mM MgCl2 and (for endogenous substrates) 8–10 µCi γ-32P-ATP (NEN). After 1 min at 30°C, quenched samples were electrophoresed on 10% SDS-polyacrylamide gels (Invitrogen), which were stained with SYPRO Ruby (Invitrogen), and dried under vacuum. 32P β-emissions of bands migrating slower than a 25-kDa protein marker (Invitrogen), were imaged and quantified per lane after 6-h phosphor-screen exposure (Storm, Molecular Dynamics). Peptide arrays were incubated as above, 60 min at 30°C, but with addition of phosphatase inhibitors and 1-mM cold ATP rather than 32P-ATP. Arrays were then stained with Pro-Q Diamond (Invitrogen), and phosphorylation was quantified by fluorescence imaging (excitation/emission at 550/580 nm) with a ScanArray 5000 (GSI Lumonics).
Total protein (20 µg), extracted from each strain as above, was loaded onto NuPAGE 4–12% gradient gels (Invitrogen) and electrophoresed 1 hour at 200 V. Phosphoproteins were quantified by Pro-Q Diamond (Invitrogen) fluorescence, which depends linearly on protein concentration (>1000-fold range). Protein load was assessed by Coomassie Blue (BioRad) staining. Phosphorylated (23.6, 45.0 kDa) and unphosphorylated (14.4, 18.0, 62.6, 116.2 kDa) protein standards (BioRad) furnished positive and negative controls.
Expression of selected genes was assessed by real-time polymerase chain reaction after an initial round of reverse transcription. Total RNA was purified from each strain (RNeasy, Qiagen), and cDNAs reverse-transcribed (SuperScript III, Invitrogen), followed by RT-PCR (Opticon2, MJ Research, using SYBR Green, Roche).
|
10.1371/journal.pgen.1006829 | Amino acid metabolites that regulate G protein signaling during osmotic stress | All cells respond to osmotic stress by implementing molecular signaling events to protect the organism. Failure to properly adapt can lead to pathologies such as hypertension and ischemia-reperfusion injury. Mitogen-activated protein kinases (MAPKs) are activated in response to osmotic stress, as well as by signals acting through G protein-coupled receptors (GPCRs). For proper adaptation, the action of these kinases must be coordinated. To identify second messengers of stress adaptation, we conducted a mass spectrometry-based global metabolomics profiling analysis, quantifying nearly 300 metabolites in the yeast S. cerevisiae. We show that three branched-chain amino acid (BCAA) metabolites increase in response to osmotic stress and require the MAPK Hog1. Ectopic addition of these BCAA derivatives promotes phosphorylation of the G protein α subunit and dampens G protein-dependent transcription, similar to that seen in response to osmotic stress. Conversely, genetic ablation of Hog1 activity or the BCAA-regulatory enzymes leads to diminished phosphorylation of Gα and increased transcription. Taken together, our results define a new class of candidate second messengers that mediate cross talk between osmotic stress and GPCR signaling pathways.
| Just as organisms must adapt to a challenging environment, cells must respond to chemical or physical changes that occur within the organism. Regardless of the environmental cue, all cells use molecular signaling pathways to respond to those changes. Many are transmitted by G protein-coupled receptors (GPCRs) or the high osmolarity glycerol (HOG) pathway. While these pathways have been studied independently for decades, less is known about how they coordinate with each other to carry out the proper response, particularly when conflicting signals are present. One way coordination can be achieved is through “second messenger” molecules produced by one pathway to regulate another pathway. Here, we identify candidate second messengers of osmotic stress by global metabolite profiling analysis of the yeast S. cerevisiae. We find that three branched-chain amino acid (BCAA) metabolites increase in response to osmotic stress and require the stress response MAPK Hog1. We show that these BCAA derivatives are necessary and sufficient to recapitulate the effects of osmotic stress on the GPCR pathway. Our results identify a new way that HOG and GPCR pathways communicate, and may in the future guide better treatment strategies for stress-related cell damage.
| Cells routinely experience changing and often unfavorable conditions in their environment. The ability to adapt to environmental stress and re-establish homeostasis is essential not only to the survival of a cell, but also to the well-being of the organism. The response to such physical or chemical stresses is mediated by well-defined signaling networks. For example, changes in nutrient availability switch signaling between the opposing target of rapamycin (TOR) and AMP-activated protein kinase (AMPK) pathways [1, 2]. Stressors such as UV irradiation, inflammatory cytokines, and osmotic shock promote signaling through activation of the p38 and c-Jun N-terminal Kinase (JNK) MAPK pathways [3, 4]. While much is known about the mechanisms of stress-dependent signaling, less is known about coordination between the stress response and other cell signaling processes. In this study, we investigate cross-talk between osmotic stress and G protein-coupled receptor (GPCR) signaling pathways.
Hyperosmotic stress causes water efflux and cell shrinkage in order to normalize the osmotic balance between the intracellular and extracellular space. Depending on the severity of the stress, cell shrinkage can lead to macromolecular crowding and alterations in cellular protein activity [5], the production of reactive oxygen species (ROS), DNA damage, cell cycle arrest, and apoptosis [6]. In addition to these negative effects, cells also initiate signaling events that promote adaptation. Most prominently, osmotic stress activates the MAPK p38, which in turn phosphorylates myriad downstream targets that coordinate osmotic stress adaptations. Targets of p38 include the transcription factor NFAT5, which promotes the expression of proteins associated with the synthesis and transport of osmolytes, antioxidants, and molecular chaperones [7, 8]. Such changes ensure the survival of the cell, and they are likely to have important consequences for other signaling pathways via cross-talk mechanisms.
As the largest receptor family in humans [9], GPCRs are likely targets of cross-pathway regulation. These receptors respond to a wide variety of homeostatic cues, such as hormones and neurotransmitters, as well as to environmental signals such as odors and light. They signal primarily through intracellular heterotrimeric G proteins, comprised of Gα and Gβγ subunits. G proteins in turn activate downstream effectors leading to the production of second messenger molecules, such as calcium or cAMP, which bind to and activate intracellular protein kinases. Another mechanism of GPCR signaling entails the direct activation of protein kinases upstream of MAPKs [10, 11].
The G protein α subunit is a molecular on/off switch for signaling processes. As such, it is likely to be a critical target for post-translational modifications that regulate GPCR signaling. In fact, several studies have shown that Gα proteins are phosphorylated, resulting in altered affinity for Gβγ subunits or guanine nucleotides [12–20]. In some cases, phosphorylation is the direct result of pathway activation, and thus constitutes a positive or negative feedback. In other cases, phosphorylation is triggered by parallel pathways, and thus constitutes a mechanism of signal coordination or cross-talk. Previously, we showed that Gα is phosphorylated in response to nutrient limitation [21]. Our focus here is Gα regulation by osmotic stress.
Identifying how environmental stress can promote post-translational modification of Gα subunits is necessary to fully understand the mechanisms by which the pathways are coordinated and integrated. Studying the response to environmental stress is often challenging however, due to the expression of multiple protein isoforms and differences in expression among various tissues and cell types. Given these complexities, much can be learned from the analysis of orthologous signaling processes in simpler eukaryotes.
The budding yeast Saccharomyces cerevisiae has a stress response pathway and a GPCR signaling pathway with component proteins that are evolutionarily conserved across eukaryotes. The High Osmolarity Glycerol, or HOG, pathway is comprised of a MAPK (Hog1), a MAPK kinase (Pbs2), and MAPK kinase kinases (Ste11, Ssk2/Ssk22). Upon activation, Hog1 phosphorylates cytoplasmic and nuclear proteins that aid in the restoration of osmotic equilibrium through osmolyte synthesis and the induction of stress response genes [22–25]. Hog1 is the yeast ortholog of mammalian p38 [26, 27].
Yeast use another, parallel MAPK pathway to initiate haploid cell fusion, or mating. This pathway is activated by pheromone binding to a GPCR to initiate exchange of GDP for GTP in the Gα subunit (Gpa1) and subsequent dissociation of Gα from Gβγ. Gβγ activates a MAPKKK (Ste11, shared by the HOG pathway), a MAPKK (Ste7) and a MAPK (Fus3, or Kss1). Once activated, Fus3 promotes transcription of genes to initiate cell mating [28]. Fus3 is the yeast ortholog of mammalian ERK1 and ERK2 [29–32].
In the present study, we use yeast as a model system to investigate how crosstalk regulates G protein signaling during osmotic stress. We have shown previously that osmotic stress dampens the pheromone response pathway, and does so by Hog1-dependent and Hog1-independent mechanisms [33]. We have also shown that glucose limitation dampens the pheromone response pathway, and does so by reducing intracellular pH [34]. The increase in proton concentration is detected by the G protein directly, resulting in increased phosphorylation of Gpa1 and a dampened mating signal. Additionally, we have identified a family of three kinases (Elm1, Sak1, and Tos3) and a PP1 phosphatase complex (Reg1/Glc7) as the molecular machinery responsible for phosphorylating and dephosphorylating Gpa1 [21]. We show here that Gpa1 is likewise phosphorylated in response to osmotic stress, and that phosphorylation of Gpa1 requires the same protein kinases, but does not entail any changes in intracellular pH. In a search for alternative mediators of cross-pathway regulation, we conducted an unbiased metabolomics screen and found that 2-hydroxy branched chain amino acid metabolites are produced in a salt- and Hog1-dependent manner. Finally, we show that these metabolites are necessary and sufficient to promote Gpa1 phosphorylation and dampen downstream signaling. We propose that these metabolites represent a new class of second messengers of the stress-responsive HOG pathway.
To understand how cells adapt to environmental stresses, we sought to identify conditions that impact pheromone signaling through the phosphorylation of Gpa1. We recently established that Gpa1 is phosphorylated by a family of three AMPK kinases (Elm1, Sak1, and Tos3), and dephosphorylated by the phosphatase complex Reg1/Glc7 [17, 21]. These proteins were previously shown to phosphorylate and dephosphorylate the yeast AMPK, Snf1 [35–37]. Snf1, is phosphorylated and activated in response to nutrient limitation, as well as heat shock, hyperosmotic shock, reactive oxygen species, ethanol, and changes in extracellular pH [38]. Accordingly, we asked whether the same environmental stressors would lead to phosphorylation of Gpa1. We treated wild-type cells with the indicated stressor in a 2-hour time-course (see Materials and methods), and analyzed cell lysates by western blot. As shown in Fig 1, Gpa1 and Snf1 were phosphorylated in all stress conditions tested (see also S1 Fig). However, among the stressors there were differences in the both the magnitude and duration of phosphorylation. In glucose-limiting conditions, approximately 90% of Gpa1 was phosphorylated by 2 minutes, with a gradual decline after 10 minutes (Fig 1A). Heat shock (at 42°C) also promoted rapid phosphorylation but slow dephosphorylation. Osmotic stress promoted slow phosphorylation, but fast dephosphorylation. Heat and osmotic stress also promoted the phosphorylation of Snf1, but the effects were comparatively weak and transient (Fig 1B and 1C) [38]. These data reveal that Gpa1, like Snf1, is phosphorylated in response to various stress signals. More broadly, the results indicate that physico-chemically distinct stimuli have a common ability to promote phosphorylation of two functionally distinct proteins, Snf1 and Gpa1.
It is well-established that the MAPK Hog1 is phosphorylated and activated in response to osmotic stress [22]. Hog1 is also activated by heat shock [39], cold stress [40], oxidative stress [41], and hypoxia [42]. Given that many of these conditions also lead to phosphorylation of Gpa1 and Snf1, we asked if Hog1 activation was required in either case. To this end, we replaced Hog1 with a mutant documented to lack catalytic activity, hog1K52R [43], and then treated the cells with 0.5 M KCl. Whereas Snf1 phosphorylation was unperturbed, the phosphorylation of Gpa1 was almost completely abrogated in the hog1K52R strain (compare Fig 1C, blue curve vs. red curve). It is unlikely that Hog1 phosphorylates Gpa1 directly, since the relevant site (Ser200) does not adhere to the MAPK consensus sequence (Ser/Thr-Pro) [17]. Thus, Hog1 catalytic activity is required for the salt-induced phosphorylation of Gpa1 but not Snf1. More broadly, these results implicate at least two distinct signaling pathways, and potentially two distinct second messengers, that mediate the response to osmotic stress.
One potential second messenger is pH. Indeed, it is well established that glucose limitation leads to a substantial decrease in intracellular pH (pHi) [44]. We have shown recently that Gpa1 is a pH sensor, and that pH-dependent changes in conformation result in phosphorylation of the protein [34]. Since other stressors trigger phosphorylation of Gpa1, we asked whether any of those conditions also cause a change in pHi. To that end we expressed the ratiometric fluorescent pH biosensor, pHluorin, in wild-type cells [34, 45, 46]. Consistent with earlier studies [34], we observed a decrease in pHi from 7.0 to 6.4 upon glucose limitation (Fig 1A, inset). In contrast, cells subjected to osmotic stress exhibited no change in pHi over the course of 60 minutes (Fig 1C and S1 Fig). These data indicate that low glucose and osmotic stress each promote Gpa1 phosphorylation, but glucose alone affects pHi. We therefore postulated the existence of an additional second messenger of the osmotic stress response.
The data presented above reveal that osmotic stress has no effect on pHi, yet is a potent inducer of Gpa1 phosphorylation. To identify potential second messengers of osmotic stress, we conducted a global, unbiased metabolomics analysis [47]. Based on results from the Gpa1 phosphorylation experiments above, we sought to identify metabolites that increased with osmotic stress and did so in a Hog1-dependent manner. To this end, we subjected wild-type and Hog1-deficient yeast cells to 0.5 M KCl for 20 minutes and then analyzed cell extracts by LC-MS/MS and GC-MS (Fig 2A). This analysis identified 296 distinct entities representing each major class of biochemical molecules—amino acids, peptides, carbohydrates, lipids, nucleic acids, vitamins and cofactors, and xenobiotics (S1 Table). Consistent with past findings, we found that the osmolytes trehalose [48] and glycerol [49] were induced substantially (32-fold and 2.5-fold, respectively) (Fig 2C). Using a comparable (2-fold) cut off, we identified an additional 26 metabolites that increased in response to osmotic stress, and 13 that increased in the presence of Hog1. Of these, only three required osmotic stress and Hog1 together (Fig 2B and 2C, S2 Table): 2-hydroxyisovalerate (HIV), 2-hydroxyisocaproate (HIC), and 2-hydroxy-3-methylvalerate (HMVA). All three compounds are 2-hydroxy carboxylic acid derivatives of the branched-chain amino acids (BCAAs) valine, leucine, and isoleucine, respectively (Fig 2C and 2D). Thus, our analysis points to 2-hydroxy BCAA derivatives as candidate second messengers of osmotic stress.
Our metabolomics study demonstrated that BCAA derivatives are produced in response to osmotic stress, and that their production requires Hog1 (Fig 2D). In principle, deleting Hog1 could alter the production of additional second messengers that may not have been detected in our metabolomics screen. However, as BCAA derivatives were the most robustly increased metabolites that met our criteria for osmotic stress, we examined the consequences of disrupting BCAA catabolism through the so-called Ehrlich pathway in yeast [50]. The first step in the Ehrlich pathway is transamination to a 2-keto acid by the branched-chain amino acid transaminases, Bat1 and Bat2. The second step is decarboxylation of the 2-keto acid to an aldehyde, which is subsequently converted to a fusel acid or fusel alcohol. The BCAA derivatives identified here retain the same carbon skeleton as the parent amino acids, suggesting the existence of an alternative metabolic route consisting of transamination followed by reduction to the 2-hydroxy acid (Fig 3A). Products of the Ehrlich pathway are exported from the cell by the ABC transporter Pdr12 [50, 51].
To test whether BCAA derivatives are required for phosphorylation of Gpa1 and/or Snf1, we deleted the BAT1 and BAT2 genes individually (Fig 3B). After osmotic stress, we observed a modest, but significant reduction in maximal phosphorylation of Gpa1 in the bat1Δ and bat2Δ mutants, as compared to wild-type cells (Fig 3B and 3C). As expected, Snf1 phosphorylation was unaffected (Fig 3B). Cells harboring deletion of both BAT genes are reported to be viable [52, 53]; however in our hands, bat1Δbat2Δ double mutants arose at a lower-than-predicted frequency after tetrad dissection and likely harbored suppressor mutations. As an alternative approach, we attempted to use a tetracycline-repressible BAT1 in a bat2Δ background. However, the doxycycline used to repress BAT1 expression also promoted the phosphorylation of Gpa1. Gpa1 phosphorylation was unaffected by loss of the transporter gene PDR12 (Fig 3D), suggesting other routes of removal or of further metabolism. Together these results indicate that either Bat1 or Bat2 is necessary for cell viability. Both proteins, as well as their catalytic products, are necessary for a full response to osmotic stress.
Our results indicate that Hog1 activity and BCAA catabolism are both needed for a full response to osmotic stress. In particular, we have shown that osmotic stress-dependent Gpa1 phosphorylation is reduced in mutants lacking Bat1 or Bat2, and is eliminated in cells lacking Hog1 catalytic activity. Based on these findings, we hypothesized that Hog1 phosphorylates one or more components of the BCAA pathway. Indeed, Bat1 has five MAPK consensus sites (S/TP), and Bat2 has three such sites. In support of our hypothesis, replacement of the MAPK consensus sites in Bat1 and Bat2 led to a significant reduction in Gpa1 phosphorylation (Fig 3B and 3E). However, there were no changes in the electrophoretic (phosphorylation-dependent) mobility of Bat1, Bat2, Bat15A, or Bat23A, either in the absence or presence of salt stress. There was also no effect of osmotic stress on Bat23A in cells lacking Bat1 (bat1Δ bat23A) or Bat15A in the absence of Bat2 (bat15A bat2Δ) (S2 Fig). Taken together, these results suggest that Hog1 does not target the transaminases, and instead plays an indirect role in promoting the production of BCAA derivatives. That role could be to induce phosphorylation, or regulate the transcription, of some other component of the metabolic pathway. One potential target is the reductase(s) (as yet unidentified) that converts the 2-keto acid to the 2-hydroxy acid.
An intracellular second messenger should, by definition, be sufficient as well as necessary to evoke the response of the extracellular first messenger. Having demonstrated that BCAA derivatives are necessary for a full response to osmotic stress, we tested the ability of the BCAA derivatives to promote phosphorylation in the absence of salt stimulus. To better enable these compounds to traverse the cell membrane, we grew the cells at pH 5, which is closer to the pKa of the metabolites. By favoring the protonated, uncharged species, the BCAA derivatives can more easily traverse the plasma membrane. Importantly, the lower external pH does not change the intracellular pH (Fig 4B, inset) [34]. Using this approach, we found that HIV, HIC, and HMVA promoted Gpa1 phosphorylation, but with varying efficacy. HIC showed the strongest effect while HIV had the weakest effect (Fig 4A and 4B). Addition of HIC (but not salt) to Hog1-deficient cells promoted the phosphorylation of Gpa1, consistent with the idea that BCAA derivative production is a consequence of Hog1 activation (Fig 4A, 4C and 4D). Snf1 was likewise unaffected, consistent with the idea that it is regulated by a distinct second messenger (Fig 4A). Taken together, these experiments indicate that BCAA derivatives are sufficient to promote the phosphorylation of Gpa1 and thus meet the criteria for second messengers of osmotic stress.
Our results so far show that BCAA derivatives promote the phosphorylation of Gpa1. We next attempted to delineate the mechanism by which BCAA derivatives act. We demonstrated previously that protons interact directly with the G protein α subunit, causing a conformational change that promotes its phosphorylation. Moreover, the pH dependent change is conserved in yeast and human Gα proteins [34]. We hypothesized that BCAA derivatives might likewise act by binding to the Gα subunit. To test this we collected 1H-15N 2D heteronuclear NMR spectra of Gα, both in the absence and presence of BCAA derivatives. These spectra allow for the detection of protons directly bonded to a 15N, including both backbone and side-chain NH resonances. As an NH resonance can be detected for every residue, with the exception of proline, the spectrum contains a “fingerprint” of the protein backbone and allows perturbations resulting from interactions to be detected on a per-residue basis. This approach is widely considered as a definitive method for detecting low- to intermediate-affinity binding of ligands to proteins [54]. Accordingly, we acquired the NMR spectra of 15N-enriched Gαi in its GDP-bound state, alone or in the presence of a 25-fold excess of individual BCAA derivatives. As shown in Fig 5, there were no significant peak shifts when BCAA derivatives were present (Fig 5A–5C). As a positive control, we acquired NMR spectra of Gαi-GDP at pH 6 and at pH 7. As shown in Fig 5D, a substantial number of peaks were shifted at the lower pH, consistent with proton-dependent conformational changes in Gαi. These results indicate that BCAA derivatives likely act on another component of the G protein signaling pathway.
Gpa1 is phosphorylated by the AMPK kinases Elm1, Sak1, and Tos3. Whereas Elm1 phosphorylates Gpa1 in a cell-cycle-dependent manner [17], Sak1 is responsible for phosphorylation during glucose limitation [21]. Our data presented above indicate that Gpa1 is likewise phosphorylated in response to osmotic stress. To determine which, if any, of the AMPK kinases mediates the response to osmotic stress, we compared Gpa1 phosphorylation in cells lacking each of the three kinases, alone or in combination. As shown in S3A Fig, deletion of ELM1 resulted in the greatest reduction of Gpa1 phosphorylation, while deletion of SAK1 had a comparatively small effect. We then performed the same experiment using BCAA metabolites in place of osmotic stress. As with salt stimulation, HIC promoted the phosphorylation of Gpa1 in cells, and phosphorylation was diminished in the elm1Δ mutant (Fig 6 and S3B Fig). Taken together, these results indicate that both the primary messenger (osmotic stress) and the putative second messenger (the BCAA derivatives) act through Elm1. More broadly, these results confirm a fundamental difference between glucose- and salt-dependent changes in the cell. While both conditions lead to Gpa1 phosphorylation, they lead to the production of two distinct second messengers (protons and BCAA derivatives) and to phosphorylation by two distinct protein kinases (Sak1 and Elm1).
We have shown that osmotic stress leads to a diminished pheromone response [33] and phosphorylation of the Gα protein (this work). Based on our model, the BCAA derivatives are responsible for many of the intracellular effects of osmotic stress signaling, including Gα phosphorylation. According to our proposed mechanism, the same metabolites should also dampen the response to pheromone. To test this hypothesis, we employed a transcriptional reporter assay using GFP under control of the FUS1 promoter, which is specific to the pheromone response pathway [55]. We then measured fluorescence in response to increasing concentrations of the α-factor mating pheromone, alone or in combination with KCl or the BCAA derivatives. Consistent with previous reports [33], osmotic stress dampened the pheromone response by approximately 40%. Consistent with our present model, the addition of HIV, HIC, or HMVA also led to a diminished response of up to 40% (Fig 7A). The capacity of each BCAA derivative to dampen transcription correlated directly with its ability to promote Gpa1 phosphorylation (Fig 4B). Deletion of the Gpa1 kinases conferred an elevated signal at all but the highest concentrations of pheromone. At 10 μM pheromone the mutant strain was less sensitive to KCl and HIC (a reduction of 27% and 26%) compared to wild type (35 and 41%, respectively). At low and intermediate concentrations, the mutant strain was less responsive to salt and largely unresponsive to the BCAA derivatives (Fig 7B). Thus, BCAA derivatives are produced in response to an osmotic stress stimulus and, by every measure used, approximates the biochemical effects of salt on Gpa1. By these criteria the BCAA derivatives could function as second messengers of the osmotic stress response pathway and account for part of the osmotic stress response program.
Here, we present several novel features of the pheromone response pathway that we believe will be generally applicable to other MAPK signaling systems. First, we show that multiple environmental stressors lead to G protein phosphorylation. Phosphorylation of Gpa1 is accompanied by attenuated signaling through the effector MAPK, Fus3 [21, 33, 34]. Second, we show that many of these same stressors trigger the activation of another MAPK, Hog1. When Hog1 is activated, Fus3 signaling is inactivated. Third, we present the results of a comprehensive screen for small molecule metabolite second messengers, and show that 2-hydroxy BCAA derivatives are generated in response to osmotic stress and Hog1 activation. We show further that these metabolites are sufficient to trigger Gpa1 phosphorylation and a dampening of the Fus3 pathway. Finally, we show that the protein kinase Elm1 is required for phosphorylation of Gpa1 in response to osmotic stress and by addition of the metabolites. These processes are clearly distinct from those reported previously for glucose stress, which leads to a decrease in cellular pH, direct binding of protons to the Gα subunit, and direct phosphorylation of Gα by Sak1. While the target of the BCAA metabolites remains to be identified, we have largely excluded the kinase and Gα subunit substrate as candidates.
Based on our findings, we propose that BCAA metabolites represent a newly described “second messenger” of stress-activated signaling. The concept of second messenger signaling stems from the work of Earl Sutherland in 1957 [56] when he discovered that the activity of liver phosphorylase is stimulated indirectly by hormones, requiring a “heat-stable factor” that was later identified as cAMP [57]. That work established a paradigm of cell signaling whereby a first messenger (e.g., hormone or neurotransmitter) activates a receptor on the cell surface (canonically a GPCR) and activation of an intracellular effector protein that produces the second messenger molecule. This process serves to greatly amplify the intracellular response since activation of one receptor can lead to the production of multiple second messenger molecules. Since the discovery of cAMP, several other second messengers have been identified, including cGMP [58], inositol trisphosphate [59, 60], diacylglycerol [61], and calcium [62, 63]. Each of these molecules was painstakingly identified through rudimentary biochemical methods. With advances in metabolomics technologies however, we now have the ability to examine a broad complement of small molecules in a single experiment.
In yeast, BCAAs are catabolized through the Ehrlich pathway. The end products of this pathway are fusel alcohols or fusel acids [50]. Much like the catabolism of BCAAs by the Ehrlich pathway in yeast, BCAAs in mammals are metabolized to 2-keto acids by the branched-chain amino acid transaminases. The 2-keto acids primarily undergo oxidative decarboxylation by branched-chain keto acid dehydrogenase to yield substrates for further oxidation and generation of anaplerotic compounds for the TCA cycle [64]. However, the molecules characterized here appear to have undergone an alternative route, wherein 2-keto acids are reduced to form 2-hydroxy acids. Excess levels of 2-hydroxy acids are found in human patients with maple syrup urine disease, also known as branched-chain ketoaciduria. This is an autosomal recessive disorder caused by a deficiency in dehydrogenase activity. Without this enzyme, 2-keto acids accumulate and are shunted towards formation of 2-hydroxy acids [65, 66]. Accumulation of 2-keto and 2-hydroxy acids often results in brain damage due to impaired neurotransmitter function caused by inhibition of glutamate uptake [67, 68], and neuronal energy metabolism dysfunction [69, 70]. Although 2-hydroxy acids are produced, the accumulation of BCAAs and 2-keto acids seems to have the greater impact on the pathophysiology of maple syrup urine disease [71].
Previously we showed that osmotic stress dampens and delays the mating pheromone response in yeast [33]. Here we describe potential mechanisms of this cross-pathway regulation. While our analysis focused on yeast, several tissues routinely experience osmotic stress, and can develop disease if osmoregulation is impaired. For example, osmotic stress can promote dry eye disease [72] and diabetic retinopathy [73]. High osmolarity in the vasculature can lead to hypertension [74] and a hyperosmolar hyperglycemic state in diabetics [75]. Importantly, BCAA metabolism is also conserved in humans [76]. Reduced levels of the BCAAs are observed in heart failure, sepsis, trauma, and burn injury [77]. Moreover, a reduction in the expression of branched-chain amino acid transaminase and keto acid dehydrogenase, as well as an increase in the levels of 2-keto acids, have been identified as hallmarks of heart failure [78]. However the connection between osmotic stress signaling and BCAA metabolism is not clearly understood. Collectively, these examples highlight the need for a more complete understanding of the osmotic stress response and of BCAA metabolism.
In summary, we identified 2-hydroxy BCAA derivatives as candidate second messengers of the osmotic stress pathway. As second messengers, these molecules are likely used to amplify the osmotic stress response and coordinate responses to hormones and neurotransmitters. A challenge for the future is to determine the mechanism by which Hog1 (or p38) promotes BCAA derivative accumulation, their cellular target(s) in both yeast and humans, as well as their potential as lead molecules for pharmacological control of the stress response in a mammalian system.
All strains (S3 Table) were generated from the BY4741 wild type strain (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) [79]. Gene deletion strains were generated by homologous recombination of PCR-amplified drug resistance genes from the pFA6a-KanMX6 [80] or pFA6a-hphMX6 plasmids [81], with flanking sequence homologous to the gene of interest [82], or by the delitto perfetto method, leaving no selection marker [83]. Similarly, Flag-tagged strains were generated by homologous recombination of the PCR-amplified cassette from pFA6a-6xGly-3xFlag-HIS3MX6 [84], with flanking sequence homologous to either side of the stop codon of the gene of interest. Bat15A Bat23A non-phosphorylatable mutants were generated using the delitto perfetto method. BAT1 was replaced with the counter selectable marker and reporter gene cassette and then with synthesized bat15A. The same steps were then used to replace BAT2 with bat23A. All cells were grown at 30°C unless otherwise noted.
The pRS426-PFUS1-YeGFP3 plasmid was generated by subcloning the YeGFP3 gene [85] under control of the yeast FUS1 promoter from pDS30 (from Daria Siekhaus, University of California, Berkeley) [86] into pRS426 [87], by digestion with BamHI and XhoI, and ligation of gel-purified products. pYEplac181-pHluorin (2μ, ampR, LEU2+) was the gift of Rajini Rao (Johns Hopkins University) (S4 Table) [34, 45, 46].
Cells were grown to saturation overnight in SCD medium, diluted to OD600 = 0.10, grown to OD600 ~0.6–0.8, then diluted again and grown to OD600 ~1.0. One third volume of SCD containing 3x stress stimulus was added to the experimental cell cultures. Control samples were mixed with 1/3 volume of SCD alone. Aliquots were collected at the times indicated, mixed 19:1 with 6.1 N trichloroacetic acid (TCA) and placed on ice. Cell pellets were collected by centrifugation at 1962 x g for 2 minutes, and resuspended in 10 mM NaN3. Cells were collected by centrifugation at 16,060 x g for 1 minute, the supernatant was removed, and cell pellets were stored at -80°C until use.
Heat shock experiments were carried out by growing cells as indicated above, then transferring the cultures to a 42°C water bath incubator/shaker and adding 1/3 final volume of SCD medium pre-warmed to 42°C. Control samples were mixed with 1/3 volume of media at 30°C.
For glucose limitation experiments, wild-type cells were grown as above to OD600 ~0.8, collected by centrifugation at 1962 x g for 2 minutes, resuspended with one-quarter volume of glucose-free SC medium, centrifuged again and resuspended in the original volume of SCD medium containing either 2% or 0.05% glucose. Note that the centrifugation step leads to partial and transient Gpa1 phosphorylation (Fig 1A, 2% Glucose curve).
Cell pellets were thawed on ice, and resuspended in ice cold TCA buffer (10 mM Tris-HCl, pH 8.0, 10% TCA, 25 mM ammonium acetate, 1 mM ethylenediaminetetraacetic acid). Cells were vortexed for 10 minutes, then collected by centrifugation at 16,060 x g for 10 minutes at 4°C. Pellets were reconstituted in resuspension buffer (100 mM Tris-HCl, pH 11.0, 3% sodium dodecyl sulfate (SDS)), heated at 99°C for 10 minutes, cooled to room temperature for 10 minutes, and centrifuged at 16,060 x g for 1 minute. Lysates were transferred to new tubes and 5 μL was used in a Bio-Rad DC Protein Assay (Bio-Rad #5000112), carried out according to the manufacturer’s protocol, and compared against a bovine serum albumin standard curve. Lysates were normalized to 2 μg/μL with resuspension buffer and 6x SDS sample buffer (350 mM Tris-HCl, pH 6.8, 30%(v/v) glycerol, 10%(w/v) SDS, 600 mM dithiothreitol, 0.012%(w/v) bromophenol blue), and used immediately or stored at -80°C.
Cell lysates were heated at 99°C for 10 minutes, then 40 μg of protein was loaded onto 10% SDS-PAGE gels. Gels were then run in SDS electrophoresis buffer (25 mM Tris base, 20 mM glycine, 0.1%(w/v) SDS) at room temperature for 20 minutes at 20 mA/gel after which, current was increased to 25 mA/gel for 110 minutes. Electrophoresed proteins were then transferred to nitrocellulose membranes at 100 V for 90 minutes at 4°C in transfer buffer (20% methanol, 25 mM Tris base, 200 mM glycine). Membranes were blocked in TBS-T (100 mM Tris Base, pH 7.5, 150 mM NaCl, 0.1% Tween-20) containing 5% (w/v) milk and 10 mM NaN3 for 1 hour unless otherwise indicated. Western blots were probed with antibodies raised against Gpa1 (in-house rabbit polyclonal antibody, 1:1,000 ratio) [88], phospho-Snf1 (phospho-AMPKα (Thr172) 40H9 Rabbit mAb, Cell Signaling Technology #2353, 1:2,000 ratio), Snf1 (poly histidine HIS-1 mouse mAb, Sigma-Aldrich #H1029, 1:3,000 ratio), Hog1 (Santa Cruz Biotechnology #sc-6815, 1:500 ratio), phospho-Hog1 (phospho-p38 MAPK (Thr180/Tyr182) 28B10 Mouse mAb, Cell Signaling Technology #9216, 1:500 ratio), and Glucose-6-phosphate dehydrogenase as a loading control (G6PDH, Sigma # A9521, 1:50,000 ratio). Blots were incubated with primary antibodies for 1 hour to overnight, washed 3 x 5 minutes with TBS-T, then incubated with horseradish peroxidase-conjugated secondary antibodies raised against rabbit (Bio-Rad #1662408), mouse (Bio-Rad #1721011), or goat (Santa Cruz Biotechnology #sc-2768) at a 1:10,000 ratio in TBS-T containing 5% (w/v) milk, and washed 3 x 5 minutes with TBS-T. Blots were imaged on a Bio-Rad ChemiDoc MP imaging system after a 5 minute incubation with Clarity ECL Western Blotting Substrate (Bio-Rad #1705061).
Wild type yeast were transformed with plasmid pYEplac181-pHluorin [34, 45, 46] and grown in SCD-Leu medium. For cells treated with BCAA derivatives (30 mM) the medium was titrated to pH 5.0 with HCl. Experiments and pHi calculations were carried out as in [34] using the indicated stressor or metabolite at 3x stock concentration.
Wild type and hog1Δ cells were grown to saturation overnight, diluted to OD600 = 0.10 grown to OD600 ~0.6, diluted again to OD600 = 0.00075, incubated overnight to OD600 ~0.9. Cultures were then split in half and grown to OD600 ~1.0 and mixed 1:4 with SCD or SCD plus 2.5 M KCl. After 3 minutes the cultures were transferred to 250 mL conical bottles (Corning #430776) and centrifuged for 3 minutes at 1819.3 x g in a Sorvall RC3C Plus centrifuge using an H6000A swinging bucket rotor. After aspirating the supernatant the cell pellets were snap-frozen in place with liquid nitrogen and stored at -80°C. The cells were exposed to KCl for a total of 20 minutes. Frozen pellets were submitted to Metabolon, Inc. for GC-MS and LC-MS/MS analysis of metabolites (see S1 Methods).
For NMR measurements, 15N-enriched Gαi-Δ31 produced as in [89] was exchanged into NMR buffer (20 mM sodium phosphate, pH 7.0, 50 mM NaCl, 2 mM MgCl2, 200 μM GDP, 5% D2O). Each NMR sample contained 50 μM Gαi-Δ31 and 1.25 mM ligand. NMR spectra were acquired at 25°C on a Bruker Avance 850 NMR spectrometer. Two-dimensional 1H–15N HSQC experiments were recorded with 1024 and 128 complex points in the direct and indirect dimensions, respectively, 44 scans per increment and a recovery delay of 1.0 seconds. Spectral widths used were 13586.957 Hz (1H) and 3015.682 (15N) Hz. Spectra were processed and analyzed using NMRPipe (NIDDK, NIH) and Sparky (UCSF).
Four colonies of the same strain transformed with plasmid pRS426-PFUS1-YeGFP3 and one colony of the untransformed background strain (to use for background fluorescence subtraction) were grown to OD600 ~1.0. Samples were added in duplicate to black clear-bottomed 96-well plates containing 10x stocks of serially diluted α-factor mating pheromone ranging in concentration from 1x10-4.5 M to 1x10-10 M prepared in sterile water, and 5x stocks of stimulus solution prepared in growth medium. The OD600 for each well was measured for cell number normalization. After 3 hours, GFP fluorescence was measured at an excitation wavelength of 485 nm, and emission wavelength of 538 nm, using a cutoff of 530 nm, in a Molecular Devices Spectramax M5 plate reader. For data presentation, raw fluorescence values from each well were normalized to the number of cells in that well (represented by the OD600) using the shorthand Taylor Series 11+x where x = OD600. Normalized values of each technical duplicate were averaged, and normalized values from the background strain (containing no fluorescence reporter) were subtracted. Finally, each well was normalized as a percent to the average maximum fluorescence value in the α-factor treated positive control. Dose-response curves were fitted using a nonlinear Boltzmann function.
All data are reported as mean ± the standard deviation. Statistical significance was determined by an unpaired two-sided student’s t-test. In all cases, a p-value ≤ 0.05 was considered to be statistically significant.
|
10.1371/journal.pbio.1000244 | The Multicopy Gene Sly Represses the Sex Chromosomes in the Male Mouse Germline after Meiosis | Studies of mice with Y chromosome long arm deficiencies suggest that the male-specific region (MSYq) encodes information required for sperm differentiation and postmeiotic sex chromatin repression (PSCR). Several genes have been identified on MSYq, but because they are present in more than 40 copies each, their functions cannot be investigated using traditional gene targeting. Here, we generate transgenic mice producing small interfering RNAs that specifically target the transcripts of the MSYq-encoded multicopy gene Sly (Sycp3-like Y-linked). Microarray analyses performed on these Sly-deficient males and on MSYq-deficient males show a remarkable up-regulation of sex chromosome genes in spermatids. SLY protein colocalizes with the X and Y chromatin in spermatids of normal males, and Sly deficiency leads to defective repressive marks on the sex chromatin, such as reduced levels of the heterochromatin protein CBX1 and of histone H3 methylated at lysine 9. Sly-deficient mice, just like MSYq-deficient mice, have severe impairment of sperm differentiation and are near sterile. We propose that their spermiogenesis phenotype is a consequence of the change in spermatid gene expression following Sly deficiency. To our knowledge, this is the first successful targeted disruption of the function of a multicopy gene (or of any Y gene). It shows that SLY has a predominant role in PSCR, either via direct interaction with the spermatid sex chromatin or via interaction with sex chromatin protein partners. Sly deficiency is the major underlying cause of the spectrum of anomalies identified 17 y ago in MSYq-deficient males. Our results also suggest that the expansion of sex-linked spermatid-expressed genes in mouse is a consequence of the enhancement of PSCR that accompanies Sly amplification.
| During meiosis in the male mouse, the X and Y chromosomes are transcriptionally silenced, and retain a significant degree of repression after meiosis. Postmeiotically, X and Y chromosome–encoded genes are consequently expressed at a low level, with the exception of genes present in many copies, which can achieve a higher level of expression. Gene amplification is a notable feature of the X and Y chromosomes, and it has been proposed that this serves to compensate for the postmeiotic repression. The long arm of the mouse Y chromosome (MSYq) has multicopy genes organized in clusters over several megabases. On the basis of analysis of mice carrying MSYq deletions, we proposed that MSYq encodes genetic information that is crucial for postmeiotic repression of the sex chromosomes and for sperm differentiation. The gene(s) responsible for these functions were, however, unknown. In this study, using transgenically delivered small interfering RNA, we disrupted the function of Sly, a gene that is present in more than 100 copies on MSYq. Sly-deficient males have major sperm differentiation problems together with a remarkable postmeiotic derepression of genes encoded on the X and Y chromosomes. Furthermore, the epigenetic modifications normally associated with sex chromosome repression are altered. Our data thus show that the SLY protein is required to mediate postmeiotic repression of the X and Y chromosomes. It is likely that the sperm differentiation problems in Sly-deficient males are largely a consequence of the derepression of the sex chromosomes in spermatids. We propose that the postmeiotic repressive effect of Sly on genes encoded on the X and Y chromosomes drove their massive amplification in the mouse.
| During spermatogenesis, germ cells progress through three phases to become functional sperm: proliferation, meiosis, and spermiogenesis. In the latter phase, haploid germ cells (spermatids) undergo dramatic remodeling and DNA compaction as they differentiate into spermatozoa.
The X and Y chromosomes are transcriptionally silenced during meiosis by a process termed meiotic sex chromosome inactivation (MSCI), and postmeiotically, the spermatid X and Y chromosomes remain largely repressed [1]. Nevertheless, there is substantial X and Y gene expression in spermatids, and based on their analysis of X gene expression in spermatids, Mueller and colleagues have argued that gene amplification plays a key role in compensating for postmeiotic sex chromatin repression (PSCR) [2]. Although the chromatin modifications associated with MSCI and PSCR are not the same [1],[3], PSCR is thought to be a downstream consequence of MSCI [4],[5].
In 2005, we reported the surprising finding that deletions of the long arm of the mouse Y (MSYq) lead to the up-regulation of several spermatid-expressed X and Y chromosomal genes [6]; this suggests that one (or more) of the multicopy genes known to be located on MSYq is involved in PSCR. Aside from this, MSYq deficiencies cause sperm head malformations, with severity correlating with the extent of the deficiency and ultimately leading to infertility [7]–[11]. Intriguingly, males with an approximately two-thirds deletion of MSYq (2/3MSYq−) are fertile but produce offspring with a sex ratio distortion in favor of females; this has been considered a manifestation of a postmeiotic intragenomic conflict between the sex chromosomes that led to the amplification of sex ratio distorter and suppressor genes [12]–[14].
Our favored candidate for the MSYq factor needed for normal sperm differentiation and a balanced sex ratio has been Sly, one of the four multicopy genes identified on the mouse Y long arm [6],[15],[16]. Sly encodes a protein that is very highly expressed in round spermatids, and among the proteins with which it interacts are the acrosomal protein DKKL1 and the chromatin modifier and transcriptional coactivator KAT5 (aka TIP60) [17]. More than 70 copies of Sly that retain an open reading frame, and 30 copies annotated as “noncoding” are predicted to be present on MSYq (Entrez Gene database from the National Center for Biotechnology Information [NCBI]; http://www.ncbi.nlm.nih.gov/sites/entrez?db=gene) as a result of the amplification of a >500-kb repeat unit encompassing at least two copies of Sly [16] (J. Alfoldi and D. C. Page, personal communication). Sly expression is consequently reduced in proportion to the extent of MSYq deficiency [15],[17]. Interestingly, the X chromosome carries multiple copies (∼25) of Slx (Sycp3-like, X-linked) [2], a gene related to Sly that encodes a cytoplasmic spermatid-specific protein of unknown function [18]. Slx is one of the X-linked genes found to be up-regulated in MSYq-deficient males, and Slx and Sly have been suggested to be key players in the postmeiotic X-Y genomic conflict [6].
Because Sly is present in multiple copies, traditional gene targeting is not an option for investigating its function. In the present study, by using an in vivo RNA interference approach, we have produced male mice with a dramatic reduction in Sly expression. The analysis of these mice has enabled us to demonstrate that SLY is a key regulator of sex chromosome gene expression during sperm differentiation.
The major challenge in the study of MSYq gene functions is the highly repetitive nature of MSYq, which contains coamplified multicopy genes organized in clusters over several megabases [16] (J. Alfoldi and D. C. Page, personal communication). This precludes the use of conventional gene targeting, so we used a transgenic approach to deliver short hairpin RNAs (shRNA) [19] designed to generate Sly-specific small interfering RNAs (siRNAs) (Figure 1).
Sly short hairpin target sequences were selected to be shared by most Sly copies, and to be specific to Sly; particular care was taken to avoid sequences that might target related genes such as Slx. The selected Sly short hairpin sequences (shSLY) were cloned under the control of the U6 promoter [20]. We chose U6, a ubiquitous polymerase III promoter, in order to achieve sufficient expression of shSLY RNA to produce a substantial knock-down of the very abundant Sly transcripts. As Sly expression is restricted to the testes, no other organs were expected to be affected by its knock-down, and indeed, we did not see any phenotypic changes other than testis-related changes (see below).
The efficiency and specificity of shSLY constructs were tested in cell culture by cotransfection experiments (Figure S1). Two shSLY constructs (sh136 and sh367—see Figure 1) were used to produce transgenic mice. High expression of shSLY RNA was associated with a dramatic decrease in Sly expression at the transcript and at the protein levels (Figure 2A–2C). Transgenic mice for sh136 and sh367 constructs (hereafter, sh136 and sh367 mice) showed an ∼70% reduction of Sly transcripts (Figure 2B), with both known splice variants being affected (unpublished data). SLY protein level was even more dramatically reduced, with no protein being detected by Western blotting (Figure 2C and Figure S2) even after long exposure. This discrepancy suggests that the persisting Sly transcripts are not translated, or that they encode a variant SLY protein(s) not detected by our antibody. The sh367 transgene was also introduced into 2/3MSYq− males and Sly transcript levels fell further, to 10% of those of normal males (Figure 2B).
Four checks were made for “off-target” effects of the RNA interference. First, the levels of two testis-expressed microRNAs, mir-t3 and mir-t25 [21], were measured and found to be unchanged in testes of shSLY mice (Figure S3A), indicating that the transgenically delivered siRNAs did not affect the expression of naturally expressed small RNAs. Second, since some shRNAs or siRNAs induce an interferon response [22]–[24], the expression level of 2′,5′-oligoadenylate synthetase 1 (Oas1) was measured as a marker of an interferon response [22],[24],[25]; this was also unchanged in shSLY mice (Figure S3B). Third, microarray analyses performed on sh367 mice did not detect any significant changes in the expression of known target genes of the interferon pathway (see below). Finally, microarray analyses were performed on juvenile testes (17 d postpartum) to check for potential off-target gene activation before the onset of Sly expression. There were no statistically significant differences in gene expression between juvenile sh367 mice and controls (unpublished data).
In view of the substantial and specific knock-down of Sly expression in the shSLY mice, we proceeded to analyze their phenotypes in order to assess the extent to which Sly depletion mimicked the phenotypic consequences of MSYq deficiencies.
Mice with MSYq deficiencies have an increased incidence of sperm head abnormalities, correlated with the extent of the deletion [7]–[11]. In mice lacking 9/10ths of MSYq (9/10MSYq−) or with no MSYq (MSYq−), 100% of the sperm are abnormal, and this is thought to be the cause of their sterility [10],[11]. Analysis of epididymal sperm from Sly-deficient mice (i.e., shSLY mice from either line) revealed that over 92% of the sperm had head abnormalities, and defects were similar to those observed in mice with MSYq deficiencies (Figure 3A and 3B). The proportion of abnormal sperm heads (grouped into three categories: slightly flattened, grossly flattened, and other gross abnormalities) was very significantly increased (p<0.0001) in shSLY mice compared to controls (transgene-negative siblings). In terms of severity, the sperm head abnormalities of shSLY mice from both sh136 and sh367 lines were very similar and fell between those of 2/3MSYq− and 9/10MSYq− mice (Figure 3A). The presence of the sh367 transgene in the context of 2/3MSYq− led to a more severely abnormal sperm phenotype than that seen in 2/3MSYq− males or in shSLY males with a normal YRIII chromosome (Figure 3A and 3B, Figure S4).
Another spermiogenic defect described for 9/10MSYq− males [11] and seen in MSYq− males (unpublished data) is a delay in sperm shedding. This is also the case for shSLY mice (unpublished data).
Together these results show that the key spermiogenic defects observed in MSYq− mice are also seen in shSLY mice, thus demonstrating that Sly deficiency is the underlying cause.
Previous studies have shown that 9/10MSYq− and MSYq− males are sterile [10],[11], whereas males carrying less extensive deletions of MSYq (such as 2/3MSYq− and B10.BR-Ydel) are fertile, but their sperm have markedly reduced in vitro fertilizing ability [12],[26]–[28]. We therefore checked for impaired fertility in shSLY males, initially focusing our study on sh367 males since the phenotypes of both lines were similar. As is the case for 9/10MSYq− and 2/3 MSYq− males, testis weights for the shSLY males did not differ from controls; sperm numbers were reduced but within the fertile range (Table S1). However, when mated for a period of 7 mo, sh367 mice had markedly fewer offspring and litters when compared to transgene-negative siblings (Table 1). The in vitro fertilizing ability of epididymal sperm samples was also dramatically reduced relative to controls with only one of 662 eggs developing to the two-cell stage (Table 1), and the quality of sperm motility appeared to be impaired as shown by the increase in the proportion of non-progressively motile sperm (Table S1). A reduced quality of motility has recently been reported for B10.BR-Ydel males [29]. Overall, the fertility defects were intermediate in severity between those of 2/3MSYq− and 9/10MSYq− mice, as was the case for the sperm head abnormalities. Surprisingly, two sh136 and sh367 transgenic males that were obtained early in the backcrosses were exceptionally fertile, whereas Sly-deficient males obtained in subsequent generations are all markedly subfertile or sterile (see Table S2).
Because of the poor fertility of Sly-deficient males, offspring sex ratio data are very slow to accumulate. After pooling data for all matings involving sh136 and sh367 males and for nontransgenic males arising in the same breeding program, 55.5% (79 of 142) of the offspring of the transgenics are female and 47.8% (176 of 368) of the offspring from nontransgenics are female. The sex ratio distortion in favor of females (7.7%) approaches significance at the 0.05 level (p-value = 0.0569). However, a major caveat is the fact that a large proportion of the data come from just two males (Table S2). It may be necessary to produce an shSLY line with an intermediate knock-down to establish whether or not Sly deficiency contributes to the sex ratio distortion observed in 2/3MSYq− mice.
In our initial gene expression study of MSYq-deficient mice, 18 sex chromosome genes (of which 16 were X- and two Y-linked) were found up-regulated [6]. We decided to investigate the consequences of Sly deficiency on gene expression in our new mouse model, and to reexamine gene expression in MSYq-deficient males, using a more exhaustive array. Microarray analyses were performed on adult testes of sh367 mice, sibling controls, 2/3MSYq−, 9/10MSYq−, and MSYq− mice (see Figure S5). We found 230 differentially expressed genes that were grouped into five categories based on their expression ratios across all genotypes (Figure 4A). The largest category (127 genes) comprises genes that are up-regulated in sh367 mice and in mice with MSYq deficiencies (category 2); 67.7% (86/127) of these up-regulated genes are X-linked, increasing to 81% (68/84) of those up-regulated at least 1.5-fold (Figure S5). Many of these X-linked genes are present in multiple copies, such as Slx, Cypt, Asb, Ssxb, and Rhox; of these, Slx, Cypt, Ssxb, Rhox3, and Rhox11 are specifically expressed in postmeiotic cells (Table 2). Several of the up-regulated single-copy X-linked genes are also known to be involved in the differentiation of postmeiotic germ cells (i.e., spermatids) (Table 2). Indeed, Actrt1 and Spaca5 are components, respectively, of the perinuclear theca and the acrosome, two highly differentiated structures of the sperm head [30],[31]. X-linked genes encoding variants of histone H2A (LOC100045423, a copy of H2al1, and LOC100046339, which is closely related to H2A.Bbd) were also derepressed in shSLY and MSYq-deficient mice (cf. Table 2).
The array also shows up-regulation in sh367 mice of 27 Y-linked gene loci, almost exclusively representing the multicopy genes Ssty1 and Ssty2 (category 4). Ssty1 and Ssty2 are spermatid-specific MSYq-encoded genes [32] and are consequently reduced in mice with MSYq deletions (Figure S5). Overall, more than 65% of the genes up-regulated in shSLY mice, and >80% of the ones that are highly up-regulated (>1.5-fold increase), are located on the sex chromosomes (Figure 4A). No X or Y genes (except one pseudogene provisionally mapped to the Y) were found to be down-regulated.
The microarray results for a number of genes were confirmed by real-time PCR (Figure 4B). Spermatid-specific X- and Y-linked multicopy genes, such as Slx, Slx-like, H2al1, Ssty1, and Ssty2 are all markedly derepressed in shSLY mice, and this is also the case for the spermatid-specific single-copy genes Actrt1 and 1700008I05Rik (an X-linked homolog of the t-complex gene Tcp11). Of two other MSYq-encoded spermatid-expressed genes that were not on the array, Asty was not significantly up-regulated, and Orly was not as dramatically up-regulated as Ssty1 and Ssty2. Zfy2, another gene predominantly expressed in spermatids [33], was found markedly up-regulated by real-time PCR, even though not picked up in our array analysis. Zfy2 is encoded by the Y chromosome short arm, and its up-regulation shows that the derepression of the Y is not restricted to its long arm. Levels of expression of autosomal genes, including the spermatid-specific Acrv1 and Protamine1 genes, are not significantly changed, in agreement with the sex-linked gene bias identified in the microarray (Figure 4B). Similar results were obtained for sh136 transgenic mice (unpublished data).
A further microarray analysis was performed on purified round spermatids from sh367 transgenic mice, sibling controls, and 2/3MSYq− mice. The vast majority of the X and Y genes found up-regulated before were also found significantly up-regulated in the new comparison (109 of 113, Figure S6). In addition, this new dataset identified a greater number of up-regulated X-linked genes (126 vs. 68 genes, >1.5-fold increase; see Figure S6). The identification of a greater number of genes in the new dataset is probably due to the increase in sensitivity when the analysis is restricted to the cell type in which up-regulation occurs. Additional Y-encoded up-regulated transcripts were also identified, such as Orly, Rbm31y, and H2al2y (a Y-encoded histone H2A spermatid specific variant).
This demonstrates that the derepression of sex chromosome genes occurs in spermatids and also provides a control for differences in the cellular composition between Sly-deficient, MSYq-deficient, and control testes.
The up-regulation of sex-chromosome spermatid genes observed at the transcript level is also detected at the protein level, as shown for SLX and SSTY1 proteins (Figure 5A and 5B and Figure S2). SLX immunostaining of testis sections confirmed that the derepression is restricted to spermatids (Figure S7).
All these data point to a global derepression of postmeiotic sex chromatin (PMSC) following Sly deficiency; the X and Y genes that are up-regulated are those already expressed in spermatids. Thus, it is clear that SLY has a key role in PMSC repression.
Several studies have demonstrated that the PMSC of X and Y spermatids is enriched in histone modifications known to be associated with transcriptional repression, such as hypermethylation of lysine 9 of histone H3 (H3K9) [3]–[5],[34],[35]. The heterochromatin proteins CBX1 and CBX3 (aka HP1β and HP1γ) also accumulate on PMSC [4],[5],[35]. As shown in other contexts, the heterochromatin proteins are recruited via binding to methylated H3K9 [36],[37] and mediate gene repression [38],[39].
In view of our microarray results implying global PMSC derepression in Sly-deficient spermatids, we decided to examine these repressive chromatin marks in our mouse model. The analysis of shSLY mice revealed a significant (p<0.005) decrease of trimethylated H3K9 (H3K9me3) staining on PMSC as compared with the chromocenter (85% of spermatids with less staining vs. 46.5% in control mice). Similarly, CBX1 accumulation on PMSC was significantly (p<0.05) reduced in mutant mice relative to the chromocenter (82% of spermatids with less staining vs. 57% in control mice) (Figure 6 and Figure S8). A recent study shows a comparable reduction of PMSC-associated H3K9me3 and CBX1 staining in MSYq-deficient males [40].
These observations imply that the repressive effect of SLY on sex chromosome gene expression in spermatids is due to a global effect on PMSC via ubiquitous mediators of heterochromatinization/transcriptional silencing.
Despite the fact that SLY has been shown to interact with the histone acetyl transferase KAT5 [17], so far no obvious change of histone acetylation was detected in Sly-deficient round spermatids (unpublished data). KAT5 is highly expressed in spermatocytes but poorly expressed in spermatids ([41] and unpublished data), and it is possible that the effect of SLY on KAT5 function is too subtle to be observed with our current tools.
The SLY protein is related to SYCP3 and XLR, two nuclear proteins thought to associate with chromatin via their conserved COR1 domain (NCBI Conserved Domains Database; http://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml) [15]. Based on cytoplasmic/nuclear protein extracts, SLY protein is predominantly located in the cytoplasm of round and early-elongating spermatids, but a significant fraction is nevertheless observed in the nucleus [17]. However, the nuclear localization has not been documented by immunostaining. Using modified immunohistochemistry protocols, we have now been able to observe SLY protein in the nuclei of spermatids from stage II–III until early stage IX, with the intensity of the signal increasing through spermatid development. SLY nuclear staining is then excluded from the nuclei at the onset of spermatid elongation (from stage IX, see Figure 7 and Figure S9). SLY nuclear localization is consequently specific to round spermatids (probably excluding stage I round spermatids in which SLY nuclear staining was not detectable above background). At higher magnification, SLY seemed to localize to a DAPI-dense subnuclear structure that could be the PMSC (Figure S9C). This was confirmed in spread spermatids in which SLY clearly colocalized with either the X or the Y PMSC in 66% of round spermatids (Figure 8 and Figure S10). In addition to PMSC colocalization, SLY protein was sometimes observed outside PMSC (ectopic, Figure 8); the reason for this non-PMSC localization remains to be determined. The 31.5% of spermatid nuclei without an SLY signal may be accounted for by the absence of nuclear SLY in early-stage spermatids.
These data strongly suggest that SLY induces gene repression via direct interaction with the PMSC or with PMSC protein partners such as histone-modifying enzymes.
In a previous study, we observed that SLY interacts with the acrosomal protein DKKL1 [17]; the severe defects of Sly-deficient sperm could be a consequence of the disturbance of this interaction. However, DKKL1 intracellular localization was unchanged in Sly-deficient testes (Figure S11), and the global level of DKKL1 protein in purified spermatids was also unaffected (Figure S2).
In addition to the massive up-regulation of sex chromosome genes expressed in spermatids, some autosomal genes were identified as up-regulated in Sly- and MSYq-deficient mice (see Figure 4A; and category 2, Figure S5). Intriguingly, these included five members of the multicopy Speer gene family and several autosomal genes encoding variants of histones H3 and H4 (Table 2). Genes that were down-regulated in shSLY and in MSYq-deficient mice (category 3) were exclusively autosomal (aside from a provisionally Y-linked pseudogene). These down-regulated genes include Chaf1b, a chromatin assembly factor (Figure S5). The microarray analysis on purified round spermatids confirmed the change of expression for the majority of these autosomal genes (Figure S6).
Other autosomal genes were found to be up-regulated or down-regulated in shSLY mice, but not in MSYq− mice (category 1a and 1b, Figure S5). This could be due to off-target effects of shSLY transgene expression, but no particular pattern/pathway was apparent. Genes of the interferon pathway, previously reported to be activated by some shRNAs [22]–[24], were not up-regulated in shSLY mice, supporting the evidence from the microarray analyses performed before the onset of Sly expression that there are no off-target effects. Alternatively, the changes in autosomal gene expression that are specific to shSLY mice could be a consequence of the up-regulation of other MSYq genes such as Ssty1 and Ssty2 transcripts; in MSYq-deficient models, all MSYq genes show reduced expression.
Using transgenically delivered shRNAs, we have successfully disrupted the function of Sly, an MSYq-encoded multicopy gene. This demonstrated first, that SLY has a critical role in PSCR, and second, that Sly deficiency is the major underlying cause of the spectrum of anomalies identified 17 y ago in MSYq-deficient males.
Our previous study suggesting that MSYq encodes information required for the repression of PMSC identified 18 up-regulated sex-linked genes in testes of MSYq-deficient mice: 16 from the X chromosome and two from the Y chromosome short arm [6]; many of these genes were exclusively expressed in spermatids. Here, our more extensive analysis of MSYq-deficient mice together with Sly-deficient mice identified 113 up-regulated sex-linked genes: 86 X-linked and 27 Y-linked. The latter mice served to establish that the up-regulation is due to Sly deficiency and that this up-regulation includes multicopy MSYq genes such as Ssty1 and Ssty2. A further microarray confirmed the up-regulation in purified spermatids. SLY is consequently the MSYq factor required for PSCR in the mouse, and the generality of the sex-linked gene repression demonstrates that SLY acts to globally repress the PMSC.
How does SLY mediate this global repression of the PMSC in X and Y spermatids? Importantly, we have shown that SLY is nuclear from stage II to IX and localizes to the PMSC of X and Y spermatids. This localization may involve the SLY COR1 domain, which is proposed to mediate association with chromatin (NCBI Conserved Domains Database) [15]. The localization to X-bearing spermatids will have been facilitated by the sharing of gene products via intercellular bridges [42].
In the male germline, the X and Y chromosomes are initially transcriptionally inactivated at the beginning of pachytene (meiotic sex chromosome inactivation [MSCI]); this inactivation is triggered by the phosphorylation of the histone H2AX [1]. During the transition from MSCI to PSCR, there are changes in nucleosomal histones, in epigenetic histone marks, and in the recruitment of heterochromatin proteins; these chromatin features are lost from the PMSC during stage XI [3]–[5],[34],[35],[43]. This loss at stage XI is unsurprising since it is the stage when the replacement of histones with protamines is initiated.
Of importance in the current context is the recruitment of the heterochromatin protein CBX1 during diplotene, coincident with the loss of H2AX phosphorylation, that is presumed to be responsible for sex chromosome repression following the shutdown of MSCI [4],[5],[43],[44]. It is thus significant that we have found that SLY subsequently plays a role in maintaining CBX1 enrichment in the PMSC, and thus, in maintaining a substantial degree of transcriptional repression. Hypermethylated H3K9 is known to be a platform for CBX1 recruitment [36],[37], and in view of the changes in nucleosomal histones during the MSCI-PSCR transition [3], maintenance of CBX1 enrichment is likely to require continuing H3K9 methylation as the new histones are introduced. It is therefore noteworthy that SLY is also involved in maintaining H3K9 trimethylation.
In our microarray screening, several autosomal and sex-linked genes coding for histones H2, H3, and H4 variants (including recently identified spermatid-specific H2A variants, H2al1 and H2al2y [45],[46]) were found up-regulated when Sly expression is reduced. Conversely, Chaf1b, which encodes a chromatin assembly factor, appears down-regulated. All these changes could contribute to the derepression of sex chromatin. The importance of histone H3 (and particularly of variant H3.3) in germline function has been recently described in the mouse and Drosophila [3],[47].
What is the molecular basis for the link between Sly deficiency and the spermiogenic defects in MSYq-deficient mice? SLY interacts with the acrosomal protein DKKL1 [17], but our study shows that DKKL1 level and pattern of expression are not noticeably affected by Sly deficiency. However, the reduction/absence of SLY leads to a dramatic up-regulation of many X and Y genes in spermatids. This up-regulation is almost certainly not benign, and we propose this to be the major contributing factor to the spermiogenic defects of Sly- and MSYq-deficient mice.
Actrt1 [30], Spaca5 [31], and Cypt [33], which are up-regulated in Sly-deficient spermatids, encode proteins of the perinuclear theca and the acrosome, two specific structures of the sperm head, and thus are candidates for contributing to sperm head defects. Candidates for effects on sperm function include 1700008I05Rik, an X-linked homolog of Tcp11, which codes for a receptor of a fertilization promoting peptide thought to promote sperm capacitation/function [48]; Rhox3a, Rhox3h, and Rhox11, related to Rhox5, which has been implicated in sperm production and motility [49]; and the A-kinase anchoring protein Akap14, which is predicted to regulate flagellum function, and consequently, sperm motility [50]. Future work on these many candidate genes will be required to determine their involvement in MSYq-deficient spermiogenesis phenotypes.
If the sperm abnormalities in MSYq-deficient mice are solely a consequence of Sly deficiency, then there should be a correlation between the extent of Sly reduction and the severity of the sperm defects. The relevant genotypes in order of decreasing transcript levels (given as percentage of control) are: 2/3MSYq− (40%), sh367 (30%), sh367 2/3MSYq− (10%), and 9/10MSYq− (<1%) [6]; this is the same order as that for increasing severity of sperm head defects. However, the latter three genotypes were indistinguishable with respect to the expression of SLY protein, since none could be detected by Western analysis. From sequence data for the remaining transcripts, it appears that the majority encode variant (but presumably functional) SLY proteins that are unlikely to be detected by our antibody, thus providing an explanation for the seeming discrepancy between RNA and protein levels. Nevertheless, it is important to bear in mind that 9/10MSYq− mice differ from shSLY mice in that the former are deficient in other MSYq-encoded transcripts (i.e., Ssty1, Ssty2, Asty, and Orly), which could contribute to the severity of their sperm defects.
There is now substantial data documenting that many sex-linked spermatid-expressed genes in the mouse are highly amplified [2],[33],[46], with MSYq genes being especially highly amplified [15],[16] (J. Alfoldi and D. C. Page, personal communication). There are currently two hypotheses that seek to explain this amplification: first, that it is a response to PSCR enabling sufficient expression of some X and Y genes with critical postmeiotic functions [2], and second, that it is driven by a genomic conflict involving postmeiotic competition between X- and Y-encoded gene products that affect sex ratio [13],[14].
Given the variety of genes involved, the PSCR-amplification hypothesis is attractive since it explains why so many different genes have become simultaneously amplified in the mouse: it seems unlikely that they could all affect sex ratio. However, a challenge for the hypothesis is to explain why the same degree of amplification is not seen on the Y chromosome of other species (J. Alfoldi and D. C. Page, personal communication). This can be resolved by our finding that one of the amplified mouse-specific genes, Sly, regulates PSCR (our present data). We therefore propose that the mouse-specific expansion of sex-linked spermatid-expressed gene copy number is a downstream consequence of the enhancement of PSCR that accompanied Sly amplification.
So what drove Sly amplification? The straightforward answer would be that this was necessary in order to maintain Sly function in the face of the enhancement of PSCR, but this creates the paradox that Sly has become amplified in order to escape its own repressive effects. This in turn implies that the enhancement of PSCR must also have been of selective advantage; otherwise, this function of Sly would have been lost. One possibility is that the enhancement of PSCR was a weapon in a postmeiotic genomic conflict, where one or more of the genes on the X chromosome acts to distort the sex ratio in favor of females, whereas Sly acts via PSCR to repress the distorter gene(s) and restore a normal sex ratio. The fact that 2/3MSYq− mice have a sex ratio distortion in favor of females is strong evidence that MSYq does encode a factor or factors that are suppressing sex ratio distortion. For shSLY mice, we observed a mild sex ratio skew of borderline significance. It may be possible to create further shSLY lines with a milder phenotype more comparable to that of 2/3MSYq− mice to enable us to obtain more extensive breeding data than that obtained with the severely subfertile mice in the present study. Our proposals concerning the role of Sly in driving sex-linked spermatid-expressed gene amplification are summarized in Figure 9.
In conclusion, SLY has a predominant role in postmeiotic sex chromatin repression, as it is required for the maintenance of the heterochromatin protein CBX1 on PMSC. Sly deficiency recapitulates almost all of the phenotypes observed in mice with MSYq deletions. Thus, Sly encodes the spermiogenesis factor identified 17 y ago on the Y long arm [10]. Future studies of the many genes that we found differentially expressed in shSLY mice will help in understanding the direct cause(s) of the multiple spermiogenesis defects observed in Sly- and MSYq-deficient mice.
Here, we have used transgenic delivery of siRNAs to disrupt the function of a multicopy Y gene, and the same approach could be used for multicopy genes on other chromosomes, for example, Slx, a gene related to Sly. Furthermore, despite numerous attempts in several laboratories, no one has reported the successful disruption of the function of a single-copy Y gene using traditional gene targeting strategies [51]; transgenic delivery of Y gene–specific siRNAs should be an effective alternative.
To generate the U6shSLY constructs, we used a PCR-based approach similar to that described in Harper et al., 2005, using primers designed to generate the shSLY sequences [20] (cf. Table S3). The PCR products were cloned into the pCR2.1 vector and sequenced (TOPO TA Cloning, Invitrogen). The U6shSLY cassettes were then subcloned into pCX-eGFP plasmid [52].
Prior to injection, the plasmids were linearized at ApaL1 and BamH1 sites and on-column purified from agarose gels (Gel Extract II kit, Macherey Nagel). Fertilized eggs from CBA/Ca×C57BL/10 mating were microinjected with the construct, using standard protocols. Transgenic founders carrying the pCX-eGFP-U6-shSLY construct (shSLY mice) were identified by the ubiquitous expression of eGFP. Two female founders with “strong” eGFP expression were obtained (one transgenic for the sh136 construct, the other for the sh367 construct) and crossed with XYRIII males on a random-bred MF1 albino (National Institute for Medical Research colony) background. These females transmitted the transgene and gave rise to two lines of transgenic mice. The lines are maintained by further backcrossing shSLY transgenic females to MF1 mice and generate XYRIII males with (tsgic) and without (neg sib) the transgene. To produce 2/3MSYq− sh367 transgenic mice, sh367 transgenic females were crossed with XYRIIIqdel males on an MF1 background [12]. The breeding strategy to obtain 9/10MSYq−, MSYq−, and control mice was described previously [15]. Animal procedures were in accordance with the United Kingdom Animal Scientific Procedures Act 1986 and were subject to local ethical review.
To obtain cell fractions enriched in spermatids, an adapted protocol from the trypsin method described by Meistrich [53] was used. Testes from a group of four to five adult mice from the same genotype (i.e., three groups of sh367 transgenic mice, three groups of sh367-negative siblings, and two groups of 2/3MSYq−) were used for the study. Testes were dissected and chopped in 20 ml of DMEM (GIBCO) and treated with 2.5 mg/ml trypsin (GIBCO) and 50 µg/ml DNase I (Sigma) for 30 min at 31°C with stirring. After adding fetal calf serum (GIBCO) (final concentration of 8%), the cells were passed through a 100-µm filter. Cells were then centrifuged at room temperature (500 g, 15 min), resuspended in DMEM 0.5% bovine serum albumin (Sigma) with 50 µg/ml DNase I, and cooled on ice. Cells were counted and checked for clumps before proceeding with the elutriation. Cell integrity was checked using Trypan blue. Fractions enriched in different testis cell types were separated with a JE-6B elutriator (Beckman) with conditions described before [54]. Collected fractions were washed in PBS, and cell pellets were frozen down at −80°C. Fraction content was assessed based on cell morphology after DAPI staining. Fractions #6 contained >90% round spermatids and were used for RNA and protein analyses.
Transgenically delivered shRNAs were detected using the Northern blot protocol optimized for short transcripts described by Shukla et al. [55]. Sh136 reverse primer and sh367 reverse primers were used as probes to detect sh136 and sh367 RNAs. All sequences are available in Table S3.
Western blot analyses were performed as described previously [18]. Briefly, 10 to 15 µg of testis or spermatid fraction protein extracts were run on a 12% SDS/polyacrylamide gel. Following transfer and blocking, membranes were incubated overnight with a primary antibody (anti-SLX antibody [18] and anti-SLY antibody [17] were used at 1/3,000; anti-SSTY1 antibody, i.e., anti-YMT2b [32], and anti-DKKL1 [R&D Systems] were used at 1/1,000 and anti-β-actin [Sigma] at 1/50,000). Incubation with the corresponding secondary antibody, coupled to peroxidase and detection by chemiluminescence, were carried out as described by the manufacturer (SuperSignal West Pico, Pierce).
Immunofluorescence experiments were performed on testis material fixed in 4% buffered paraformaldehyde as described previously [18]. Anti-SLX [18], anti-DKKL1 (R&D Systems), and anti-SLY [17] antibodies were used at 1/100, and a preimmune rabbit serum was used as a control. For nuclear detection of SLY, an additional step of 15-min permeabilization with 0.5% Triton X-100 (Sigma) was performed prior to antigen retrieval, and blocking was performed using 5% fetal calf serum (GIBCO). Alexa Fluor 594–conjugated peanut agglutinin lectin (Invitrogen) was used to stage the testis tubules [56].
A portion of testis (approximately 25 mg) was chopped in 1 ml of RPMI medium (GIBCO) and transferred to a round-bottomed tube. Five milliliters of fixative solution (2.6 mM sucrose, 1.86% formaldehyde) were added to the cells drop by drop, and cells were mixed by inverting the test tube three times. The cell suspension was incubated at room temperature for 5 min before proceeding to centrifugation (1,200 rpm, 8 min). The fixative was then removed and the cells resuspended in six drops of PBS (GIBCO). Two drops of the cell suspension were spread on Superfrost Plus slides (BHD) and air dried for 2 min. The cells were permeabilized by adding 0.5% Triton X-100 to the slides for 10 min, and washed twice in PBS before incubation in blocking buffer (PBS, 0.15% BSA, 0.1% Tween-20) for 30 min at room temperature in a humid chamber. Incubation with the primary antibody (anti-SLY [17], anti-CBX1, or anti-H3K9me3 [Upstate] diluted 1/100) was carried out for 2 h in a humid chamber at 37°C. Three washes of 2 min in PBS were performed before proceeding with secondary antibody detection as described previously [17]. As a control for specificity, SLY antibody was preabsorbed with 8 µg of SLY immunogenic peptide or with 8 µg of a noncompeting peptide (SLX peptide [18]). Controls are described in Figure S10.
For quantification of CBX1 or H3K9me3 signals, the signal intensity over the PMSC and over the chromocenter was measured using the DeltaVision SoftWoRx software, and PMSC/chromocenter ratio was calculated for each cell. (See Figure S8.)
Chromosome painting was performed as described previously [57].
Testes were fixed in Bouin (Sigma) and wax-embedded. Five-micron sections were stained with periodic acid–Schiff (PAS). For the analysis of the sperm shedding delay, ten tubules of stage IX to XI were analyzed per mouse, for five mice per genotype. Silver staining of sperm smears obtained from the initial caput epididymis was performed as described previously [11].
In vitro fertilization (IVF) was performed with sperm from three sh367 transgenic males and three negative controls, using oocytes from B6D2F1 (C57BL/6×DBA/2) hybrid and MF1 outbred females. Each male was tested in duplicate following initial semicastration. In each IVF session, sperm from each male were incubated in parallel with oocytes from the two types of females. The method for IVF has been reported before [28]. Briefly, sperm were released from a single epididymis directly into T6 medium and capacitated for 1.5 h, prior to addition of the oocytes-cumulus complexes obtained from hormonally stimulated females. The gametes were co-incubated for 4 h with sperm density ∼2–3×106/ml. After fertilization, the oocytes were washed and cultured; the number of two-cell embryos was recorded after 24 h.
To analyze sperm number and motility, a small portion of sperm suspension was placed in a hemacytometer chamber. Three independent scorings were done per sample, and the final result was a mean of these scorings.
The fertility of shSLY mice was assessed over a period of 7 mo by mating two sh367 transgenic males and two negative siblings with MF1 females. Mating was confirmed by the presence of copulatory plugs.
For real-time reverse transcription-polymerase chain reaction (RT-PCR), total testis RNA was extracted using Trizol and then DNaseI-treated (Invitrogen). Reverse transcription of polyadenylated RNA was performed with Superscript Reverse Transcriptase II, according to the manufacturer's guidelines (Invitrogen). Real-time PCR was performed using Absolute qPCR SYBR Green ROX mix (ThermoFisher) on an ABI PRISM 7500 machine (Applied Biosystems). PCR reactions were incubated at 95°C for 15 min followed by 40 PCR cycles (5 s at 95°C, 20 s at 60°C, and 45 s at 68°C). Primer sequences are available in Table S3. Samples from four transgenic mice and three nontransgenic siblings (negative controls), all at 2 mo of age, were analyzed. All reactions were carried out in triplicate per assay, and β-actin was included on every plate as a loading control. The difference in PCR cycles with respect to β-actin (ΔCt) for a given experimental sample was subtracted from the mean ΔCt of the reference samples (negative siblings) (ΔΔCt). For the quantification of Sly knock-down, values were further normalized to ΔΔCt values of the spermatid-specific control Acrv1. This was to have a more robust analysis when compared with 2/3MSYq− mice, which have variability in spermatid content.
For microarray analyses, absolute expression values were obtained by single-color hybridizations (Illumina BeadChip, mouse whole-genome array, v2) for three sh367 transgenic individuals and matched littermate controls (negative siblings), and RNA from each individual was hybridized separately. A similar analysis was performed on 2/3MSYq−, 9/10MSYq−, and MSYq− samples and appropriate age/strain-matched controls. In each case, pooled RNA from two or three individuals was used as the sample. Differentially expressed genes were grouped into five categories based on their expression ratios across all genotypes (see Figure S5). Similar microarray analyses were performed on juvenile testes (17 d postpartum) from three sh367 males and three littermate controls (negative siblings). There was no significant change of gene expression between the two groups. Microarray analyses were also performed on purified spermatid fractions from two groups of sh367 transgenic mice, two groups of sh367 negative siblings, and two groups of 2/3MSYq−.
For comparisons of the incidence of sperm head abnormalities and of sperm motility (after conversion of percentages to angles), and of the CBX1 and H3K9me3 PMSC/chromocenter intensity ratios, differences between genotypes were assessed by ANOVA using the Generalized Linear Model provided by NCSS statistical data analysis software. Chi-square analysis was used to compare sex ratio and IVF data; for the sex ratio, we used a one-tailed test of significance since we sought to test whether there was a sex ratio distortion in favor of females (as seen in 2/3MSYq− mice). Student t-test was used to compare the data obtained for fecundity, sperm number, testis weight, Northern and Western blot quantification, and real-time PCR (performed on the ΔΔCt values). For microarray analysis, quantile normalization of all expression data was performed using BeadStudio (Illumina). Data for the normal/mutant sh367 animals were compared in BeadStudio, using the Illumina custom error model with a false discovery rate of 5%. For the cluster analysis, normalized data for all samples were imported into Inforsense Discovery Studio (Inforsense), log2-transformed, and expression ratios calculated relative to the appropriate controls. Hierarchical clustering was then performed on the ratio values, using Pearson correlation as the distance metric.
|
10.1371/journal.pgen.1000204 | Intronic Alus Influence Alternative Splicing | Examination of the human transcriptome reveals higher levels of RNA editing than in any other organism tested to date. This is indicative of extensive double-stranded RNA (dsRNA) formation within the human transcriptome. Most of the editing sites are located in the primate-specific retrotransposed element called Alu. A large fraction of Alus are found in intronic sequences, implying extensive Alu-Alu dsRNA formation in mRNA precursors. Yet, the effect of these intronic Alus on splicing of the flanking exons is largely unknown. Here, we show that more Alus flank alternatively spliced exons than constitutively spliced ones; this is especially notable for those exons that have changed their mode of splicing from constitutive to alternative during human evolution. This implies that Alu insertions may change the mode of splicing of the flanking exons. Indeed, we demonstrate experimentally that two Alu elements that were inserted into an intron in opposite orientation undergo base-pairing, as evident by RNA editing, and affect the splicing patterns of a downstream exon, shifting it from constitutive to alternative. Our results indicate the importance of intronic Alus in influencing the splicing of flanking exons, further emphasizing the role of Alus in shaping of the human transcriptome.
| The human genome is crowded with over one million copies of primate-specific retrotransposed elements, termed Alu. A large fraction of Alu elements are located within intronic sequences. The human transcriptome undergoes extensive RNA editing (A-to-I), to higher levels than any other tested organism. RNA editing requires the formation of a double-stranded RNA structure in order to occur. Over 90% of the editing sites in the human transcriptome are found within Alu sequences. Thus, the high level of RNA editing is indicative of extensive secondary structure formation in mRNA precursors driven by intronic Alu-Alu base pairing. Splicing is a molecular mechanism in which introns are removed from an mRNA precursor and exons are ligated to form a mature mRNA. Here, we show that Alu insertions into introns can affect the splicing of the flanking exons. We experimentally demonstrate that two Alu elements that were inserted into the same intron in opposite orientation undergo base-pairing, and consequently shift the splicing pattern of the downstream exon from constitutive inclusion in all mature mRNA molecules to alternative skipping. This emphasizes the impact of Alu elements on the primate-specific transcriptome evolution, as such events can generate new isoforms that might acquire novel functions.
| Alternative splicing enhances transcriptomic diversity and presumably leads to speciation and higher organism complexity, especially in mammals [1]–[3]. There are four major types of alternative splicing: exon skipping, which is the most prevalent form in higher vertebrates; alternative 5′ and 3′ splice site (5′ss and 3′ss) selection; and intron retention, which is the rarest form in both vertebrates and invertebrates [4],[5]. At least 74% (and probably much more) of human genes that contain introns produce more than one type of mRNA transcript through alternative splicing; however, it is unclear which of these products are biologically functional and which are non-functional products of inaccurate splicing [3], [6]–[8]. Thus, understanding the changes in the genome that dictate fixation of beneficial alternative splicing events or deleterious events (e.g., mutations leading to genetic disorders or cancer), or aberrant splicing events (noise in the system) is of great interest.
There are three known origins of alternatively spliced exons: 1) exon shuffling, which is a form of gene duplication [9]–[11]; 2) exonization of intronic sequences [12]–[16]; and 3) change in the mode of splicing from constitutive to alternative splicing during evolution [17],[18]. One mechanism responsible for the shift from constitutive to alternative splicing is accumulation of mutations in the 5′ splice site region. Here we set out to examine additional mechanisms involved in the transition from constitutive to alternative splicing.
The primate-specific retrotransposons called Alu are ∼280 nucleotides long. These are the most abundant retrotransposed elements in the human genome with about 1.1 million copies [16],[19],[20]. A large fraction of these Alu elements are located within intronic sequences, in both the sense and the antisense orientation relative to the mRNA, and can potentially form long regions of double-stranded RNA (dsRNA) [16], [21]–[25]. There are indications that extensive secondary structure occurs between Alu elements. The evidence is embedded in analyses of the RNA editing mechanism: The human transcriptome undergoes extensive adenosine to inosine RNA editing [23],[26],[27]. RNA editing is directed by an adenosine deamination mechanism catalyzed by specific adenosine deaminases, termed dsRADs (double-stranded RNA adenosine deaminases) or ADARs [26],[28],[29]. ADARs are required for the formation of the dsRNA molecules that serve as substrates for the deamination process [30],[31]. Hence, dsRNA regions formed between two Alus in opposite orientation within 2000 nucleotides of each other may serve as substrates for ADAR [21]–[25],[27],[28],[32].
More than 90% of known editing sites are found in Alu elements and editing occurs in sense and antisense pairs of Alus but not in flanking non-Alu sequences [23],[24],[33]. Recently, it was shown that a pair of inverted Alus located within the 3′UTR of EGFP mRNA serves as a substrate for A-to-I RNA editing that stabilizes the binding of the p54 protein to the mRNA. This causes nuclear retention of the mRNA and the silencing of EGFP expression [34]. Another example comes from the NARF gene, where formation of Alu-Alu dsRNA and its subsequent editing generates a functional 3′ splice site that is essential for exonization of that intronic Alu; moreover, editing in that Alu eliminates a stop codon and modulates the strength of exonic splicing regulatory sequences (ESRs). Interestingly, the nucleotides surrounding the editing site are important not only for editing of that particular site but also for editing at other sites located downstream in the same exon. It was also shown that the C nucleotide thought to pair with the edited site on the dsRNA is important for editing [35]. There is emerging evidence that the secondary structure of precursor mRNA plays a role in regulation of alternative splicing. However, in most studies the double-stranded structure was made up of only 10–40 base pairs and sequestered exonic or splice site sequences [36]–[47].
In this study, we have bioinformatically and experimentally evaluated the effects of intronic Alu elements on splicing. We found that different regulatory constraints act on Alu insertions into introns that flank constitutively or alternatively spliced exons. We further demonstrated that two Alu elements which were inserted into introns in opposite orientation have the potential to undergo base-pairing, as evident by RNA editing, and affect the splicing patterns of a downstream exon, by shifting it from constitutive to alternative. Finally, as Alu elements are abundant in introns, the findings we present suggest that the effect of intronic Alu elements on the transcriptome could be substantial, and could result in transcriptomic novelties. The new isoforms could then be subjected to purifying selection which will determine their fixation.
To examine potential effects of insertion of Alu elements into introns on splicing of the flanking exons, we downloaded data of human introns (hg18) and Alu elements and determined the intersected set using the UCSC genome browser and GALAXY [48],[49]. Overall, 730,622 Alu elements that reside within introns and 185,534 introns were extracted. This analysis showed that there are 85,126 introns that contain at least one Alu element; of these, 5009 introns contained at least two Alu elements in opposite orientation. The median length of introns containing at least one Alu element is 3829 base pairs (bp), whereas the median length of introns that do not contain an Alu is 521 bp (for intron length distribution see Figure S1). This suggests that double strand formation as a result of base pairing between two nearby Alu elements in opposite orientation might be common.
An antisense Alu and a sense Alu that are within 2000 nucleotides of each other can form dsRNA and be subjected to mRNA editing [21]–[24],[50]. To examine the configurations of possible Alu-Alu dsRNA we extracted data on 10,113 nucleotides that undergo mRNA editing in the human genome from Levanon et al. [24]. Intersection with data on all Alu elements in the human genome and human RefSeq intronic sequences yielded 953 Alu elements that are embedded in intronic sequences that undergo mRNA editing (see Materials and Methods). For each of the 953 edited Alu elements, the nearest Alu element in the opposite orientation was identified. For the vast majority of edited Alu elements (880 out of the 953; 92%), the closest Alu element in the opposite orientation resided within the same intron, with an average distance of 682 bp from the edited Alu. However, we found 73 cases (8%) where the nearest Alu element in the opposite orientation was in a different intron; in 61 of these cases it is at least 500 bp closer to the edited Alu element than the nearest Alu element in the opposite orientation in the same intron (Figure S2). In these cases, the average distance between the edited Alu and the nearest Alu in the opposite orientation is 1993 bp. In fact, in 43 out of the 61 cases the distance was less than 2000 bp (averaging 1122 bp). This close proximity between the two Alu elements, along with the evidence that at least one of them undergoes editing, suggests that these regions may base pair.
Since in 92% of edited Alu elements, the closest Alu element in the opposite orientation resided within the same intron, we decided to examine the splicing process in these cases. But first, we set to examine the distribution of Alu elements within datasets of exons conserved within human and mouse having different splicing patterns.
We analyzed three datasets of human-mouse orthologous exons and their flanking introns and exons: 1) conserved constitutively spliced exons (constitutively spliced in both species, 45,553 exons), 2) conserved alternatively spliced exons (alternatively spliced in both species, 596 exons), and 3) exons that are alternatively spliced in human and constitutively spliced in mouse (species-specific alternative exons; 354 exons). Analysis of Alu insertions into introns flanking these exons revealed that species-specific alternative exons exhibited the highest level of Alu insertions, followed by conserved alternative exons; the group with the fewest intronic insertions were conserved constitutively spliced exons. We calculated the density of Alu insertions, namely the number of Alus divided by the total intron length (and then multiplied by 1000 for convenience), in order to control for the fact that different intronic lengths might influence Alu insertion (see Materials and Methods). On average, 0.42 Alu elements were found per 1000 bp within the upstream introns and 0.41 Alu elements were found per 1000 bp within the downstream introns of constitutively spliced exons; 0.49 and 0.44 Alus per 1000 bp were found in the upstream and in the downstream introns of alternatively spliced exons, respectively; and 0.66 and 0.65 Alus were found per 1000 bp in the upstream and in the downstream introns of species-specific alternatively spliced exons, respectively. Thus, the number of Alu elements present in species-specific alternative exons differed significantly from that found in constitutively spliced exons (p-value = 7.16E-10, p-value = 5.22E-09, for upstream and downstream introns, respectively) and also differed from that in the alternatively spliced exons (p-value = 0.000503, p-value = 0.000014, for upstream and downstream introns, respectively).
Furthermore, analysis of the distribution of antisense and sense Alus upstream of conserved constitutively spliced exons revealed a selection against the presence of Alu elements adjacent to exons, specifically, against Alu elements in the antisense orientation (Figure 1). There are significantly fewer antisense Alus compared to sense Alus within 150 bp of constitutively spliced exons (p-value = 0.000012). Examination of the downstream intron did not reveal a significant bias (p-value = 0.056). This implies that a selective pressure exists against insertion of Alus in close proximity upstream to constitutively spliced exons; this bias is stronger against Alus in the antisense orientation than against the sense orientation.
B1 is a rodent-specific retrotransposed element of ∼150 nucleotides that has the same ancestral origin as Alu: the 7SLRNA [51],[52]. Like Alus, large numbers of B1 elements (a total of 331,015) reside within intronic sequences [16]. We found 236,036 B1 elements within 177,766 mouse introns that are found in GenBank (see Materials and Methods). In mouse, 70,516 introns (39.6%) contain B1 elements. Overall, there are 1.32 B1 elements per intron. The median length of introns containing at least one B1 element is 3278 bp, whereas the median length of introns that do not contain B1 is 636 bp. These results indicate that Alu and B1 containing introns are substantially longer than other introns.
We observed a significant correlation between the number of insertions of Alu and B1 elements within orthologous introns (Pearson correlation coefficient of 0.73 with p-value <0.0001). Namely, orthologous introns show the same tendencies for Alu and B1 insertion, although these events happened independently after the split of the mouse and human lineages. We then set out to analyze whether insertion of B1 into rodent introns was biased in terms of location and orientation as was the case for Alu in primates. Analysis of B1 insertions within the flanking introns of conserved constitutively spliced exons, conserved alternatively spliced, and species-specific alternatively spliced exons yielded the same trend as that of Alu insertions in human. There was no statistical difference in the density of B1 between conserved constitutively spliced and conserved alternatively spliced exons; however, the upstream introns of the 258 species-specific exons (alternatively spliced in mouse, but constitutive in human) were significantly more enriched with B1 elements than were the upstream introns of conserved constitutively spliced exons (p-value = 0.0012) or constitutively spliced downstream introns (p-value = 0.014). This was also the case when the regions upstream of exons that are alternatively spliced in mouse and constitutively spliced in human were compared to the upstream introns of conserved alternatively spliced exons (p-value = 0.042) but not the downstream introns (p-value = 0.155). Therefore, insertion of retrotransposed elements into intronic sequences is correlated with the mode of splicing of the flanking exons.
Five reports [53]–[57] indicate that de novo Alu insertions into intronic sequences in antisense orientation and in close proximity to the affected exon (between 19–50 nucleotides) cause the downstream exon to shift from constitutive splicing to full exon skipping (three cases) or to alternative splicing (two cases) (Table 1). This effect of Alu elements on adjacent exons may be due to the Alu structure. Alu elements are comprised of two very similar segments, termed left and right arms. When an Alu is located in a gene in the antisense orientation and transcribed it contributes two poly-T stretches to the mRNA precursor. These poly-T regions might act as polypyrimidine tract (PPT) and, in combination with downstream 3′ and 5′ pseudo splice sites, might act as pseudo-exon [58]. Hence, such antisense Alus that act as pseudo-exons might compete with nearby exons for the binding of splicing factors. These five cases of de novo Alu insertions imply that Alus located in close proximity to exons might affect splicing of adjacent exons. This and the finding of de-novo Alu insertions that affect splicing imply that this is an on-going evolutionary process, which may result in novel transcripts that are deleterious and inflict genetic diseases. On the other hand, a shift in the splicing pattern from constitutive to alternative might be advantageous in some cases, and could enable testing new mRNA options without eliminating the old ones. Moreover, such a shift could introduce a premature termination codons enabling the expression of truncated proteins at certain needed times or in specific cell types and could be delicately regulated by the levels of splicing regulatory proteins [59],[60].
In order to determine how many alternatively spliced exons are potentially regulated by the insertion of an antisense Alu, we used the alternative splicing track in the UCSC genome browser ([48] see also Materials and Methods). In 269 events (∼1.5% out of 17,151 alternatively spliced cassette-exons), an antisense Alu was found within 100 bp upstream of an exon (150 have additional sense Alu within 2000 bp), 491 (∼2.8%) events in which an antisense Alu was found within 150 bp (273 have additional sense Alu within 2000 bp), and 689 (∼4%) events in which an antisense Alu was found within 200 bp (373 have additional sense Alu within 2000 bp). Out of these 689 alternative exons, 525 (76.1%) are conserved between human and mouse (23 events were recorded as alternatively spliced also in mouse alternative splicing track in version mm9). Within the human genome, almost 85% of alternative cassette exon skipping events are conserved in mouse, however only 76% of the cassette exon skipping events that have an adjacent Alu in opposite orientation are conserved within mouse genome. This is statistically significant, implying that there is a bias for Alu in antisense orientation in the regulation of alternative exons within non-conserved alternative splicing events (χ2 test p-value = 1.6×10−8).
The above results suggest that stable insertion of Alus into introns is associated with the mode of splicing of the flanking exons—especially the downstream exon—and that most Alu-Alu dsRNA is formed between sequences within the same intron. To test this hypothesis, exon 3 of the human RABL5 gene was analyzed experimentally to examine the connection between intronic Alu and alternative splicing. A minigene containing exons 2 through 4 of the human RABL5 gene (a conserved gene within all vertebrate genomes) was cloned. Exon 3 of RABL5 is alternatively spliced in human and constitutively spliced in mouse, rat, dog, chicken, and zebrafish (see Figure 2 in [17]). Based on the phylogenetic relationships among the analyzed organisms, we conclude that the alternatively spliced variant is a derived form and the constitutively spliced variant is the ancestral one.
Six Alus have been inserted into the flanking introns of exon 3 since the last common ancestor of human and mouse: two in the upstream intron and four in the downstream intron (Figure 2A). Alu4, which is located in the downstream intron, resulted from insertion of an Alu within another Alu and will be regarded as one Alu (see also Text S1). The minigene was transfected into 293T cells and the splicing products were examined following RNA extraction and RT-PCR analysis. Exon 3 in the RABL5 minigene is alternatively spliced with approximately 40% inclusion (Figure 2B, lane 1). Removal of all intronic Alus shifted splicing from alternative to constitutive (Figure 2B, compare lanes 1 and 2). This indicates that the insertion of Alus into the flanking introns during primate evolution shifted exon 3 splicing from constitutive to alternative.
Our experiments revealed that the orientation and position of the Alus within the upstream intron affected splicing of exon 3. Deletion of the Alus in the upstream intron, namely Alu1 and Alu2 (Δ1+2), had the same effect as deleting all Alus (Figure 2B, lanes 7 and 2). The same effect was observed if one of the Alus was deleted and the other was replaced with a 270-nucleotide non-Alu intronic sequence (Figure 2B, lanes 18 and 21, see also Text S2). The replacement of each of the intronic Alus with a non-Alu intronic sequence of a similar length eliminated the possibility that the effect observed after deletion related to shortening of the intron.
The Alus in the downstream intron, however, had a little or no effect on splicing (Figure 2B, lane 13). Taken together, it seems that the shift from constitutive to alternative splicing in the lineage leading to human is mediated mainly by Alus 1 and 2. Deletion of Alu1 or replacement with a 270-nucleotide non-Alu intronic sequence resulted in almost complete exon skipping (Figure 2B, lanes 3 and 17, respectively). Replacement or deletion of Alu2 resulted in constitutive exon splicing (lanes 4 and 20).
Interestingly, Alu1 and Alu2 have opposite effects on splicing. Deletion of both Alus has the same effect as deleting only Alu2. Therefore, we concluded that Alu2 is dominant over Alu1. The dominance of Alu2 is also supported by two other observations. First, if all Alus except Alu2 are removed, we observe almost total exon skipping (Figure 2B, lane 14). This indicates that Alu2 is a negative regulator of exon 3 recognition, unless Alu1, which is in opposite orientation to Alu2, is present (compare lanes 13 and 14). Furthermore, deletion or replacement with a 270-nucleotide non-Alu intronic sequence of Alu2 in combination with any additional intronic Alus leads to constitutive splicing (lanes 7, 8, 10, 18–22). In the absence of Alu1 and the presence of Alu2, the dominance of Alu2 over the other Alus is observed, leading to exon skipping (lanes 9, 12, 14, 15, and 16). As expected, in the presence of both Alu1 and Alu2, deletion of Alus from the downstream intron had a marginal effect on splicing (Figure 2B, lanes 5, 6, 11, and 23).
We demonstrated that the antisense orientation of Alu2 is essential for alternative splicing of exon 3. We first noted that the exact Alu family is not an important factor in determining splicing pattern: replacement of Alu1 of the Jo family with the sequence of Alu3 from the Sx family (both Alus are in the sense orientation) did not affect the splicing pattern (Figure 2B, lane 24). Thus, the important factor is the presence of Alu1 in sense orientation. Our analysis showed that only when Alu2 is in the antisense orientation and Alu1 is in the sense orientation is alternative splicing of exon 3 observed (Figure 2B, lanes 25–26). These results indicate that the two Alus in the upstream intron regulate alternative splicing of exon 3, whereas the three downstream intronic Alus have no apparent effect on splicing of that exon. Moreover, Alu2 in the antisense orientation suppressed inclusion of exon 3, whereas Alu1 in the sense orientation antagonized the effect of Alu2.
How do the two intronic Alus regulate alternative splicing of exon 3? It is apparent that if Alu2 alone is present in the mRNA precursor, the exon is always skipped. We therefore postulated that in a population of the mRNA precursors that contain both Alu1 and Alu2, the two might form dsRNA formation (Figure 3A). This sequestration leads to exon inclusion; in the fraction of mRNA precursors with no base pairing between Alu1 and Alu2 the exon is skipped. To support this hypothesis, we examined whether RNA editing occurred in intron 2; editing would be indicative of formation of dsRNA. We searched the human EST/cDNA dataset and found five different mRNAs sequences containing intron 2 and comparison with genomic sequence indicated extensive RNA editing in both Alu1 and Alu2 (marked in red in Figure 3B). The region of the editing was found to be in the middle of both Alus. This suggests that these two Alu regions are in a double-stranded form.
To further confirm the formation of dsRNA, we generated cDNA from a neuroblastoma cell-line using specific primers (Figure 3C). By using primers that hybridize in the exons flanking intron 2, we were able to observe a small population of mRNA molecules that contain intron 2; the majority of mRNAs are spliced (Figure 3C lane 1). We enriched the intron-containing fraction using primers that hybridize within the intron and within the downstream exon (Figure 3C, lane 2). Sequencing of the higher molecular weight PCR product using primer to Alu2 allowed us to identify four editing sites within it (Figure 3D).
To confirm the importance of pairing between Alu1 and Alu2 on editing, we used the ΔAlu1 mutant (see Figure 2 lane 3) that led to a full exon skipping and examined the effect on editing within Alu2. There is one editing site within Alu2 that is dependent on the presence of Alu1; without Alu1 no editing at this site was observed (Figure 3E, the site is also highlighted in green in Figure 3B). The other putative editing sites found in EST/cDNA show relatively low level of editing in the minigene.
We next set to examine if the distance between exon 3 and the intronic Alus and the distance between Alu1 and Alu2 were important factors in splicing of exon 3. Alu2 is located 24 nucleotides upstream of exon 3. We identified the putative branch site of intron 2 and inserted an 800-nucleotide non-Alu intronic sequence upstream of the branch sequence and downstream of intronic Alu1 and Alu2 (marked B in Figure 4A; see Text S3). This insertion caused a shift from alternative to constitutive inclusion of exon 3 (Figure 4B, compare lane 1 and 2). Only when this insertion was shortened to less than 68 nucleotides did we begin to detect restoration of alternative splicing of exon 3; the level of skipping was further elevated when the inserted sequence was shortened to 56 or to 44 nucleotides (Figure 4B, lanes 3–9). To rule out the possibility that the sequence that was inserted contained intronic splicing regulatory sequences, we designed a fragment of 25 nucleotides free from known splicing regulatory sequences (see Materials and Methods). We inserted this sequence into site B and also duplicated and triplicated this sequence to generate 50 and 75 nucleotides insertions. The longer is the inserted sequence, the higher is the inclusion level (Figure 4B, lanes 10–13). This indicates that the distance between the intronic Alu2 from exon 3 affects the mode of splicing.
We also examined the effect of the distance between Alu1 and Alu2 on the splicing of exon 3. The insertion of the same fragment of 800 nucleotides between the two elements (marked A in Figure 4A) led to substantial reduction in the inclusion level of exon 3, although alternative splicing was still observed (Figure 4C, compare lane 1 and lane 2). We then shortened this sequence, ultimately to 24 nucleotides; when the distance was shorter than 550 nucleotides, almost complete inclusion of exon 3 was observed (Figure 4C lanes 2–9). Our results indicate that the distance between Alu1 and Alu2 is important for maintaining the alternative splicing of exon 3; however, it is not as important as the distance between the intronic Alu elements and exon 3. We also note that increasing the distance between exon 3 and Alu2 leads to exon inclusion, whereas increasing the distance between Alu1 and Alu2 enhances exon skipping.
Figure 2 demonstrates that Alu2 suppresses the inclusion of exon 3. We therefore analyzed the sequence of Alu2 to determine what regions might be critical for this effect. Figure 5A shows the sequence of Alu2 and the mutations made. We deleted each of the two arms of Alu2 separately (Figure 5B, lanes 6 and 8) and mutated putative splicing signals (Figure 5B, lanes 2–5, 9–11). We found that the left arm of Alu2 is involved in the constitutive-to-alternative shift. Deletion of the left arm enhanced inclusion, whereas deletion of the right arm caused only a marginal effect (Figure 5B, compare lanes 1, 6, and 8). Mutations in the putative splicing signals in the right and left arms of Alu2 did not affect the splicing pattern (Figure 5B, lanes 2–5, 9–13). Our results do not support the possibility that the left arm functions as a pseudo-exon that abolishes or reduces selection of the exon 3 by competing with splicing factors (see also [58],[61]). However, analysis of this data is not straight forward, because deletion of the entire right arm together with the left-arm-polypyrimidine tract (LPPT), which produced a short Alu2 sequence, caused complete skipping of exon 3 (Figure 5B, lane 7). Insertion of a complementary sequence to the short Alu2 sequence along with its upstream intronic sequence (to complete an Alu-like length of 280 nucleotides), 105bp upstream to the original short Alu2 (mimicking the original distance of Alu2 from Alu1), either in the sense or antisense orientation, did not affect full skipping of exon 3. Deletion of the right arm alone or the LPPT alone had a marginal effect on splicing of exon 3 (Figure 5B, compare lanes 6 and 9 to 7). These results imply that multiple sequences along Alu2 combine to suppress the recognition of exon 3.
Based on the location of the editing sites shown in figure 3B we concluded that in a large part of the left arm of Alu2 there is no editing, suggesting that this part might not participate in a dsRNA structure. Within the left arm we identified a potential sequence, which is not part of the Alu1-Alu2 putative pairing alignment. This region has the potential to form an internal stem-loop structure (Figure 5C, upper alignments). Deletion of this region, replacement of this sequence with a similar stem-loop structure of a different sequence, creation of fully paired stem structure, or disruption of the stem structure all caused full exon 3 skipping. These results imply that the sequence, rather than the potential secondary structure, of this region is important for the inclusion of exon 3. This sequence contains two putative SC35 binding sites. However, mutations that eliminated these potential binding sites without generating another known splicing regulatory sites had no effect on splicing of exon 3, indicating that this is not the sequence involved in the regulation (Figure 5C, lane 2). There are no other potential binding sites for known splicing regulatory factors in this sequence (based on [62]–[64]). Finally, addition of a potential complementary sequence to Alu1 did not effect splicing of exon 3 (not shown). Although Alu2 functions primarily to inhibit exon 3 selection, the above sequence within Alu2 enhances the inclusion of exon 3. Formation of a duplex between Alu1 and Alu2 is needed in order to present this intronic enhancer sequence properly for its effect on splicing of exon 3.
There are over 0.5 million copies of Alu elements in introns of human protein coding genes [65], yet their function in regulation of gene expression is largely unknown. Here we show that intronic Alus are not ‘neutral’ elements; they affect splicing of flanking exons. Some of these effects can be directly linked to the shift from constitutive to alternative splicing during primate evolution. The regulation demonstrated here involves both positive and negative effects of Alu element in antisense orientation, in close proximity, and upstream to the regulated exon. This complex regulation causes the downstream exon to shift from constitutive to alternative splicing. There are several examples [53]–[57] of de novo insertions of Alu elements within introns that result in skipping of the adjacent exons. In three of the reported cases, the insertion of the Alu in the antisense orientation caused a total skipping of the adjacent exon.
Exon 3 of RABL5 gene, analyzed in this study, is alternatively spliced in human and constitutively spliced in mouse, rat, dog, chicken, and zebrafish. Six Alus have been inserted into the flanking introns of exon 3 since the last common ancestor of human and mouse. Alu2 was inserted in the antisense orientation just upstream of exon 3 and functions as a negative element that suppresses exon 3 selection. This negative effect is partially reversed by another Alu present in the same intron in the sense orientation. Although we were not able to fully resolve the mechanism by which the two Alus regulate alternative splicing of the downstream exon, we provide evidence that regulation requires the formation of a double-stranded region between the two Alus and a combination of negative and positive sequences located in Alu2. The end result of this complex regulation is a shift from constitutive to alternative splicing of the downstream exon. This results in a new primate-specific mRNA isoform that could acquire novel functions, as well as maintaining the original mRNA. Moreover, such a shift could introduce a premature termination codon resulting in truncated proteins that might have regulatory roles at certain times or in specific cell types as could be delicately determined by the levels of splicing regulatory proteins [59],[60].
Introns in humans are considerably longer than their mouse counterparts, mostly due to the presence of Alu elements [66]. Introns that flank alternatively spliced exons are longer than introns that flank constitutively spliced ones [4],[67]. We found a correlation between the splicing pattern of exons and the presence of Alus in the flanking introns. First, alternatively spliced exons are flanked by introns containing more Alus compared with introns flanking constitutively spliced ones, even when controlling for the difference in intron lengths. Second, more Alus are present in human introns than are corresponding mouse B1 elements in the orthologous mouse introns. This second observation correlates with the finding that there are more species-specifically, alternatively spliced exons in human than in mouse (354/612 alternatively spliced events in human and 258/612 alternatively spliced events in mouse; χ2, p –value <0.01).
It was suggested that intron complementarities formed by multiple copies of Alu could help define and increase the splicing efficiency of very large metazoan introns [68]. However, it may be also possible that formation of a long and stable double-stranded structure in the upstream intron, especially near the splice site as in the case studied in this manuscript, reduces the ability of the splicing machinery to properly recognize the downstream exon, leading to slower splicing kinetics or suboptimal exon selection and, thus, to intron retention or exon skipping. Supporting the hypothesis that formation of dsRNA in introns might slow splicing is a recent publication showing that formation of dsRNA during pre-microRNA formation can slow splicing of the intron where the microRNA resides [69].
Although the effect of intronic retroelements on the splicing of flanking exons is presumably not a general trend that applies to all exons, it is relevant to a certain fraction of alternatively spliced exons (1.5% to 4% of the alternatively skipped exons in human).
Our analysis indicated that the presence of Alu elements is correlated with the mode of splicing of adjacent exons. There is an ‘exclusion zone’ in intron sequences flanking exons, where insertion of Alu elements is presumably under purifying selection. The length of this ‘exclusion zone’ is similar to that of the human-mouse conserved sequences flanking alternatively spliced exons (∼80–150 nucleotides). This is presumably indicative of regions where the presence of intronic splicing regulatory sequences can affect alternative splicing of the adjacent exon [5],[18],[70],[71]. Alus might be excluded from the proximal intronic sequences flanking constitutively spliced exons because Alus were never inserted into these regions or because Alus were inserted in an equal proportion in all gene regions (intronic and exonic) but we currently observe only those Alus that have escaped purifying selection. The major burst of Alu retroposition took place 50–60 million years ago and has since dropped to a frequency of one new retroposition for every 20–125 new births [72],[73]. As some of these insertions were deleterious and thus selected against, we probably detect intronic Alus that are neutral, mildly deleterious, or beneficial to human fitness. Some of these beneficial intronic Alus presumably altered splicing of the flanking exons and resulted in the generation of new isoforms that presented an advantage during primate evolution and were thus fixated in our genome. The research described here sheds light on how Alu elements have shaped the human genome.
A dataset of 596 alternatively spliced exons, conserved between human and mouse, was derived from a previously compiled dataset [5]. In addition, 45,553 human-mouse conserved constitutively spliced exons were obtained from Carmel et al. [74]. Species-specific exons (354) were extracted from a dataset of 4,262 human-mouse orthologous exons that are suspected to splice differently in human and mouse based on initial EST analysis [74]. For details of how the datasets were built see [17].
Introns and exons for human (Homo sapiens, Build 35.4) and mouse (Mus musculus, Build 34.1) were extracted from the Exon-Intron Database (http://hsc.utoledo.edu/depts/bioinfo/database.html) [75]. These intron sequences were analyzed with RepeatMasker software version 3.1.0 [76] (www.repatmasker.org) using Repbase update files [77].
Since Alus are primate specific, the distribution was computed only from human flanking introns. The density of Alu elements was calculated per 1000-bp intron length according to the following equation:N = number of Alus within the intron; L = the length of the intron; Alu density = Alu density. For the detection of retrotransposed elements, we used the RepeatMasker (http://www.repeatmasker.org) software version 3.1.0 [76] and Repbase update [77].
A T-test was used to calculate statistical differences between two populations; for χ2 test with 2×2 contingency table, Fisher's exact test was used.
The alternatively skipped exons in the human genome (build hg18) were extracted by downloading the knownAlt table from UCSC genome browser [48]. The presence or absence of Alu within the upstream intron was determined using RepeatMasker tables downloaded from UCSC. The conservation of these introns was analyzed using MAF pairwise alignments between the human genome (build hg18) and the mouse genome (build mm9) downloaded from UCSC genome browser. The intersections between these tables were done using the Galaxy sever [49].
We extracted data of on 10,113 nucleotides that undergo mRNA editing in the human genome from Levanon et al. [24]. The UCSC genome browser [58] was then used to extract data of all Alu elements in the human genome (build hg18) using the RepeatMasker [76] annotations, and to extract human RefSeq intronic sequences. Intersection of these three datasets yielded 953 Alu elements that are embedded in intronic sequences and undergo mRNA editing.
The RABL5 (RAB member RAS oncogene family-like 5) minigene was generated by amplifying a human genomic fragment using PCR reaction. Each primer contained an additional sequence encoding a restriction enzyme. The PCR product was restriction digested and inserted into the pEGFP-C3 plasmid (Clontech) and sequenced to confirm that the desired construct was obtained. The RABL5 minigene, contains exons 2 through 4 (2.7 kb). The intron replacements with the RABL5 Alu1, Alu2, and Alu3 were done by PCR opening of the plasmid lacking the specific Alu (#1, #2 or #3) and ligation with a fragment of 270 intronic-nucleotides taken from a PCR amplification directed to the IKBKAP gene intron number 20 (primer forward, 5′AGAATCGTGACACTCATCATATAAAGGAGG3′; and primer reverse, 5′CAAAACATTAGTATAGATCTTTCCAATACA3′). The 800-nucleotide insertion1 was taken from PCR amplification directed to the IMP gene intron number 11 (primer forward, 5′ATCACTCTGCACTTTCTCCCAT3′; primer reverse 5′ACCATGTCCACTTCATCCAGTTC3′). Insertion2 is a 25-bp sequence, free of any known splicing regulatory elements, that was doubled or tripled into 50-bp and 75-bp sequences, ( 5′CTATCTGATAAGCTGCGAGCAATT3′).
Endogenous PCR amplification was done on a cDNA template originating from a neuroblastoma cell-line (SH-SY5Y). Amplification was performed for 30 cycles, consisting of denaturation for 30 seconds at 94°C, annealing for 45 seconds at 52°C or 56°C, and elongation for 1 minute at 72°C. The products were separated in a 1.5% agarose gel. The upper PCR product was Topo-ligated (Invitrogen) and sequenced. The primers used were: forward (exon 2), 5′CAGAATCTTCTGACATCACTG3′; or forward (intron 2), 5′GTGAGCCCTGACAAATCTGTGT3′; and reverse (exon 3) 5′GTTGCTGGTAACATGCGGGTTC3′.
For details see [78].
|
10.1371/journal.pgen.1002977 | The p38/MK2-Driven Exchange between Tristetraprolin and HuR Regulates AU–Rich Element–Dependent Translation | TNF expression of macrophages is under stringent translational control that depends on the p38 MAPK/MK2 pathway and the AU–rich element (ARE) in the TNF mRNA. Here, we elucidate the molecular mechanism of phosphorylation-regulated translation of TNF. We demonstrate that translation of the TNF-precursor at the ER requires expression of the ARE–binding and -stabilizing factor human antigen R (HuR) together with either activity of the p38 MAPK/MK2 pathway or the absence of the ARE-binding and -destabilizing factor tristetraprolin (TTP). We show that phosphorylation of TTP by MK2 decreases its affinity to the ARE, inhibits its ability to replace HuR, and permits HuR-mediated initiation of translation of TNF mRNA. Since translation of TTP's own mRNA is also regulated by this mechanism, an intrinsic feedback control of the inflammatory response is ensured. The phosphorylation-regulated TTP/HuR exchange at target mRNAs provides a reversible switch between unstable/non-translatable and stable/efficiently translated mRNAs.
| For immediate response and better control of gene expression, eukaryotic cells have developed means to specifically regulate the stability and translation of pre-formed mRNA transcripts. This post-transcriptional regulation of gene expression is realized by a variety of mRNA-binding proteins, which target specific mRNA sequence elements in a signal-dependent manner. Here we describe a molecular switch mechanism where the exchange of two mRNA-binding proteins is regulated by stress and inflammatory signals. This switch operates between stabilization and efficient translation of the target mRNA, when the activator protein of translational initiation binds instead of the phosphorylated destabilizing protein, and translational arrest and degradation of the target, when the non-phosphorylated destabilizing protein replaces the activator. This mechanism is specific to the mRNA of the inflammatory cytokine tumor necrosis factor (TNF)-α and the mRNA of its regulator protein TTP and, hence, enables fast inflammatory response and its stringent feedback control.
| TNF is a master cytokine of inflammatory signaling of macrophages. Its biosynthesis is tightly controlled to allow rapid secretion but also to avoid delay or leakiness in its down-regulation, which could result in exaggerated or persistent inflammation. The levels of regulation comprise transcription, processing, nuclear export and stability of the TNF mRNA, translation of pro-TNF, and shedding of TNF (reviewed e.g. in [1]). Pro-TNF contains a leader sequence of 79 amino acids (for mouse TNF) and is synthesized as a type II membrane protein [2], [3]. After initiation of ribosomal translation of TNF mRNA, followed by SRP-mediated arrest of ribosomal synthesis of the nascent protein chain and docking to the ER membrane, the C-terminal part of TNF containing the potential cleavage site is synthesized into the lumen of the ER. Subsequently, pro-TNF is transported in an LPS-stimulated manner from the trans Golgi network to the cell surface using tubular carriers that fuse with the recycling endosome [4]. At the cell surface, pro-TNF is cleaved and released by the TNF-converting enzyme TACE/ADAM17 [5].
The p38 MAPK/MK2/3 pathway [6] regulates TNF-biosynthesis mainly at the translational level. Inhibition of this pathway by small molecules, such as SB203580 or SB202190, or deletion of its components, such as the downstream protein kinase MK2, lead to a significant reduction of LPS-induced TNF production of macrophages although the LPS-stimulated increase in TNF mRNA concentration remains almost unaltered and the stability of mature TNF mRNA in these cells is only sightly reduced [7]–[9]. The post-transcriptional regulation of TNF biosynthesis by the p38 pathway depends on the AU-rich element (ARE) in the 3′ non-translated region of TNF mRNA [9], [10]. So far the molecular mechanisms regulating ARE-dependent translation of pro-TNF via phosphorylation are not understood. However, various mRNA- and ARE-binding proteins have been identified as substrates of the p38 MAPK pathway, e.g. hnRNP A0, tristetraprolin (TTP), KSRP and poly(A)-binding protein 1 [11], [12], [13], [14], [15]. The mRNA-ARE and corresponding ARE-binding proteins (ABPs), such as TTP, KSRP, HuR, TIA-1 and AUF1, are mainly held responsible for the regulation of mRNA metabolism governed by exosome-, PARN- and CCR4/Not1-dependent degradation of mRNAs or by their storage in discrete cytoplasmic foci (reviewed in [16] and [17]–[19]). For example, KSRP stimulates the rapid decay of ARE-containing mRNAs and its activity is inhibited via direct phosphorylation by p38 MAPK providing a mechanism of stress-dependent stabilization of ARE-containing mRNAs [13]. In contrast, HuR (ELAV) is a factor of constitutive nuclear and cytoplasmic stabilization of ARE-containing mRNAs [20], [21], which also binds to the ARE of TNF mRNA [22], [23]. However, besides its mRNA stabilizing function, HuR also influences translation of specific mRNAs as measured by the association of these mRNAs to the polysomal fractions of cell lysates separated by density centrifugation. Gene-silencing of HuR blocks polysomal localization of cytochrome C- and nucleolin-mRNA indicating a positive effect of HuR on translation of these mRNAs [24], [25]. In contrast, deletion of HuR increases the long-term shift of TNF mRNA to polysomal fractions after LPS treatment of macrophages [26] and its overexpression leads to a reduction of TNF-mRNA in polysomal fractions [27] indicating also an inhibitory effect of HuR on translation.
The inhibitory effect of TTP in the regulation of TNF production first became obvious by significant cachexia in the TTP knockout mouse [28], which was explained by increased TNF concentrations and a feedback effect of TTP on TNF production by its binding to the ARE and destabilization of the TNF mRNA [29]. Subsequently, TTP could be identified as a destabilizing factor for various ARE-containing mRNAs, including its own mRNA (reviewed in [16]). In LPS-induced TNF biosynthesis TTP is genetically downstream to p38 MAPK and MK2, since its deletion neutralizes the defect in LPS-induced TNF production seen in MK2 knockout mice [30]. MK2 directly phosphorylates TTP [11]. It is proposed that phospho-TTP/mRNA complexes are sequestered by 14-3-3 binding proteins [15] and that phospho-TTP is unable to recruit deadenylases [17], [19] resulting in target mRNA stabilization. Via phosphorylation of SRF, MK2 also contributes to transcriptional activation of the TTP gene [31]. Interestingly, in MK2 knockout and, especially, in MK2/MK3 double knockout (DKO) macrophages a strong reduction of the TTP concentration is observed [32] suggesting a major role of the p38/MK2/3 pathway in the regulation of TTP expression.
The predominant translational regulation of TNF by the p38 MAPK/MK2/3 pathway raises the question of the role played by certain direct substrates of these protein kinases and the mechanisms involved in this translational regulation. Here, we reconstituted MK2-dependent translational regulation of TNF in immortalized MK2/3-deficient mouse macrophages by re-introducing MK2, its catalytic dead mutant or, as a control, GFP alone. We analyzed the requirements of MK2-dependent translation of native TNF mRNA for the presence of ABPs, such as TTP and HuR, and for the mRNA-binding MK2-substrate Ago2. In this analysis, the combination of cytosol/ER-fractionation and subsequent polysome profiling provides additional mechanistic insights into the translational regulation of pro-TNF. The regulatory mechanisms of translation of TNF mRNA elucidated by this approach are also valid for TTP mRNA and contribute to the stringent feedback regulation of the inflammatory response.
We generated an immortalized macrophage cell line from MK2/MK3 double-deficient mice [32] by expression of v-raf and v-myc in bone marrow derived macrophages (BMDM). For restoration of the “wild type” situation in these macrophages, we subsequently expressed MK2 by stable retroviral transduction using pMMP-MK2-IRES-GFP. As knockout control for this cell line, we used pMMP-IRES-GFP. To differentiate between the effects of catalytic activity of MK2 and the effects of MK2-dependent stabilization of p38 in the binary complex [33], we also rescued the cell line with the catalytic dead mutant of MK2 (pMMP-MK2K79R-IRES-GFP) [31]. The generation of different cell lines by retroviral transduction after initial immortalization excludes the unwanted influence of different random events of immortalization in the cell lines to be compared. After retroviral transduction, cell lines were sorted and selected by preparative FACS for comparable expression of GFP. The level of expression of MK2 in the rescued cell lines was comparable to its level in immortalized wild type BMDM and in RAW264.7 cells (Figure S1).
We compared basal and LPS-dependent expression and phosphorylation of the central components of the pathway (p38, MK2) and of relevant substrates of MK2/3 (TTP, NOGO-B) in the cell lines (Figure 1A and Figure S2A). As controls, we also analyzed expression of HuR, TIA-1, KSRP, GFP and GAPDH. While these controls showed comparable expression in both cell lines and were not induced by LPS, we detected strong induction of p38 activity by LPS in both cell lines. Although decreased stabilization of p38 protein by the lack of MK2/3 leads to reduced p38 concentrations in GFP-transduced MK2/3-deficient cells, the activity of p38, detected by the antibody which only detects the dual phosphorylation in the activation loop, was similar in MK2-rescued and GFP-transduced cells. Obviously, increased p38 activation compensates for its lower expression not only in neurons [34], but also in macrophages. As expected, MK2 is only detected and activated by LPS-treatment in MK2-rescued cells. Interestingly, there is a rapid induction of the MK2/3 substrate TTP by LPS-treatment, qualifying TTP as an immediate early gene. This LPS-induced expression of TTP is strongly reduced in GFP-transduced cells, a fact already known from MK2- and MK2/3-deficient primary cells [32]. In addition, a lack of phosphorylation of the ER membrane resident MK2 substrate NOGO-B, which is phosphorylated in its cytoplasmic domain [35], is detected in GFP-transduced compared to MK2-rescued cells based on the loss of the slower migrating phosphorylated isoform of NOGO-B.
Since the kinetics of p38/MK2 activation and TNF production in macrophages is fast (MK2 activity peaks after 20 min, maximum of TNF production is reached after about 60 min), we determined TNF-mRNA and -protein concentration 1 h after LPS-treatment. The relative intracellular TNF mRNA concentration as represented by the TNF mRNA/actin mRNA ratio and the length of the polyA-tail of TNF-mRNA do not significantly differ between MK2-rescued and GFP-transduced cells (Figure 1B and Figure S3). In contrast, both pro-TNF protein and secreted TNF are significantly increased in MK2-rescued cells compared to the GFP-transduced control (Figure 1C, 1D). Thus, the translational control of TNF biosynthesis by MK2 is clearly reflected in this cellular system. The catalytic activity of MK2 is necessary for this control, since rescue of the macrophages with the kinase-dead mutant MK2-K79R does not release the translational repression of TNF (Figure S2).
To monitor ribosome occupancy of TNF mRNA in these cells as a measure of translational initiation and elongation, we applied density gradient centrifugation to distinguish between TNF mRNPs, TNF mRNA containing monosomes and polysomes. Furthermore, since pro-TNF is synthesized as a type II membrane protein by ER-directed translation, we decided to combine cytosol/ER-fractionation by saponin treatment, which was modified after [36], with subsequent density centrifugation. A typical fractionation is shown in Figure 2A. While mRNPs, ribosomal subunits, scanning ribosomal subunit, initiated translation and stalled ribosomes with signal recognition particle (SRP) bound to the nascent peptide chain before docking to the ER membrane should distribute between fractions 1–4, polyribosomes are expected in fraction 6 and above. Polyribosomes containing mRNAs of secreted or membrane proteins, such as pro-TNF, are expected in the polyribosomal fractions of the ER subfraction, while polyribosomes containing mRNAs of cytosolic proteins together with all monosomes including SRP-stalled monosomes with nascent proteins are expected in the cytosolic subfraction. The overall distribution of TNF mRNA after LPS-stimulation differs between MK2-rescued and GFP-transduced cells as seen for the cytoplasm/ER ratio, which is significantly lower in MK2-rescued cells (Insert in Figure 2A and Figure S4 for absolute mRNA levels), while the distribution of ß-actin mRNA is not significantly different. Thus MK2 is required for preferential ER localization of TNF mRNA. We then performed polysome profile analyses of specific mRNAs. To ensure biological significance, we always performed at least two independent experiments comprising separate LPS-stimulation of cells, density gradient fractionation of cell lysates and qRT-PCR detection of specific mRNAs. Data were only considered significant in the instances where the biological repeats yielded the same qualitative results. The repeats of key experiments are displayed in Figure S5. As seen in the polysome profiles for TNF mRNA (Figure 2B), its cytosolic population occurs mostly in the free mRNP fraction. Only a low concentration of monosomal complexes of TNF mRNA is detected in the cytosolic fraction of MK2-transduced cells and, to a slightly lower degree, also in GFP-transduced cells. In the ER fraction, a clear peak of larger polysomal complexes of TNF mRNA (fractions 9–10) is seen in the presence of MK2 and this peak is completely missing in the absence of MK2 in cells transduced with GFP only. In parallel, the free mRNP signal is decreased for TNF mRNA in the presence of MK2. The existence of the larger polysomal complexes of TNF mRNA depends on catalytic activity of p38 and MK2, since the peak corresponding to larger polysomal complexes is reduced after treatment of the cells with the p38 inhibitor SB202190 and completely disappears for cells expressing the catalytic dead MK2 mutant instead of wild type MK2, respectively (Figure 2B). As a control, efficient translation of actin mRNA in ER and cytosol fractions is detectable for both MK2-rescued and GFP-transduced cells (Figure 2C). The finding that ß-actin mRNA coding for a cytosolic protein is also translated at the ER indicates the fact that not all proteins synthesized at the ER are secreted or integral membrane proteins [37] and that ER-associated mRNAs serve a global role also for translation of cytosolic proteins [38]. As further control, we monitored the mRNA distribution of the secreted chemokine KC (Cxcl1), which is also synthesized as pro-protein at the ER and which is known to be regulated by MK2 at the level of mRNA stability, but not at the level of translation [32]. KC mRNA distribution is independent of the presence of MK2 (Figure 2D). Efficient translation of this mRNA proceeds mainly at the ER. However, KC mRNA peaks in earlier polysome fractions (around 7–8) compared to TNF mRNA (peak at fractions 9–10), since the ORF of KC mRNA is 291 nucleotides only, compared to 708 nucleotides of TNF mRNA, allowing only a smaller number of ribosomes to elongate at the same mRNA molecule. In contrast to ß-actin mRNA, KC mRNA in the cytosolic fraction is mainly detected in monosomes. These monosomes probably represent translationally initiated monosomes with the SRP-arrested nascent protein chain. In cytosolic fractions the amount of ribosome-free KC mRNPs is comparable to the amount of KC mRNA in monosomes. In contrast, the amount of ribosome-free TNF mRNPs is significantly higher than the amount of monosomal TNF mRNA, indicating that initiation is the critical step for TNF mRNA translation. Taken together, this fractionation analysis clearly demonstrates a specific role for the catalytic activity of the protein kinase MK2 in the regulation of ER-directed translation of pro-TNF in macrophages.
To demonstrate that the cell lines generated reflect the situation in primary BMDMs we extended the analysis of LPS-induced MK2-dependent translation of TNF mRNA to wild-type, MK2-deficient and MK2/3 double-deficient primary BMDM. There are no qualitative differences in the overall polysome profiles of wild type and MK2/MK3 double-deficient BMDM (Figure 2E). However, after 1 h of LPS-stimulation TNF mRNA is detected in a clear polysomal peak for wild type cells only, while in MK2- and MK2/MK3-deficient cells this peak is missing (Figure 2F). As control, distribution of ß-actin mRNA in the polysome profile does not show clear differences between wild type and knockout BMDM (Figure 2G).
We postulated that proteins involved in the p38/MK2-dependent regulation of translation should exist in specific fractions of the monosome/polysome profile or the cytosol/ER fractions of LPS-treated macrophages depending on the presence of MK2. Therefore, we analyzed the relative concentration of various candidate proteins in the different fractions of lysates of LPS-stimulated MK2-rescued or GFP-transduced macrophages (Figure 3). In the density centrifugation of total cell lysates (Figure 3A), MK2 and p38 - as freely diffusible small proteins - are mainly present in the ribosome-free fraction, which contains also mRNPs. The small ribosomal protein S6, which was used to monitor ribosome distribution, is present in monosomal and polysomal fractions independent of the presence of MK2, indicating that there is no general effect of MK2 on translation (cf. also Figure 2C, 2D). Transcript regions free of bound ribosomes are cleaved by RNaseA treatment resulting in destruction of polyribosomes (Figure 3B) and in S6 being shifted to monosomes (Figure 3A).
The MK2 substrate NOGO-B is present in the ribosome-free mRNP, the monosomal and polysomal fractions, indicating that monosomes and polysomes may dock to the ER. A double band for NOGO-B, characteristic of the slower migrating phosphorylated and faster migrating non-phosphorylated isoform, is only seen in MK2-rescued cells in all fractions indicating that NOGO-B is a specific substrate for MK2 also in macrophages (cf. Figure 1A) and that phosphorylation by MK2 does not change its overall distribution in the gradient. Since remaining ER structures cannot be degraded by RNase A, NOGO-B distribution is not completely changed after RNase treatment and seems not to be a ribosome-associated protein.
The distribution of various mRNA-binding proteins (TIA-1, KSRP, TTP, HuR, and Ago2) was analyzed. RNA binding of TTP, HuR and Ago2 is necessary for their distribution in the gradient, since RNase A pre-treatment leads to almost complete disappearance of these proteins from the gradient (Figure 3A). As for NOGO-B, the lack of phosphorylation of TTP by MK2 in the GFP-transduced macrophages does not significantly change the overall distribution of TTP in the gradient. Remarkably, only HuR and Ago2 show a clear two peak-distribution corresponding to the peaks of monosome and polysome fractions as represented by the S6 distribution. This observation strengthens the notion that these proteins may control ribosomal translation. There is no significant MK2-dependent redistribution detected for the proteins analyzed. This is not unexpected, since TNF mRNA and other ARE-containing mRNAs are only a minor part of the total cellular mRNAs analyzed in this overall fractionation.
We also compared the distribution of selected proteins between the cytosolic and ER fractions in dependence on the presence of MK2. There are no qualitative differences in the overall polysome profiles of MK2-rescued and GFP-transduced cells or of the ER or cytosol fractions detected after 1 h LPS-stimulation as measured by absorbance at 254 nm (Figure 3C). However, ß-actin is mainly present in the cytosolic fraction, while the major portion of NOGO-B is detected in the ER fraction (Figure 3D). The ribosomal protein S6 is present in both fractions representing cytosolic and ER-directed translation. The nuclear histone protein H3 is detectable in the total extract but not in the cytosolic or ER fractions, indicating that the fractionation indeed excludes nuclei. The distribution of ß-actin, NOGO-B, S6 and H3 is independent of the presence of MK2. p38 MAPK, MK2, and another protein kinase downstream to p38, Msk1 [39], are detected exclusively in the cytosolic fraction. Only in MK2-rescued macrophages, NOGO-B displays the double band for the phosphorylated and non-phosphorylated isoforms in the ER-fraction, indicating that NOGO-B can serve as a substrate for MK2 although both proteins are detected in different fractions. Most interestingly, TTP is detectable in the cytosol and ER of GFP-transduced cells, but shows a clear exclusion from the ER fraction in MK2-rescued macrophages, although expressed to a higher total concentration in these cells. Such complete exclusion is not seen for the other mRNA-binding proteins tested (Ago2, KSRP, TIA-1, and HuR). The MK2-dependent exclusion of TTP from the ER fraction containing ER-bound polysomes, also loaded with TNF mRNA (cf. Figure 2B), indicates a possible inhibitory role of TTP in translational regulation, which can be neutralized by MK2. In accordance with its synthesis as a type II-membrane protein, pro-TNF is almost exclusively detected in the ER franction and its level is significantly reduced in the GFP-transduced cells (Figure 3D, cf. Figure 1C)
The MK2-dependent absence of TTP in the ER fraction, where TNF mRNA is actively translated, and the fact that TTP is a substrate for MK2 [11] at least indirectly suggest an MK2-dependent role of TTP in repression of TNF translation. This notion is also in agreement with the observation that genetic deletion of TTP increases TNF production and renders it MK2-independent [30]. To prove this hypothesis, we performed TTP knockdown (Figure 4A) and analyzed ER-directed pro-TNF translation. As seen in Figure 4B, TTP knockdown does not influence translation of ß-actin mRNA. Interestingly, TTP knockdown specifically stimulates a significant shift of TNF mRNA into ER-bound polysome fractions in LPS-treated GFP-transduced macrophages (Figure 4C). Since in LPS-treated MK2-rescued macrophages TNF mRNA is already detected in the polysomal fractions, the amount of polysomal TNF mRNA is only slightly increased further by the knockdown of TTP (Figure 4D). These results clearly identify TTP as a specific repressor of TNF translation, which acts downstream of MK2.
Since the distribution of HuR in density centrifugation parallels ribosomal distribution (Figure 3A) and since it is known that HuR also binds to the ARE of native TNF-mRNA [23], we analyzed the effect of knockdown of HuR on the translation in MK2-rescued macrophages. siHuR knockdown is efficient and does not significantly influence translation of ß-actin mRNA (Figure 4E, 4F). Remarkably, upon HuR knockdown TNF mRNA is excluded from polysomal ER-bound fractions (Figure 4G), reduced in monosomal cytosolic fractions and increased in the mRNP fractions (Figure 4H). This indicates that HuR is necessary for the translation and specifically involved in translational initiation of TNF mRNA.
We then asked whether the de-repression of translation seen upon TTP knockdown also depends on the presence of HuR. We performed double siRNA knockdown of TTP and HuR in LPS-treated GFP-transduced macrophages (Figure 4I). Knockdown is efficient for both proteins. As for the TTP knockdown alone, the TTP/HuR double knockdown does not influence translation of ß-actin mRNA (Figure 4J). In contrast, the polysome localization of TNF mRNA induced by TTP knockdown (cf. Figure 4C) is not observed in the TTP/HuR double knockdown (Figure 4K, 4L) indicating that HuR is necessary for both types of de-repression, whether caused by the reduction or phosphorylation of TTP. Hence, in contrast to TIA-1, which acts as constitutive repressor [40], HuR is a constitutive activator of TNF translation in these cells.
Similar to HuR, Ago2 also parallels ribosomal distribution (Figure 3A). Ago2 has been described to be essential for TTP- and miR16-dependent TNF mRNA degradation [41] and for translational regulation of TNF mRNA upon serum starvation [42], [43]. Furthermore, Ago2 is also a substrate for MK2 [44]. Therefore, we analyzed the role of Ago2 in MK2-dependent translational control of TNF mRNA by knockdown experiment. Although Ago2 itself is a central element of the mechanism of siRNA-mediated regulation of specific mRNAs, its knockdown is quite efficient in MK2-rescued and GFP-transfected cells (Figure S6). However, knockdown of Ago2 does not significantly alter TNF mRNA distribution in MK2-rescued or GFP-transduced macrophages (Figure S6, lower panels) indicating that Ago2 is not regulating LPS-induced translation of TNF mRNA under these conditions.
We asked whether the changes in the presence of TNF mRNA in the polysomal fractions seen for the down-regulation of the different components in MK2-rescued and GFP-transduced macrophages are also reflected by changes in pro-TNF biosynthesis of these cells. For this reason, pro-TNF levels before and 4 h after LPS stimulation were semi-quantitatively determined by Western blot analysis. For two independent experiments Western blot signals were quantified and normalized by the ß-actin signal (Figure 4M, 4N). Before LPS stimulation almost no pro-TNF signal was detectable (Figure 4M), indicating that the pro-TNF detected after LPS-treatment represents newly synthesized protein. After LPS-stimulation the amount of newly synthesized pro-TNF is significantly higher in MK2-rescued compared to the GFP-transduced cells as also seen in Figure 1C and Figure 3D. For MK2-rescued cells the knockdown of TTP, which caused a slight increase in the TNF mRNA polysomal fraction (Figure 4A), also lead to a further small increase in pro-TNF synthesis (Figure 4N). Knockdown of TTP in the GFP-transduced macrophages resulted in a strong increase in pro-TNF level corresponding to the strong shift of TNF mRNA to the polysomal fraction (cf. Figure 4C). Consistently, knockdown of HuR significantly inhibited pro-TNF synthesis in MK2-rescued cells reflecting the decrease in the polysomal amount of TNF mRNA detected under this condition (cf. Figure 4G). Most importantly, parallel knockdown of TTP and HuR resulted in a strongly decreased pro-TNF synthesis compared to the single TTP knockdown reflecting the differences in polysomal TNF mRNA seen in Figure 4L. Taken together, the strict correlation between the abundance of TNF mRNA in the polysomal fraction and the production of pro-TNF after LPS treatment confirms the notions that the TNF-mRNA detected in the polysomal fractions is translationally active and that translation of pro-TNF is regulated by TTP and HuR.
Since TTP and HuR are both able to bind to the ARE of TNF-mRNA [23], [29] and since HuR is necessary for the release of the translation block after phosphorylation or knockdown of TTP (Figure 4L), we were interested in whether both proteins compete in ARE-binding and whether this competition is controlled by phosphorylation of TTP by MK2/3. To analyze this scenario in vitro, we first expressed strep-tagged HuR and TTP in HEK293T cells and purified these proteins by binding to strep-tactin beads. We subjected different amounts of purified strep-HuR and -TTP to Western blotting and detected the proteins using enzyme-coupled strep-tactin. While strep-HuR is detected in a single band of approximately 40 kDa, strep-TTP is seen in a more diffuse group of bands around 50 kDa, which probably arises from multiple PTMs [45] introduced into TTP also in the HEK 293T cells (Figure S7A). To prove ARE-binding activity of the purified proteins, we performed EMSA using a TNF mRNA ARE-derived RNA probe containing six overlapping AUUUA pentamers labeled with the infrared dye DY-681, which can be detected by imaging (Figure 5A). In EMSA, HuR displays the typical two band pattern [23] while TTP retards the ARE probe in one band, as already known [46]. Specific antibodies against HuR and TTP lead to supershifts respectively demonstrating the identity and specificity of the complexes (Figure 5A). In an in vitro kinase assay, TTP but not HuR is further phosphorylated by MK2 in the presence of its activator p38 (Figure 5B). The observation that strep-HuR is not phosphorylated by p38 in this experiment is in contrast to the published finding that GST-HuR is phosphorylated by p38 in vitro [47] and indicates that the same site is already phosphorylated by Chk2 during overexpression in the HEK293 cells as already described for HeLa cells [48].
We then incubated the purified proteins and mixtures thereof with the DY-681-labeled ARE mRNA, stabilized protein-RNA-complexes by UV cross-linking and visualized ARE-probe bound to protein after SDS-PAGE by infra-red imaging (Figure 5C). Although cross-linking of the ARE-probe to HuR seems more efficient, binding of the probe to both proteins can be detected. In the mixture both proteins show competitive binding to the probe with similar binding in comparable concentrations and a significant displacement of the competitor at about threefold molar excess of the other protein (Figure 5C). Competitive binding of TTP and HuR to the ARE was then analyzed in the presence of active MK2 (Figure 5D). Remarkably, p38/MK2 activity strongly and robustly shifts the binding equilibrium towards HuR in this assay. Even in the presence of an about four-fold molar excess of TTP (100 ng TTP vs. 25 ng HuR), almost exclusive binding of HuR to the ARE is detected in the presence of p38 and MK2. We also tested whether MK2 or p38 alone are able to shift the binding towards HuR (Figure S7B). It became clear that only MK2 together with its activator p38 is able to initiate the TTP-HuR-exchange at the ARE-probe, indicating the necessity of catalytically active MK2 in the assay. We then analyzed the ARE binding affinity of HuR and TTP, respectively, phosphorylated by p38/MK2 activity by EMSA and cross-linking experiments using CIP-dephosphorylated protein as control. By titrating increasing concentrations of these proteins (HuR: 2.5×10−7 M–2×10−6 M) with the same concentration of the ARE probe (7.5×10−14 M) approximate Kd values were determined as described in [49]. In line with being no substrate for MK2 (Figure 5B), no significant change in binding affinity of HuR to ARE (kd 5.0–7.5×10−7) could be observed in the presence of catalytically active MK2 (Figure 5E). In contrast, catalytic activity of p38/MK2 lead to a clear reduction of the affinity of TTP to ARE represented by a shift in the approximate Kd from 5×10−7 for the non-phosphorylated protein to 6–8×10−6 in the presence of active MK2 (Figure 5F). At a fixed TTP concentration of 4×10−6 M significant changes in ARE-binding depending on p38/MK2 phosphorylation could be demonstrated by EMSA (Figure 5G). The reduction in affinity was dependent on phosphorylation of TTP by MK2, since the MK2 phosphorylation site mutant TTP-AA bound with the same affinity to the ARE like the CIP-treated TTP (not shown) and did not show altered affinity in the presence of p38/MK2 (Figure S7D). Furthermore, the binding equilibrium between HuR and the TTP single or double mutants was not changed in the presence of p38/MK2 (Figure S7E). The fact that phosphorylation by MK2 weakens the affinity of TTP for ARE while HuR affinity remains nearly unchanged qualitatively explains that the competitive binding equilibrium between TTP and HuR to the ARE is shifted towards HuR in the presence of MK2.
To examine whether the MK2-dependent shift of the ARE-binding equilibrium between TTP and HuR is relevant in vivo in LPS-stimulated macrophages, we monitored the binding of endogenous HuR to TNF mRNA by immunoprecipitation of endogenous HuR (RNA-IP). MK2-rescued and GFP-transduced macrophages were stimulated for 1 h with LPS and the IP was carried out using HuR- and, as negative control, GFP-antibodies. TNF mRNA in the IPs was quantified by qRT-PCR. Compared to the GFP-IP, specific accumulation of TNF mRNA in the HuR-IP is detected (Figure 5H). Since it is known that HuR constitutively binds to an U-rich element in the 3′UTR of ß-actin mRNA [50] and since ß-actin mRNA is not induced by LPS, we quantified ß-actin mRNA in the HuR-IP to monitor the efficacy and further specificity of the IP (Figure 5I). When normalizing the HuR-IP by ß-actin mRNA, a significantly increased TNF mRNA/ß-actin mRNA ratio is observed in LPS-treated MK2-rescued macrophages compared to the GFP-transduced cells (Figure 5J). This finding suggests that an MK2-dependent shift of the binding equilibrium between HuR and TTP at the ARE of TNF mRNA might also occur in vivo.
We could not obtain complementary data for TTP binding to TNF mRNA, since the efficacy of TTP-IP using the available antibodies is low. In addition and more importantly, the strong difference in the expression level of TTP between MK2-rescued and GFP-transfected cells (Figure 1A, Figure 3D) leads to large variations in IP-efficiency which make a comparison between these cell lines impossible.
Since the mechanism elucidated regulates initiation of translation before docking to the ER, we were interested in whether synthesis of cytoplasmic proteins is regulated in the same manner. TTP itself is a cytoplasmic protein and it is known that TTP also binds to its own mRNA, which contains a 3′ UTR ARE of three clustered AUUUA pentamers [51]. Therefore, we analyzed translation of TTP mRNA in the cytosolic fraction for its dependence on MK2 catalytic activity and on the presence of the TTP antagonist HuR (Figure 6). The content of TTP mRNA in polysomal fractions is significantly increased in MK2-rescued macrophages when compared to GFP-transduced cells (Figure 6A). The p38 inhibitor blocks this increase and the catalytically dead mutant of MK2 is not able to increase TTP translation (Figure 6B). Both treatments increase the concentration of TTP mRNA in monosomes. These findings indicate that catalytic activity of MK2 also stimulates translation of TTP mRNA in the cytosol. Interestingly, knockdown of HuR strongly inhibits MK2-mediated translational stimulation of TTP mRNA and leads to the accumulation of TTP mRNA in mRNPs (Figure 6C). This inhibition is accompanied by a decrease in TTP biosynthesis as detected by Western blot 4 h after LPS-stimulation (Figure 6D). These findings strongly support the notion that translation of TTP mRNA is regulated by a mechanism similar to that described for TNF mRNA.
We provide evidence that translational control of TNF-mRNA in macrophages might be achieved by a phosphorylation-regulated competitive mRNA association of the two ABPs TTP and HuR. In this scenario TTP acts as a translational repressor while HuR is a translational activator. By LPS-induced MK2-mediated phosphorylation of TTP the competitive binding equilibrium between TTP and HuR is shifted towards HuR and leads to a stimulation of translation (Figure 7). This mechanism is not restricted to the translation of the type II membrane protein pro-TNF, but is also valid for the translation of the cytosolic protein TTP itself, indicating that this regulation by ABP exchange occurs in an early common step of both processes, probably during translational initiation. This notion is strengthened by the observation that knockdown of HuR leads to an accumulation of cytosolic TNF- and TTP-mRNPs (Figure 4H and Figure 6) and to a reduction of TNF mRNA-containing monosomes (Figure 4H).
In regard to the possible molecular mechanisms of translational initiation (reviewed in [52]), the binding of TTP to the ARE in the 3′ region of the mRNAs probably interferes with eIF4G-dependent mRNA circulation or with the binding of the 43S pre-initiation complex (PIC) to the mRNA (Figure 7). Since no significant accumulation of mRNAs in the 43S-fraction is observed, inhibition of 43S PIC scanning or of joining of the 60S ribosomal subunit seems unlikely. Furthermore, it is not clear whether TTP directly interacts with components of the general translation machinery or whether it recruits further proteins which interfere with translational initiation. From the analysis of TTP's mRNA destabilizing function it is known that TTP can act as a binding platform for various proteins involved in mRNA decay [53]. Furthermore, an mRNA-dependent interaction between TTP and poly(A)-binding proteins (PABPs) has been detected [19], [53], [54] and, interestingly, a direct interaction of TTP and PABP C1 was recently demonstrated in a Y2H screen [55]. Hence, one may speculate that TTP-PABP-interaction could interfere with PABP-eIF4G-interaction and could prevent PABP-eIF4G-eIF4E-mediated circularization of the mRNA as a prerequisite for translation (Figure 7). Taking into account that PABP-1 is also a direct substrate of MK2 [14], its phosphorylation by MK2 could further contribute to the weakening of the TTP-PABP-interaction.
The molecular mechanism of the phosphorylation-driven change in the competitive binding equilibrium of HuR and TTP at the ARE is probably based on the fact that the affinity of TTP is reduced after phosphorylation. However, one should also take into account that cytoplasmic concentrations of HuR and, especially, of the immediate early gene TTP [31] are highly flexible and contribute to the binding equilibrium. A possible scenario of LPS-stimulated regulation could be that newly synthesized TTP is prevented from ARE binding by early phosphorylation via MK2/3 allowing efficient translation of the target mRNAs by the help of HuR. Subsequent decrease in MK2-activity, which peaks after a mere 30 min period, and dephosphorylation of TTP by protein phosphatase 2A [56] could then lead to increased binding of TTP to the ARE, resulting in feedback regulation by translational arrest and destabilization of the target mRNA. Upon LPS-stimulation of macrophages, RNA-IP experiments indicate an early MK2-dependent shift of the binding equilibrium towards HuR supporting the idea that the regulatory mechanism proposed might be relevant in vivo.
Recently, a translational repression by TTP has been demonstrated using reporter constructs carrying an ARE in the 3′ UTR in transfected 293T HEK cells [57]. In this system, TTP cooperates with the general translational repressor RCK/P54, which belongs to the DEAD-box helicase family and displays ATP-dependent RNA-unwinding activity [58]. Hence, it cannot be excluded that RCK/p54 also contributes to TTP-dependent translational repression of TNF mRNA in macrophages. The ubiquitin E3 ligase cullin 4B is a recently identified TTP interacting protein which slightly changes polysome loading of TNF mRNA [59]. However, it is also possible that cullin 4B is involved in ubiquitination and subsequent degradation or functional modification of TTP, as already known for its MEKK1-induced TRAF-2-mediated K63 ubiquitinylation [60].
It has also been described that TTP-facilitated binding of miRNAs, such as miR16 or miR369-3, are involved in regulation of the decay of ARE-containing mRNAs [41] and stimulation of translation of TNF mRNA upon serum starvation [42], respectively. A recent screen for miRNAs, which target the 3′ UTR of TNF mRNA, revealed miR-125b and miR-939 as further candidates. However, repression of these miRNAs did not influence TNF mRNA expression, making it unlikely that these miRNAs regulate translation of TNF mRNA [61]. Furthermore, knockdown of Ago2, a target of MK2 and an essential component of the miRNA system, does not influence LPS-induced TNF translation in macrophages (Figure S3). This makes it rather unikely that miRNAs regulate MK2-dependent, LPS-induced translation of TNF mRNA in macrophages.
It is known that TTP destabilizes ARE-containing cytokine mRNAs of IL-1ß, IL-2, IL-3, IL-6, IL-10, TNF, GM-CSF and that, in many cases, this destabilization is relieved by the p38 pathway (reviewed in [16]). Even for the global LPS-stimulated regulation of mRNA decay by TTP it has been convincingly shown that the activity of the p38 pathway inversely correlates with TTP's mRNA degrading activity which arises about 3–9 h after LPS-treatment when p38/MK2/3 activity is back to basal levels [62]. The p38/MK2/3-dependent replacement of TTP by HuR at ARE-containing mRNAs would provide a simple and elegant explanation for this regulation: The binding of TTP to a specific mRNA is the prerequisite for its constitutive degradation. When TTP is displaced from the ARE by HuR (or other ABPs), the specific mRNA is no longer targeted to the constitutive degradation pathway and stabilized. It is interesting to note that the stability of KC mRNA is also regulated by TTP [63], even in a MK2-dependent manner [32] (Figure S8A, S8B), while KC mRNA is not translationally regulated by MK2 (Figure 2D) and we do not observe disappearance of the majority of KC mRNA from the polysome fraction upon HuR knockdown (Figure S8C). KC mRNA contains three isolated AUUUA motifs and two doubles of AUUUAUUUA, which are sufficient for TTP binding and regulation of stability [64]. Interestingly, the isolated AUUUA motifs of KC mRNA are not sufficient for HuR binding. Furthermore, although at least its AUUUAUUUA stretch fits rather well to the HuR consensus UUUUUUU, there is no HuR binding reported for KC mRNA in the transcriptome wide screens [65], [66]. Hence, KC is an example of separation of TTP function from HuR action, since HuR is not able to compete with TTP binding to KC mRNA and is not necessary for its translation.
TTP has been postulated as the essential component of the feedback regulation of the LPS-stimulated TNF-response [29]. Increased TTP phosphorylation by MK2, which neutralizes TTP repressor function [15], [30], is paralleled by transcriptional activation of the TTP gene, which belongs to the group of immediate early genes (Figure 1A and [67]), as a result of phosphorylation of the transcription factor SRF by MK2 [31]. Interestingly, the feedback regulation by TTP also comprises TTP's binding to an ARE in the 3′UTR of its own mRNA [51]. Here, we have demonstrated that translation of TTP mRNA is also stimulated by MK2, probably by the same mechanism of ARE-replacement of phospho-TTP by HuR at the ARE of TTP mRNA, since knockdown of HuR inhibits translation and protein expression of TTP (Figure 6). Hence, MK2 not only stimulates transcription but also translation of TTP and rapidly enables the re-synthesis of non-phospho-TTP, which again limits TNF expression and TTP expression itself. The parallel translational stimulation of TNF and TTP by the same MK2-dependent mechanism for the first time explains the paradox observation that both, TNF biosynthesis and TTP expression are strongly reduced in MK2-deficient macrophages [32], whereas complete deletion of TTP leads to increased TNF biosynthesis [29]. In addition, since both TNF and TTP are further reduced in MK2/3 double-deficient macrophages when compared to MK2-deficient cells, it is highly likely that MK3 shares these functions of MK2.
The fact that ABPs interact with their cognate mRNAs and regulate their own expression has been comprehensively described [68]. Furthermore, a complex network of post-transcriptional cross-regulation of expression between ABPs such as HuR, TIA-1, KSRP, AUF1 is known to exist [68]. However, in the experimental system of immortalized macrophages applied in this study, we could not detect significant changes in expression of other ABPs (TIA-1, KSRP) as a result of knockdown of TTP and HuR. Furthermore, the binary in vitro-binding system using purified HuR and TTP proteins, which is sufficient to observe the MK2-dependent exchange between TTP by HuR, suggests that the regulatory mechanism postulated could function without the involvement of other ABPs apart from TTP and HuR. On the other hand, specific binding of isoforms of the mRNA-binding protein AUF1 to TTP has been described [69]. Interestingly, this interaction increases RNA-binding affinity of TTP in an in vitro assay about five-fold. Hence, it cannot be excluded that further ABPs modulate the proposed regulation in vivo.
HuR is predominantly held responsible for constitutive stabilization of ARE-containing mRNAs [20] and demonstrably binds to the ARE of TNF mRNA [22], [23]. Besides the TNF mRNA stabilizing function, HuR has also been described to influence translation. In macrophages, HuR acts as a homeostatic coordinator of expression of ARE-containing mRNAs [26], which, when deleted, also increases inflammatory cytokine production of macrophages. This effect could be explained by the fact that HuR is essential for efficient expression of TTP (Figure 6) necessary for down-regulation of the inflammatory response. On the other hand, LPS-stimulated TNF production is significantly inhibited in the NZW mouse strain containing two different 3-base insertions in the TNF-ARE, which inhibit binding of HuR to the ARE [70]. This finding indicates a positive role of the specific binding of HuR to the TNF-ARE for TNF expression and is in agreement with our observations that expression of HuR is necessary for efficient translation of TNF. It is interesting to note that stress-induced release of miR-122-mediated translational repression of CAT-1 mRNA also requires HuR [71] indicating a more general role of HuR in counteracting translational repression, not only by ABPs but also by miRNAs.
HuR is mainly localized in the nucleus but continuously shuttles between cytoplasm and nucleus [72]. There are various post-translational modifications of HuR described, which result in changes in its subcellular distribution [73]. Interestingly, p38 phosphorylates HuR at T118 in a stress-dependent manner resulting in significant cytoplasmic accumulation of HuR [47]. Expression of constitutively active MK2 also leads to cytoplasmic accumulation of endogenous HuR [74]. This p38/MK2/3-stimulated increase in the cytosolic HuR concentration could be a prerequisite for translational stimulation of ARE-containing mRNAs by HuR (cf. Figure 7). Hence, the p38/MK2 pathway may regulate specific translational initiation by catalytic activity of both, p38 and MK2, in parallel at the levels of translocation and regulation of activity of ABPs.
Primary MK2/3 DKO BMDMs were immortalized as previously described [75]. Retroviral transduction of MK2/3 immortalized BMDMs with vectors encoding MK2 wild type, kinase dead MK2 (K76R) and the empty vector (GFP) was carried out as described [31], [75].
Immortalized and retroviral transduced MK2/3 DKO BMDMs were grown in DMEM containing 10% FCS, 2 mM L-glutamine, 100 U penicillin G/ml, 100 mg streptomycin/ml and 0,1 mM non essential amino acids mixture (Life Technologies/Invitrogen) under humidified conditions with 5% CO2 at 37°C. HEK293T cells were grown under the same conditions and were transfected with the calciumphosphate method. The p38 inhibitor SB202190 (Sigma) and LPS (Escherichia coli 0127:B8, Sigma) were used at concentrations of 5 µM and 1 µg/ml, respectively. Primary BMDM were derived as previously described by using 10 ng/ml M-CSF (Wyeth/Pfizer) [30].
siRNA-mediated knockdown in BMDMs was performed following the instructions of the protocol for knockdown in RAW264.7 cells using HiPerFect transfection reagent (Qiagen). 8×104 cells were seeded in 100 µl growth medium in a 24-well plate. Appoximately 187.5 ng siRNA was mixed with 3 µl HiPerFect and 100 µl Opti-MEM (Life Technologies/Invitrogen) and incubated for 5 minutes at room temperature. The mixture was added dropwise to the wells and after 6 hrs of incubation 400 µl complete growth medium was added. The next day the medium was changed to complete medium. The highest efficiency for the different knockdown experiments was achieved after 48 hrs of siRNA treatment. For siRNA transfections of 10 cm cell culture plates the protocol was upscaled according to the instructions of the HiPerFect handbook (Qiagen). The following mouse specific siRNAs (target sequences) were used (Qiagen): siTTP: 5′-CCTGAGAATCCTGGTGCTCAA-3′ (Mm_Zfp36_6), siHuR: 5′-CAGAAACATTTGAGCATTGTA-3′ (Mm_Elavl1_4), siAgo2: 5′-CACTATGAATTGGACATCAAA-3′. For control knockdown Allstars negative Control siRNA (Qiagen 1027281) was used.
Western blotting was performed as described [31]. Blots were developed with an ECL detection kit (Santa Cruz Biotechnology) and digital chemoluminescence images were taken by a Luminescent Image Analyzer LAS-3000 (Fujifilm). Primary antibodies used were: anti-eEF2 (2332), anti-Histone3 (9715), anti-MK2 (3042), anti-pMK2pT222 (9A7) (3316), anti-p38 (9212), anti-pp38pT180/pY182 (9211), anti-pPKD (4381B) and anti-S6 (5G10) 2217 (all from Cell Signaling), anti-Ago2 (M01) (Abnova), anti-GAPDH (6C5) Mab374 (Millipore/Chemicon), anti-ß-Actin (C4) sc-47778, anti-GFP (B-2) sc-9996, anti-HuR (19F12) sc-56709, anti-Msk1 (H-19) sc-9392, anti-TIA-1 (C-20) sc-1751, anti-TNF (L19) sc-1351 (all from Santa Cruz Biotechnologies). Antibodies against KSRP [76], TTP (SAK21B) [77], TTP-pS178 [56] and NOGO-B [35] were described previously and were kindly provided by Drs. A. R. Clark (London), G. Stoecklin (Heidelberg) and P. Cohen (Dundee), respectively. Streptactin-HRP conjugate (IBA BioTAGnology) and secondary HRP-conjugated antibodies (Santa Cruz Biotechnologies) were used. The mouse TNF-alpha Ready-SET-Go! kit from eBiosciences (88–7324) was used for TNF ELISA. Quantification of TNF and ß-actin Western blot signals was performed with the Multi Gauge 3.2 software (Fujifilm). For each individual experiment two different exposure times were analyzed and then normalized to the ß-actin signals. Two independent Western blot experiments with different sample loading were carried out. For inter-experimental comparison the signals were normalized by the intensity of non-stimulated MK2-rescued cells transfected with control siRNA.
BMDMs of a 6 cm plate of 80% confluence were dissolved in 150 µl extraction buffer (20 mM Tris pH 8.0, 140 mM KCl, 5 mM MgCl2, 0.5 mM DTT, 0.1 mg/ml cycloheximide and 0.5 mg/ml Heparin). Then 1% (w/v) Saponin (ICN chemicals) dissolved in DMSO was added to a final concentration of 0.1% (v/v) followed by 20 minutes incubation on ice with partial vortexing. Cells were then centrifuged 5 minutes at 500×g. The supernatant (cytsosol) was kept separately on ice. The pellet (microsomes and nuclei) was washed with extraction buffer and was centrifuged again (500×g). The pellet was dissolved in extraction buffer containing 0.5% v/v NP-40 (Fluka). Both cytosol and dissolved microsomal fraction were then centrifuged 10 minutes at 7500×g. The resulting supernatants represent the cytosolic and the microsomal ER fraction free of nuclear components.
1×107 BMDMs were used for a single gradient experiment. The cells were washed twice with 1×PBS containing 0.1 mg/ml cycloheximide and differentially lysed as described above. For RNaseA treatment controls, 1.5 mg/ml RNaseA (Carl Roth) was added to the lysates for 15 min on ice. The cytosolic and ER ribosome extracts (400 µl each) were loaded on linear sucrose gradients (12 ml) ranging from 50% (bottom) to 10% (top) sucrose containing 140 mM KCl, 20 mM Tris pH 8, 5 mM MgCl2, 0.1 mg/ml cycloheximide, 0.5 mg/ml Heparin and 0.5 mM dithiothreitol (DTT). After loading, the extracts were separated by ultracentrifugation in a SW40.1 Ti Rotor (Beckmann-Coulter) for 2 hrs at 35000 rpm and 4°C. Subsequently, 12 gradient fractions (each 1 ml) were collected using a UA-6 UV/VIS device (Teledyne/ISCO Inc.) that was connected to an optical unit allowing OD documentation. RNA isolation from the gradient fractions was performed by adding 1/10 volume of 3 M Na-acetate (Sigma) and 1 volume of isopropanol and overnight precipitation at −20°C. RNA was pelleted by centrifugation at 13000×g for 20 min at 4°C. For the analysis of co-sedimenting proteins, trichloro acetic acid (TCA) was added to each fraction (final 10% v/v). Proteins were precipitated overnight at 4°C and subsequently centrifuged for 20 min at 13000×g at 4°C. Pellets were washed twice with ice cold acetone, dissolved in 100 µl 2×SDS-loading buffer and heated for 5 minutes at 95°C.
RNAs were isolated by resolving the cells or pellets obtained from polysome gradient fractions in lysis buffer RA1 of the RNA NucleoSpin II kit and subsequent processing (Macherey+Nagel). For cDNA synthesis, the cDNA first strand cDNA synthesis kit (Thermo/Fermentas) was used. cDNAs were diluted 1∶20 for detection in qRT-PCR reactions. For detection of TTP and KC cDNA predesigned and FAM-labelled TaqMan primer mixtures from Applied Biosystems were used (Mm00457144_m1 Zfp36, Mm00433859_m1 Cxcl1). TNF cDNA was amplified using a labelled probe (5′-FAM – CAC GTC GTA GCA AAC CAC CAA GTG GA – BHQ1-3′) together with flanking primers (forward: 5′-CAT CTT CTC AAA ATT CGA GTG ACA A-3′ and reverse: 5′-TGG GAG TAG ACA AGG TAC AAC CC-3′). ß-actin cDNA was amplified with the VIC-labelled predesigned probe from Applied Biosystems (4352341E) allowing two channel detection of one cDNA. Amplifications were carried out in a 1× SensiFAST Probe No-ROX buffer system (Bioline) using a Rotor-Gene-Q device (Qiagen). The threshold cycle (CT) of each individual PCR product was calculated by the software of the instrument.
15 cm diameter plates of HEK293T cells were transfected with the expression constructs (pcDNA3-His-Strep-TTP, -TTP-S52A, -TTP-S178A, -TTP-AA (S52,178A) and pEXPR-IBA105-HuR) and lysed 24 hrs post transfection. Strep-tagged HuR and TTP protein was purified by affinity chromatography using streptactin beads (IBA TAGnologies) as described previously [76]. Proteins were eluted with desthiobiotin (IBA TAGnologies) and the protein concentration was determined by Coomassie staining of the bands in SDS-PAGE compared to BSA standards using the Multi Gauge quantification software 3.2 (Fujifilm). Dephosphorylation of purified proteins was achieved by incubation with calf intestinal phosphatase (CIP) for 15 minutes at 37°C.
Strep-tag purified proteins were dissolved in EMSA-shift buffer (20 mM HEPES pH 7.6, 3 mM MgCl2, 40 mM KCl, 5% Glycerol, 2 mM DTT, 4 µg tRNA) to give a total volume of 20 µl and were incubated with 75 fmol of an 5′-DY681-labbeled AU-rich RNA-probe (5′-AUU UAU UUA UUU AUU UAU UUA UUU A-3′) for 25 minutes at 4°C. For supershift experiments 0.2 µg of specific antibody was added to the mixture after 10 min of preincubation at 4°C. The reaction mix was then loaded onto a 4% native shift gel after a pre-run of 30 minutes at 80 V and 4°C and separated at 80 V for 90 minutes in 0.25× TBE buffer at 4°C. The detection of RNA-protein complexes was performed by visualization of DY681 on a LiCOR Odyssey infrared-scanner.
As competition assay, purified HuR protein (25–300 ng) was first incubated together with 75 fmol of the 5′-DY681-labelled AU-rich RNA-probe (see above) in a 20 µl reaction mix in 96-well plates. Where indicated, 300 ng of the purified recombinant protein kinases His6-MK2 (in 1.0 µl) and GST-p38 (in 0.3 µl) (Menon et al. 2010) and 0.5 µl 10 mM ATP were added. After 10 min at 30°C, purified TTP protein was added and incubated for another 15 min minutes at 30°C. Then, the RNA-protein crosslink was performed by UV auto cross-linking using a Stratalinker (Stratagene). The cross-linking products were separated by SDS-PAGE and detected using the LiCOR Odyssey infrared scanner.
For detection of phosphorylation, strep-tag purified proteins were mixed together with E.coli-expressed and purified His6-MK2, GST-p38, GST protein and radiolabelled gamma-33P-ATP as described previously (Menon et al. 2010). Purified recombinant Hsp25 served as a positive control for MK2 kinase activity. Samples were resolved by SDS-PAGE and analyzed by phospho-imaging on a FLA-5000 (Fujifilm) system.
MK2-rescued and GFP-transduced macrophage lines were stimulated with LPS for 1 h, UV treated for 30 seconds (120 mJ/m2) and subsequently lysed in buffer containing 30 mM HEPES (pH 7.4), 150 mM NaCl and 0.5% v/v NP-40 with protease inhibitors. For RNA-IP, 1 µg of anti-HuR (mouse IgG1) and, as negative control, 0.5 µg of anti-GFP (mouse IgG2a) were incubated overnight at 4°C with the same amounts of cross-linked lysates in a volume of 0.5 ml. Afterwards 30 µl of Protein G Sepharose (GE Healthcare) suspension blocked with 50 µg/ml t-RNA for 1 h were added. After further incubating 2 h at 4°C, the beads were washed extensively in lysis buffer containing 0.25% v/v NP-40. The associating RNAs were eluted by vigorous vortexing of the beads in 350 µl RA1 lysis buffer (RNA NucleoSpin II kit (Macherey+Nagel)). Samples of the total lysates (input) and the re-dissolved precipitates were then analyzed by qRT-PCR.
|
10.1371/journal.pcbi.1006118 | Effects of growth rate, cell size, motion, and elemental stoichiometry on nutrient transport kinetics | Nutrient acquisition is a critical determinant for the competitive advantage for auto- and osmohetero- trophs alike. Nutrient limited growth is commonly described on a whole cell basis through reference to a maximum growth rate (Gmax) and a half-saturation constant (KG). This empirical application of a Michaelis-Menten like description ignores the multiple underlying feedbacks between physiology contributing to growth, cell size, elemental stoichiometry and cell motion. Here we explore these relationships with reference to the kinetics of the nutrient transporter protein, the transporter rate density at the cell surface (TRD; potential transport rate per unit plasma-membrane area), and diffusion gradients. While the half saturation value for the limiting nutrient increases rapidly with cell size, significant mitigation is afforded by cell motion (swimming or sedimentation), and by decreasing the cellular carbon density. There is thus potential for high vacuolation and high sedimentation rates in diatoms to significantly decrease KG and increase species competitive advantage. Our results also suggest that Gmax for larger non-diatom protists may be constrained by rates of nutrient transport. For a given carbon density, cell size and TRD, the value of Gmax/KG remains constant. This implies that species or strains with a lower Gmax might coincidentally have a competitive advantage under nutrient limited conditions as they also express lower values of KG. The ability of cells to modulate the TRD according to their nutritional status, and hence change the instantaneous maximum transport rate, has a very marked effect upon transport and growth kinetics. Analyses and dynamic models that do not consider such modulation will inevitably fail to properly reflect competitive advantage in nutrient acquisition. This has important implications for the accurate representation and predictive capabilities of model applications, in particular in a changing environment.
| Relating environmental nutrient concentration and nutrient acquisition to cell growth is an important feature of numerical simulations describing ecological systems of microbes. Here we investigate the critical role of the combined effects of maximum growth rate, cell size, motion, and elemental stoichiometry on nutrient transport kinetics and thence growth kinetics. By applying mechanistic scaling of nutrient uptake our results identify fundamental shortcomings in the interpretation of empirically derived relationships used to describe nutrient uptake in microbes. While the amount of nutrient required to grow at a given rate under nutrient limited conditions increases rapidly with cell size, the maximum growth rate scales directly with the environmental nutrient concentration. Requiring less nutrient at lower maximum growth rates, cells can therefore remain healthier at lower resource abundance. Further, decreased carbon content per cell lowers demand for nutrient transport per surface area significantly. This allows larger phytoplankton cells, like diatoms, to significantly increase their competitive advantage with increasing sedimentation rates. These findings have important implications for numerical models both in a context of theoretical ecology and applied science. Our results highlight the importance of accounting for organism physiology and related feedbacks in ecological applications and climate change studies.
| The relationship between nutrient uptake kinetics and growth rate is seen as a critical determinate in competition for organisms reliant on the transport of dissolved nutrients, and often plays a key role in structuring marine ecosystem models [1–3]. Here we consider interactions between cell size and cellular carbon density (as linked to vacuolation, for example), elemental stoichiometry, motion through the water, and growth rate potential with nutrient transport. While facets of such interactions have been considered before [3–5] we present a new analysis that explores how traits at the level of nutrient transport work through to better explain how nutrient availability controls organism growth and competitive advantage.
The physiology underpinning these relationships is complex and there is scope for significant confusion in interpreting experiment design and data. Most obviously there is the difference between the short term relationship between nutrient (substrate) concentration at the cell surface (S0) and nutrient transport rate into the organism, and the longer term relationship between S0 and organism growth rate. This difference develops because nutrient transport is controlled by various feedback processes that develop during post-transport assimilation of the nutrient, and are thus related to the organisms’ physiological history and thence to its growth rate. These factors also affect the difference between S0 and the substrate concentration in the bulk water (S∞); it is the latter which is determined in chemical analyses of water and features as a variable in models, while the former is the concentration of importance for the organism itself.
Flynn (1998) [6] differentiated between transport and growth kinetics, noting that experiment design (especially with respect to the period of incubation and the type of nutrient) and the prior physiological history of the organism govern whether measured “uptake kinetics” are more in keeping with true transport kinetics or with growth kinetics [7,6]. To measure transport kinetics requires very short incubations (durations of seconds) or extrapolation of time course incubations [6]. However, for practical reasons experiments are typically run over times from a few minutes up to several hours which is sufficient for the development of some level of satiation feedback that moderates the transport process. That incubation period is also usually insufficient to allow nutrient flow through to growth to approach steady-state. In consequence, interpreting reports in the literature concerning nutrient uptake kinetics conducted on different organisms, using different experimental protocols, is fraught with difficulties.
It is often assumed that for transport, uptake and growth kinetics the relationship with the substrate may be described using a rectangular hyperbolic type 2 (RHt2) function. RHt2 describes the process rate (V) as limited to a maximum rate, (Vmax) and with a half saturation constant (K) of the substrate concentration (S).
V=VmaxSS+K
(1)
With K usually written as KM, Eq 1 describes the Michaelis-Menten equation for enzyme kinetics. An analogous equation is used to describe Monod growth kinetics. To enable us to differentiate between transport, growth and uptake kinetics, we use terminologies analogous to those of Flynn (1998) [6]. Thus, with reference to the form of Eq 1, we differentiate between pairs of constants for maximum rate and K, respectively, controlling transport (Tmax & KT), uptake (Umax & KU,) or growth (Gmax & KG). Table 1 gives a description of all abbreviations used in this work.
In reality, as we shall see, the RHt2 curve may not always be appropriate for the task at hand. However, the reciprocal value of K, as the value of S0 at which V = Vmax/2 = V0.5, nonetheless provides an index for the relative affinity of the kinetics for a given value of Vmax. Ultimately, if all else is equal, an organism which requires a lower substrate concentration to support a growth rate (G) at half that of its maximum (i.e., G = Gmax/2 = G0.5) and thus expresses a lower KG, will be at an advantage over an organism with a higher KG. While in models KG is usually set as an input constant, the real value is an emergent function of nutrient transport and whole organism physiology. For example, KG for iron-limited phytoplankton growth depends greatly on whether nitrate or ammonium is used as the N-source, and also on the incident irradiance under which the phytoplankton grow [8]. To make the linkage between transport and growth kinetics thus requires an appreciation of the underlying physiology.
Nutrient transport (e.g., of NO3-, NH4+, PO43-) typically occurs via secondary active porters that are either matched for a specific nutrient molecule type, or for similar types [12]; thus a transporter for NO3- will not transport NH4+, while similar amino acids such as the cationic group arginine, lysine, histidine and ornithine may share the same transporter [13]. In addition, individual nutrient types may be taken up by several different transporter proteins [14–16], some of which may support biphasic kinetics [16–18]. Here, to simplify discussions, we will consider transport via a single (monophasic) transporter type.
While transporter proteins are not strictly enzymes (as they typically do not change the chemical form of their substrate), they express an affinity for the nutrients they transport; by analogy with the Michaelis-Menten half saturation value of enzymes, KM, we term this substrate concentration KT. The constant KM is a function of the affinity of the enzyme for the substrate in classic Michaelis-Menten terminology and is determined assuming that all factors other than substrate availability are non-limiting. Determining KT is more complex because transporter functionality depends on the integrity of the membrane in which the transporter proteins function, ionic gradients generated by primary active transporters required to support the operation of the typically secondary-active nutrient-transporters, as well as on the aforementioned absence or presence of short and longer term feedback processes modulating transport itself into the functional cell.
Another defining criterion for enzyme functionality is the maximum level of activity, kcat, which is described in units of mole of substrate consumed (or product given) per mole of enzyme per unit of time (Table 1). The maximum rate of enzyme activity in a given sample of biological material, which is a product of kcat and the concentration of enzyme protein, sets the value of the maximum process rate, Vmax, in Michaelis-Menten kinetics. It is important to note that the amount of enzyme in an assay does not affect the value of KM, while the value of Vmax in the assay is linearly related to enzyme concentration. The value of Vmax can thus be seen as being somewhat ambiguous, only being useful for a specific assay incubation. For considerations of whole-organism physiology, the value of kcat needs to be placed in the context of the total demand for its activity, the size (mass) of the enzyme and thence for the total resource expenditure for that enzyme within a given cell (e.g., for such calculations applied to the enzyme fixing CO2, RuBisCO [19]).
The maximum rate of activity in a given cellular system (Tmax) is analogous to Vmax in an enzyme assay. Accordingly, while the value of KT is independent of the number of transporter proteins in the cell, the value of Tmax is indeed dependent on that number. The extent to which Tmax exceeds Gmax, noting that transporter activity is modulated by post-transport physiology, helps to explain why KG is lower than KT, as illustrated in S1 Fig and the adjoining online text. In reality there are many hundreds if not thousands of transporter proteins in operation across the plasma-membrane of an individual cell. Theoretical estimates of relative nitrate and phosphate transporter density suggest that a specific transporter type will generally cover less than 0.1% of the cell surface under nutrient limited conditions [20]. The number of transporter proteins, and hence the maximum rate of nutrient transport (Tmax), also varies greatly with the nutritional status of the cell and for different nutrients, with ammonium transport and assimilation being much faster than for nitrate [21]. An example of the differences between ammonium and nitrate transport potential, and concurrent needs of N assimilation at different levels of N-stress is given in S2 Fig.
The linkage between nutrient transport and assimilation, and ultimately growth, is modulated via the expression of transport capacity for specific nutrient types via end-product (de)repression signals. These events involve both short-term control, for example satiation feedback regulation upon the operation of existing transporter proteins, and longer-term control via synthesis and removal of transporter proteins. This feedback occurs more quickly following ammonium and than nitrate transport because of the rapidity of both ammonium transport and of its assimilation [6,22]. Nitrate may also be accumulated in larger cells further decoupling processes of N-assimilation from transport. Thus, depending on the organism size and nutrient status, the nutrient being tested, experiment sampling, and subsequent data processing methodology, the values of both Gmax and KG may differ significantly from Tmax and KT [6]. Experiments using a given species and nutrient, for example varying the period of N-limitation, may likely give useful information on trends. However, interpretations of inter-species and inter-nutrient differences in Umax and KU, especially when derived by different researchers, carry a high degree of uncertainty.
Estimates of KT for nutrient transport are very rare, and values for phytoplankton nutrient transporters are rarer still [23], but a value in the range of 0.5–2 μM has been reported [15]. In the following we will assume KT = 1μM. For comparison, the KM for enzymes processing biochemical transformations are typically in the mM range [24]. Just as the importance of the numeric value of Vmax needs to be placed in the context of the enzyme sample in which it has been measured, so the value of Tmax needs to be placed in the context of the cell in which it is located. The value of Tmax may be expressed per cell, as a specific transport rate either following N-source uptake using 15N, or as a C-specific rate (this is shown in S2 Fig).
Nutrient availability for the cell does not just reflect the bulk water nutrient concentration (S∞), which is readily measured, but it reflects the interactions between processes adding and removing nutrient molecules around the individual cell which affects the substrate concentration (S0) at the transporter protein. Thus S0 is also affected by turbulence, cell size and the cell’s motion [25–27]; collectively these determine the formation of a boundary layer around the cell and thence affect diffusion to the sites of transport. Cell size is a critical determinant in transport kinetics, as it affects the boundary layer thickness and hence the relationship between S0 and S∞. It is thus constructive to express Tmax in the context of the surface area of the plasma-membrane in which the transporter proteins reside. If we assume a spherical cell form, with a given equivalent spherical diameter (ESD) and an equal distribution of transporter proteins over the membrane surface, we can then report Tmax in terms of a transport rate density (TRD; Table 1). Thus, for the transport of ammonium-N, units of TRD would be g ammonium-N d-1 μm-2; that is to say that every day across every μm2 of cell plasma-membrane area so many g ammonium-N could be transported assuming no satiation feedback. The value of TRD is enabled by the activity of many transporter proteins spread over the cell surface area (SA), each of which has its own KT and kcat. TRD is thus
TRD=TmaxSA
(2)
and Tmax is
Tmax=kcat∙{transporterproteinspercell}
(3)
The larger the cell, the greater the surface area but there is no reason to necessarily expect the value of TRD to differ according to cell size. In the following we ignore changes in cell size associated with nutrient availability (e.g. N-limited cells are typically smaller, while P-limited cells are larger) and environmental conditions (e.g. growth at different temperatures and irradiance [28] affect cell shape and size).
Growth itself is not a simple function of the presence of external nutrient availability (even if estimated more accurately as S0 rather than S∞), but is primarily a function of availability of that nutrient within the cell, and the allied biochemical processes associated with its assimilation into biomass. We thus need to consider transport rates in the context of supply and demand for the cell. Depending on the nutrient status, the value of Tmax changes (S2 Fig), and consequentially so does the value of TRD. We can now define two important specific values of TRD. These are the values of TRD needed to enable G = Gmax (TRDGmax), and the maximum possible value of TRD (TRDmax); the latter defines the value of TRD which aligns with the absolute maximum value of Tmax (Tabsmax) which is usually expressed by a cell at an intermediate level of nutrient stress (S2 Fig). By analogy with the plots shown in S2 Fig, we can also consider the excess in transport potential that develops during nutrient stress as the ratio of TRDmax: TRDGmax (δTRD; Table 1).
At saturating concentrations of nutrient and plausible maximum growth rates we can assume that diffusion is not limiting the supply of substrate to the transporter proteins (S0 ≈S∞), and hence we can estimate the value of Tmax (as per S2 Fig) and hence TRD. From experimental work for ammonium and nitrate transport into the coccolithophorid Emiliania huxleyi, raphidophyte Heterosigma carterae and the diatom Thalassiosira weissflogii we compiled the data shown in Table 2. These values exploit relationships between cell biovolume measured using an Elzone (Coulter counter–like) instrument, and C-biomass derived from elemental analysis. In Table 3, comparative values for TRD are presented, calculated using the allometric relationships of cell size to C-content taken from the literature [9]. While there are significant differences between the C-, and thus the N-content of the cells computed according to these different methods, from these estimates we obtain a feel for a likely maximum value of TRDmax. For a given computational choice (Table 2 or Table 3) the value of TRDmax is not so different between organisms of markedly different taxonomy, size and maximum growth rate potential. These values suggest a decreasing scope for excess in transport potential δTRD (i.e., TRDmax: TRDGmax) with increasing size, which may be expected, given the associated changes in surface area to volume (SA:Vol) ratio.
An analysis of the data compiled by [11], which reports experimentally derived nutrient uptake maxima, and assuming Umax = Tmax, yields average TDRmax that are broadly in line with those in Tables 2 and 3. Those data yield TRD values (as pg nutrient μm-2 d-1) of 0.075 (+/- SE 0.041), 0.115 (+/- SE 0.0137) and 0.172 (+/- SE 0.069) for ammonium-N, nitrate-N and phosphate-P uptakes, respectively, with no statistical relationship with ESD. It is noteworthy that the TRD values for ammonium estimated from the data compiled by [11] are half those for nitrate; ammonium Tmax and thus TRD is expected to be much greater than the values for nitrate transport [21,34], which could indicate confounding estimation of kinetic parameters by different researchers, as explained earlier.
Ultimately the balance of supply and demand is reflected in how close an organism comes to attaining its maximum growth rate, Gmax. It is this maximum rate of growth, and the form of the functional curve relating nutrient concentration to the achieved growth rate (G) that help define competitive advantage, and certainly do so in simple mathematical models. However, while the performance of each transporter protein may be expected to conform to the RHt2 equation of Michaelis-Menten kinetics, diffusion limitation is expected to decrease potential transport at lower nutrient concentrations [35,4,36], and the satiation feedback is expected to suppress transport rates at higher concentration (S2 Fig). In short, there are various reasons to expect that a RHt2 response curve (as used in simple models) will not well describe the true functional response curve between the bulk nutrient concentration (S∞) and G. Indeed, we should likely not expect such a RHt2 relationship even between S0 and G (S1 Fig).
Let us now consider the situation that aligns with a growth rate at half the maximum value (G0.5). At this rate, the residual steady-state nutrient concentration in the bulk medium (S∞) would equate to the half saturation value for growth, which defines KG. The value of Tmax in cells growing at the N-status equal to G0.5 is much higher than the value of Tmax in cells growing at Gmax (S2 Fig; e.g. [21]). In addition, the amount of N required to support growth at G0.5 is less than that required to support G = Gmax. If, for example, we consider Gmax to be associated with a maximum cellular N:C (g:g) of 0.2, and G = 0 with a minimum N:C of 0.05, then a cellular N:C aligning with G0.5 would be expected to be ca. 0.125 gN gC-1 (S2 Fig). In such a situation, the potential excess (δTRD) in transport capacity, of Tmax, at G0.5 could be ca. 20 fold the nutrient transport rate required at Gmax. It is thus readily apparent that cells with different stoichiometries will exhibit different growth kinetics with respect to nutrient concentration, all else being the same.
There is one other important part of the jigsaw, and that concerns the relationships between cell size, the cellular carbon density as affected by vacuolation, and cell shape. For simplicity we assume a spherical cell, which then sets surface area (SA) as a simple geometric function of cell size (ESD). Vacuolation in protists, and especially in diatoms, increases markedly with ESD [9,37], and hence the demands for nutrient transport across each μm2 of cell surface does not simply relate to cell size.
Having described the physiological framework, and considered the experimental data, we now proceed to extend the analysis according to allometric constraints across a range of sizes, organism types and motilities. The questions that we consider are:
The emphasis here is on factors that impact upon KG, namely S∞ that support G0.5. This value of KG can be seen to be an emergent property of TRD, KT, cell size, Gmax, cell motility, cell vacuolation and cellular elemental stoichiometry. To our knowledge, no previous study has considered the interconnected nature of all these facets. Collectively these also embrace the core features considered in classic trait trade-off studies.
Fig 1 shows the potential growth rate at given external bulk nutrient concentrations (S∞) in terms of dissolved inorganic-N, DIN, for different cell types and configurations, all with the same fixed maximum growth rate of Gmax = 0.693 d-1. These plots clearly show the competitive advantage for nutrient transport of being small, and of motion achieved through either swimming (flagellated phototrophic protist) or sedimentation (diatom). Thus the value of S∞ supporting G0.5 (G = 0.693/2 = 0.346 d-1), which is the value of KG, decreases with cell size and with motion. At cell ESD below 5μm, at this growth rate, nutrient concentrations at the cell surface are similar to those in the bulk water. The cellular carbon density also has an important impact on the growth-nutrient kinetics; increasing vacuolation with size (for a given C:N stoichiometric configuration) decreases the requirement for N transport. It is thus apparent that diatoms can compensate significantly for increasing cell size through being more vacuolate and hence having de facto a lower than expected SA: cell-N ratio compared to a typical protist phytoplankton. While altering the value of KT (assumed by default as 1μM) changes KG, the relationship is not pro rata; thus halving KT decreases KG to ca. 75%, and doubling KT increases KG to ca. 150%.
In Fig 2, values of KG obtained with different cell configurations growing with different maximum growth rates are plotted, showing that smaller cells can attain a higher Gmax relative to KG; their value of Gmax/KG is higher. Fig 3 also shows the potential for cell motion and/or cellular carbon density to compensate for the negative impact of increasing ESD. For a given cell configuration, however, the value of Gmax/KG is invariant with changing Gmax (Fig 3). The negative relationship between Gmax/KG and ESD varies strongly between cell configurations, and becomes more variant between configurations at larger ESD (Fig 4). The power slopes between Gmax/KG and ESD are given in Table 4; assuming a cellular carbon density that is fixed (C150), or accords with a generic protist phytoplankton (Cprot) or with a diatom (Cdiat). More details regarding the organism’s configuration are given in Table 1. The slope exceeds -1.5, but motility (through swimming or sedimentation) and increasing vacuolation with ESD mitigate the slope to closer or less than -1.
To consider the implications of variable elemental stoichiometry, Fig 5 presents the relationship between cell size and the minimum N:C quota (NCmin) and the nutrient concentration that half saturates transport of dissolved inorganic-N for protists (non-diatoms) that are motile or non-motile. This assumes a fixed maximum growth rate and fixed maximum N:C quota. These plots demonstrate a linear increase in KG as the difference between NCmax and NCmin decreases; cells with a more restricted N:C quota need more N and thence are disadvantaged if DIN acquisition is the sole limiting factor.
S3 and S4 Figs show how N-specific transport (which aligns with growth rate) varies with nutrient concentration for cell configurations Cprot and Cdiat, considering different maximum growth rate potentials, ESD, and different relationships between N-status and Tmax. These plots show how the difference between bulk water and cell surface nutrient concentrations (S∞ vs S0) for a given transport rate increases with ESD and with maximum growth rate. Also apparent is that, for a given KT (all these plots assuming the same value of 1 μM) the relationship between N-status and Tmax has a very significant effect on the kinetics (as expected from S1 Fig). To consider whether these kinetics could be adequately described through application of a simple RHt2 response curve (as per Eq 1), such a curve form was fitted to the model output using an iterative approach (as supported by SigmaPlot 12.5); the fit assumed either a free maximum rate, or a maximum rate that is fixed equal to the value of Gmax. Especially notable, where Tmax increases with deteriorating N-status (Fig 6), is that the form of the response curve appears steeper and/or plateaus more abruptly than for a RHt2 curve (S3–S6 Figs). Nonetheless, the R2 values for all of these fits exceed 0.98. The RHt2 plots typically overestimate transport at nutrient concentrations aligning with the value of KG and could significantly over-estimate (free-fitting maximum; “RHt2” plots in S3–S6 Figs) or under-estimate (plateau fixed equal to Gmax; “RHt2 fGmax” plots in S3–S6 Figs) transport at higher nutrient abundance.
Rather than using simple hypothetical relationships between N-status and Tmax (S3 Fig and S6 Fig), in Fig 7 experimentally derived response curves (from S2 Fig) were deployed. Again, the importance of the form of the relationship between N-status and Tmax is clear; especially for the nitrate curves, the deterioration in transport capacity at low N-status (low N:C in S2 Fig) results in the cell-surface nutrient values being closer to the bulk water values than may otherwise have been expected. Fig 7 also shows how RHt2 curves that give statistically acceptable fits also give differences in projected transport rates for a given nutrient concentration that could be significant in simulations. This is especially so for nitrate-supported growth.
The relationship between resource abundance and growth rate (hereafter, the “RG-relationship”) is widely considered as a key factor affecting competitive advantage, as represented as a core theme in ecological research [38]. Not only does the relationship affect bottom-up regulations in a direct fashion but it affects organism health and nutritional status, and thus affects ecological stoichiometry [39,40]. The analysis presented here indicates very significant scope for variation in the RG-relationship for phytoplankton, linked to cell elementary stoichiometry, cell size, maximum growth rate potential, motility or sedimentation, cellular carbon density (vacuolation) and the enhancement of transport potential with nutrient stress. The situation is complicated further given that we now recognise that many phytoplankton are mixotrophic, not only using inorganic nutrients but also being capable of using organic compounds and contributing to their resource needs through predation [41]. Nonetheless, the RG-relationship has been, and will continue to be, deserving of attention as it impacts on so many facets of competition within plankton communities [3] and in general ecology.
We may consider that transporter proteins are specialist enzymes. There is an established literature exploring the competitive advantages, and evolution, of enzymes of different kcat and KM. Pettersson (1989) [42] considers the evolution of the value of kcat/KM noting that, beyond the initial phase that sees the expected increase in kcat and decrease in KM, enzyme evolution displays a linked increase in both kcat and KM; the value of kcat/KM approximates the diffusion control limit at the level of the enzyme molecule. Several studies [43–45] discuss the usefulness of this so-called “specificity constant” (kcat/KM) pointing out various problems both with the usefulness of the value itself, and with its some-time alternative title as a value of “catalytic efficiency”.
Interpretations of transporter kinetic parameters, operating at the site of individual transporter proteins, would be similarly implicated in such considerations. Just as trying to piece together whole organism biochemical evolution through reference to kcat/KM for all the constitutive enzymes in an organism is fraught with problems [42], so too are considerations of transport kinetics for different substrates into different species. However, it is noteworthy that our analysis indicates that, for a given cell configuration (size, motility, value of Ccell, stoichiometry; Fig 6 and Fig 7), the value of Gmax/KG is constant, as is kcat/KM expected to be constant in an evolutionary mature enzyme.
The phytoplankton literature has hitherto explored the relative competitive value of organisms under nutrient limitation through reference to (in our terminology–see Table 1) to Umax/KU. This value of Umax/KU has been termed “affinity” in parts of this literature [10,46]. Such usage of “affinity” conflicts with traditional parlance for enzyme affinity, which defines affinity by just the half saturation constant KM. The form and interpretation of Umax/KU is also different to that for kcat/KM for enzymes; while KU may approximate to KM, Umax is de facto a function of the product of transporter kcat and the number of transporter proteins. The number of transporter proteins varies with cell size, nutrient status and likely also with Gmax. In addition, there is the practical challenge of measuring Umax, being as it is a function of Tmax (the rate of transport at the start of the experimental incubation, at t0; [6]) and incubation conditions during the assay. In consequence, the values of Umax and KU, and thence of their ratio, are subject to various confounding issues. The value of Umax/KU could, under ideal conditions of measurement, perhaps be equated to Tmax/KT; however, there is still the question as to the impact of nutrient status upon Tmax (S2 Fig), and the complication that KT is the substrate value at the transporter protein (S0) while KU is the value of the substrate concentration in the bulk medium (S∞).
The underlying explanations and potential trade-offs in expression of the uptake affinity defined as Umax/KU has been argued to lack a mechanistic basis, hence leading to a potential misrepresentation of primary production in modelling approaches [3,47,5]. Our results indicate why a search for such a mechanistic basis has proven so difficult; there are too many confounding factors. An alternative approach considers nutrient uptake as a function of cell traits and actual nutrient availability in a turbulent environment [4,48,49]. The non-linear formulation describes so-called affinity as a function of cell size, density of uptakes sites at the cell surface (i.e. transporter proteins) and turbulence [5]. This diffusion- limited nutrient uptake results in a linear scaling of affinity with the cell diameter or radius (r). While some experimental results are consistent with this scaling [50], the general picture drawn by laboratory experiments over a wide range of sizes of taxa indicate a scaling closer to the square of cell radius [10,51] that is with the cells surface area, a trend that becomes more pronounced with decreasing cell size. Theoretical arguments have suggested that this mismatch might stem from the fact that cells are not “perfect sinks”, hence are not able to absorb all nutrients at the cells surface immediately as assumed by diffusion limited nutrient uptake [20], which is likely once satiation feedback develops. According to these considerations, while smaller cells are favoured by a larger surface to volume ratio, they also require a higher transporter density to achieve maximum affinity and would thus have higher relative investment costs [20]. However, Tmax increases during at least the initial phase of nutrient-limitation (S2 Fig), which demonstrates an increased synthesis cost for transporters in such nutrient-limited cells; this suggests that the investment cost in transporters is not significant. There are clearly challenges with all the above analyses, centring upon what exactly Umax and KU index as curve-fitting parameters for RHt2 curves fitted through imperfect (and only partially understood) experimentally-derived data.
With suitable methods, estimates of Umax will approach Tmax, and estimates of KU will approach KT [6]. The numeric disparity between these variables depends on the nutrient status of the cell, the size of the cell (and thus how close S0 is to S∞), the form in which the nutrient is available, and the capacity of the cell to accumulate unaltered that particular nutrient prior to the development of satiation feedback. In consequence, greater challenges could be expected when measuring the kinetics of ammonium transport, which is assimilated very rapidly [8] and not accumulated. The ability of the diatom Phaeodatylum to take up the un-metabolisable ammonium analogue methlyamine is many orders of magnitude higher than for any other N-source [21]. This likely reflects the fact that methylamine entering via the ammonium transporter is not subject to the usual very rapid accumulation of the ammonium-transport-repressor signalling amino acid glutamine [8]. Lesser problems can be expected when measuring nitrate transport into a large vacuolated diatom that may accumulate nitrate [52], in comparison to transport into a nanoflagellate that lacks such vacuoles. It may therefore likely be no coincidence that the (few) data for kinetics for ammonium transport collated by Edwards et al. (2015) [11] appear so competitively poor in comparison with those for nitrate when the converse might have been expected. Similarly we expect fewer challenges when measuring phosphate transport into a cell type that accumulates polyphosphate.
Nutrient “affinity” [10,46], which has been described in our terminology as Umax/KU, has typically not been related to the C:N:P stoichiometry of the cell nor to the cellular carbon density both of which will affect the numeric value of this index. Together, these additional data would provide links between nutrient-status and Tmax and to the level of vacuolation affecting resource demand to be satisfied by transport over the cell surface. Collectively, stoichiometry (Fig 5) and cellular carbon density (Fig 1) affect the cell’s demand for the nutrient, which is a critical factor affecting the relative importance of any index of nutrient affinity. There is, however, scope for Tmax to vary allometrically on account of the packing of transporter proteins within the plasma membrane (Fig 6); which is consistent with the suggested explanation of the discrepancy between theoretical scaling and observed values of Umax/KU [20]. Further, and of greater significance for large non-diatoms protists than for diatoms, there is scope for the maximum growth rate to be limited by TRD attaining TRDmax (Fig 8). That is, if TRDmax = TRDGmax there is no scope to further enhance transport during nutrient stress. This is important because the value of KG is a function of the potential transport over the required capacity in transport (S1 Fig), as the ratio Tmax: TGmax. This means that larger cells, and faster growing cells of a given configuration (cell type and motility), are expected to have a higher KG.
There are also additional features of ecophysiology that affect the medium term dynamics of nutrient transport. There is for example a difference in the handling of ammonium versus nitrate, that sees the uptake and assimilation of ammonium more constrained to just the light phase. Thus ammonium transport rates during light may have to be double those expected looking at the day-average value, while nitrate assimilation is more likely split over the whole day [30]. In these contexts, it is interesting to note the relationships between ESD and Gmax for different cell types [53], and that the typical value of Gmax in phytoplankton equates to a division per day (Gmax = 0.693d-1), aligning with RuBisCo activity [19]. It is not just nutrient acquisition at nutrient-limiting concentrations that may be limiting growth rate potential; maximum transport at non-limiting concentrations may also be a factor (Fig 8). While for nitrate transport, there may be the potential for the expression of high-rate transporters, endowing the cell with a biphasic kinetic capability [18,54,55], this may be less likely for ammonium transport. Ammonium is highly toxic at high internal concentrations and its transport appears, unsurprisingly, tightly regulated. Ammonium is also normally present at low (often at vanishingly low) concentrations in natural waters, as the product of N-regeneration in ecosystems with low inorganic N concentrations. If for a given cell, the ammonium transporter exists only as a high affinity system, which is incapable of supporting growth at the highest rates because of limitations in TRD for ammonium, then high growth rates in large protists may only be possible when augmented by nitrate transport. This would place an interesting new spin on our understanding of ammonium-nitrate interactions, with implications for modelling biogeochemical and ecological events.
The results of our analysis show how features relating to the regulation of the synthesis and kinetics of transporter proteins, as well as to stoichiometric and allometric features of the cell, all play a part in the story. Arguably, the competitive advantage of an organism would be best indexed by the value of Gmax/KG as this integrates over all aspects of the organism’s nutrient physiology. We thus emphasise factors affecting KG. In the following we assume for the most part that all else remains constant (i.e. KT, TRDmax, NCmax and NCmin are constant) and consider the impacts of each of these factors upon the system.
If the cellular carbon density is constant across cell sizes, then there is a clear and powerful impact of cell size on KG (Figs 4–7). Smaller cells are much better equipped than are larger cells in this regard; this is because the SA:Vol ratio directly translates to a SA:N-demand ratio, as well as to lower diffusion limitations in smaller cells [56]. However, in reality there is an important allometric relationship between cellular carbon density and cell size [9] such that larger cells have a lower cellular carbon density. For diatoms in particular, which are increasingly vacuolate at large size [37], this greatly decreases the needs for nutrient transport across a given area of plasma-membrane. According to the calculations presented here, large diatoms with high sedimentation rates appear potentially to be much better adapted to make use of low nutrient concentrations than one may expect if one was to assume a fixed cellular carbon density (i.e., Cdiat vs C150) (Fig 4).
The consequences of this decrease in cellular carbon density with cell size is actually secondary to the decrease in N-cell density; the above mentioned mitigation of cell size on KG in consequence of the lower N-cell density thus assumes that cell stoichiometry is the same. From the effects of altering the range of cell stoichiometry, shown in Fig 5, we conclude that cell stoichiometry and the form of the relationship between stoichiometry and growth rate (the quota relationship–see [57]) are also important factors to consider when reviewing calculations of KG. That is to say, if larger cells had a high NCmin, such that the value of N:C at G0.5 was elevated, then the mitigation afforded through being more vacuolated would be eroded. Conversely, if smaller cells were relatively N-rich, then the advantage of being small would be eroded. For example cyanobacteria are typically relatively N-rich [58] and would therefore not be so competitive as may at first appear.
The physiology of nutrient acquisition and stoichiometry has the potential to override, or at least partially compensate for, limitations at transport [59]. Models considering detailed explorations of nutrient uptake kinetics thus need also to relate those kinetics to variable stoichiometry and cell size, and not assume simple fixed relationships. For phosphate transport, as for ammonium transport, TRDmax is likely very much higher than TRDGmax. In addition, the strongly curved form of the P:C quota relationship [57] will also have a strong impact upon KG for P-limited growth as the P:C value in cells growing at G0.5 will be low.
Our analysis suggests that for smaller cells (ca. <5μm ESD) motion has little additional scope to moderate diffusion limitation. Above that size, the negative effect of size is greatly countered (though not negated) by motion through swimming or sedimentation (Figs 1–5). Note that sedimentation is affected directly by Stokes law; hence differences in cell mass between species, and with nutrient status may affect sedimentation rates [60]. While it may be tempting to explain motility primarily as a mechanism to enhance competitive advantage for nutrient transport (i.e., through lowering KG), the role of motility is also related to behaviour linked to vertical migration [61,62]. Motility is also important for finding prey to support mixotrophy, an activity present in even the very smallest flagellated species, with an ESD of <3 μm, Micromonas [63].
Sedimentation in diatoms is a common trait [64,65], often considered as detrimental but having clear advantages for nutrient acquisition at low concentrations in turbulent water systems (Fig 4). For diatoms, sedimentation adds significantly to the advantage of becoming increasingly vacuolate with larger ESD (Figs 4–7). Given that cell size usually also confers an anti-predator advantage, this means that larger diatoms appear better adapted to dominate in turbulent waters (in which their sedimentation de facto confers motility) than may otherwise appear.
Our analysis indicates that the relationship between Gmax/KG and Gmax is flat for a given ESD (Fig 5). This relationship is useful as it permits the estimation of KG for a given organism type, motility and size. It also means that a given organism will have a lower KG if its Gmax should decrease through adaptation, or indeed through acclimation, to different environmental conditions. The analysis also indicates that there is scope for a much greater spread in nutrient-related kinetics in larger cells (Fig 4). For smaller cells there is less effect of motility, and less variation in cell-C density; inter-species variation will thus generate increasing “noise” in the relationship between ESD and kinetics in larger cells.
The value of Gmax/KG reflects many interactions and as a summary parameter provides an index for competitive advantage in simple nutrient-competition (bottom-up controlled) systems. The value of KG itself also has important implications for the health of the cell; it defines the bulk water nutrient concentration (S∞) supporting a state of health aligning with G0.5. Health affects the intrinsic mortality rate of the cell, a factor that is typically not included in models scaled to nutrient status, but one that is important as a selective feature [66,67]. A poor health status adversely affects the operation of repair mechanisms, e.g. compensating for photo-damage [68], and explains the duration of the lag phase of growth seen when nutrient-starved microalgae are re-fed [69].
Simple models relate nutrient concentration to transport rate and thence to growth rate using a rectangular hyperbolic type 2 (RHt2) response curve, in line with Monod (1949) [70]. From our analysis (Figs 9, S1 and S3–S6) RHt2 cannot be expected to well define the actual relationship between nutrient concentration in the bulk medium (S∞) and transport. The fitting of RHT2 tends to over-estimate transport at lower nutrient availability and over or under estimate it at high availability. The expected relationship plateaus more abruptly than RHt2 can describe it. It is noteworthy that the fit of RHt2 to the modelled relationships was high (R2 > 0.95 in all instances, and most > 0.98); the “noise” in biological measurements that is inherent in experimental procedures of transport and growth rates [6] will inevitably result in a statistically acceptable fit to RHt2. Nonetheless, RHt2 does not appear to be appropriate, and the apparent subtle differences in the form of the described nutrient transport kinetics will manifest themselves in potentially important differences in competitive advantage in modelled populations. Such differences become more apparent when considering the form of the relationship between nutrient status and Tmax (Fig 7), a topic that is also of consequence when describing the ammonium-nitrate interaction [71]. It is also important to couple nutrient-light limitations in the correct way, else the expected decrease in KG with light limitation does not occur [72]. Interactions with temperature and allometry are also complex [53,73], with changes in cell size, overall growth rate, and differential impacts on transport vs metabolism [28,74]. All of this speaks to the importance of describing the relationship between multi-factor feedback interactions upon cell growth, with some attempt to simulate (de)repression of different metabolic pathways.
In general, the importance and usefulness of using a single proxy as a determinant of competitive advantage seems overstated. This applies to usage of the value of kcat/KM in enzyme kinetics, Umax/KU in studies of diffusion limitation, or Gmax/KG in whole organism growth kinetics. Similarly, only considering stoichiometry represents too great a simplification in considerations of nutrient competition [59]. We simply have too limited knowledge of the real nutrient concentrations at the scales of consequence for these organisms (proximate to the cells), while we also know that factors such as alternative nutritional routes (nitrate vs ammonium vs dissolved organic -N; phosphate vs dissolved organic -P, phago-mixotrophy), different transporter types with different affinities for a given nutrient [14,16], allelopathy [75], palatability for grazers [76] and resistance to non-predator factors affecting cell mortality [77] are all important if not critical factors affecting competition at different times and places in the real world.
Our analysis, like many other studies, makes the unrealistic caveat of all-else-being-equal across a wide range of organism types, shapes, sizes, motilities and stoichiometries. So, while Fig 4 portrays a general theoretical pattern, application of that pattern to explain species competition for growth in the same water body must be viewed with extreme caution. It is of some comfort that the approach justifies (is consistent with) a common assumption that fast growing (r-select) species are disadvantaged in mature ecosystems where their slower growing (K-select) competitors have a better nutrient affinity (lower KG). However, simply relating KG (or indeed any such parameter) to size is in any case highly problematic: many very small, non-motile cells tend to grow together (notably when P-stressed), and diatoms can grow in chains or mats, so that effective particle size (affecting boundary layer thickness and sedimentation) is often larger than it appears; the impacts of such changes are typically not included in models. Furthermore, little is known about interactions with alternative modes of nutrition, such as mixotrophy (including the use of dissolved organics), which likely vary significantly between organisms and will impact greatly upon the significance of KG for a given limiting nutrient at any instant in time.
Within simple bottom-up controlled systems operating under non-steady-state conditions, possession of a higher growth rate is expected to endow a powerful competitive advantage under conditions of nutrient abundance. Larger growing cells need not be disadvantaged in such systems. However, smaller organisms appear always to be at an advantage for nutrient acquisition within nutrient limited systems running closer to steady-state, as epitomised by chemostat experimental systems and typically observed in the oligotrophic oceans. In a chemostat, at a given dilution rate the substrate concentration converges on that which enables the growth rate to match the dilution rate. Besides the logistic challenges in running chemostats to determine KG, there is also the real risk that the organisms adapt to enforced slow growth over many months [66]. It is notable that the predicted values of KG from this study (Fig 2) are in the main very low, bordering on the level of chemical detection in the bulk media, even when assuming a transporter protein nutrient affinity (KT) of 1 μM. Interestingly, in modelled systems, the dynamics of the system may be more heavily controlled by the parameters controlling activity of zooplankton than by the value of KG for phytoplankton [78]. It is also noteworthy that factors affecting cell size, motility/sedimentation, stoichiometry and cellular carbon density impact greatly upon predation kinetics and the value of the organism as food for the predator [79,26]. Thus, while motility enhances transport potential through decreasing boundary-layer limitations, motility is rather a double-edged-sword as it raises the likelihood of encountering a predator. For sure, simple comparisons between single-factors such as nutrient competition cannot possibly determine the true competitive advantage.
We can perhaps be more secure in considering the implications of our analysis for the evolution of an individual species, where intra-species competition is important. Here, within a particular cell line of a given species, the values of KG and Gmax can be expected to be linked; a faster growing cell will have a higher KG. This observation is consistent with a general feature of enzyme activity such that high kcat is often associated with a high KM [42], in consequence of a low KM being deleterious for the rapid breakdown of the enzyme-substrate complex. Irrespective of species-species interactions, one may thus expect a trade-off between KG and Gmax and for this to be reflected in the evolution of a particular cell line. Taken alone, this is an important trade-off between traits affecting the benefit of fast growth and is consistent with the observation that cells forced to grow slowly in a low-dilution chemostat (noting that dilution rate = growth rate at steady-state, and that the residual nutrient concentration is lower at low dilution rates) evolve a lower Gmax than the parent population [66]. The complexity of trade-offs in the evolution of individual enzymes [42] perhaps warns against attempting too-tight a linkage between KG and Gmax in terms of trait trade-off arguments at the whole organism level.
In the following we assume that the transporter rate density (TRD) has a maximum possible value (TRDmax); that is to say that, the plasma-membrane can only contain so-many nutrient transporter proteins over a given area. We assume TRDmax to be the experimentally determined maximum rate of 0.4 pgNμm-2 d-1 (from the diatom Thalassiosira, using experimentally computed C-cell; Table 2). Note, that the actual expressed value of TRD, and the instantaneous operation of transporter proteins, may be down-regulated due to long or short-term feedback linked to satiation and cellular nutrient status. It is assumed that all transporter proteins, contributing to TRD, have the same transport potential irrespective of the organism; hence we assume no features of the plasma-membrane or allied cell wall structure affect the functional value of kcat or KT of the embedded transporter proteins.
The value of Tmax varies with the physiological status of the cells. Here we consider the N-status as indexed by the cellular N:C. The N-status is described as a normalised N:C quota [57] such that minimum stress is given by NCu = 1, and maximum stress by NCu = 0. The equation defining NCu is:
NCu=(1+KQ)∙(NC−NCmin)(NC−NCmin)+KQ∙(NCmax−NCmin)
(4)
NC is the current cellular mass ratio of N:C, which ranges between NCmin when G is limited to 0 by supply of nutrient-N, and NCmax when G = Gmax. KQ is a curve shaping constant, which at a KQ = 10 gives the expected near-linear relationship between N:C and the growth rate, G [6].
The value of Tmax can be derived experimentally (as in S2 Fig). Tmax can alternatively be described hypothetically as increasing with decreasing nutrient status. To achieve the latter, here we use a simple curve form that carries a minimum of Gmax × NCmax and rises rapidly as the N-status, NCu, decreases (i.e. as N:C decreases from NCmax to NCmin). This equation contains a normalised RHt2 description which for values of 0≤NCu≤1 will return a value of 0 to 1 irrespective of the value of KTcon, which is a curve setting constant (the lower the value the steeper the curve, increasing Tmax as N:C decreases with N-stress).
Tmax=Gmax∙NCmax∙(1+Tadd∙(1+KTcon)∙(1−NCu)(1−NCu+KTcon))
(5)
The value of Tadd provides a simple approach to reflect the diversity in scaling between the very highest expressed Tmax and that required to support G = Gmax, as broadly seen in real organisms (S2 Fig). Tadd acts as a multiplier for Tmax (dimensionless); e.g. Tadd = 1, will at NCu = 0 double the value of Tmax over that expressed when NCu = 1 with G = Gmax. If Tadd = 0, then Eq 5 describes a flat Tmax, as is de facto assumed in most models [72,80].
The maximum possible value of Tadd in Eq 5 is a function of the value of TRDmax permitting us to explore the allometric and allied scaling of transport potential by reference to the maximum possible TRD (which here we consider as 0.4 pg nutrient-N μm-2 d-1) and also to the value of TRD required to support Gmax, TRDGmax. From Eq 2, we obtain:
TRDGmax=Gmax∙NCmax∙CcellSA
(6)
Ccell is the C content per cell (pgC); this value as a function of ESD is described as per [9]. SA is the cell surface area (μm2), and NCmax is the mass N:C at G = Gmax.
Tadd is then given by
Tadd=TRDmax−TRDGmaxTRDGmax
(7)
From Fig 8, it can be seen how the value of TRDGmax varies between organism configurations, increasing with size and Gmax. In particular large protists with their high demands for nutrients become limited by the value of TRDmax at high growth rates, i.e. TRDGmax approaches the maximum density of 0.4 pg nutrient-N μm-2 d-1. Fig 6, for a hypothetical organism with a fixed cellular carbon density (C150), shows the potential for smaller organisms to have scope for a far higher excess transport capacity; that is TRDmax: TRDGmax = δTRD is higher for small cells, and this excess is higher again at lower Gmax. However, in realty larger cells are less C-dense [9], and this is even more apparent for diatoms as these are relatively even more vacuolated; this mitigates against the simple allometric response (Fig 9; Cf. Fig 2).
From the value of Tmax, the transport rate (T) is given by Eq 8 (Cf. Eq 1), where S0 is the nutrient concentration at the plasma membrane surface, and KT is the half saturation constant for the nutrient transporter protein,
T=Tmax∙S0S0+KT
(8)
This is rearranged to obtain S0:
S0=TKTTmax−T
(9)
In reality, the value of T is limited by diffusion at low nutrient concentrations. This limitation sets a relationship between S∞ and S0. From Eqs 16 and 17 in [25], developed from [35], the transport rate of nutrient into the cell (T, ng cell-1 d-1) is related to the gradient between the bulk nutrient concentration and the nutrient concentration at the cell surface (S∞−S0, ng L-1) via the following equation:
T=Dr(1+0.5∙rD∙c)∙4πr2(S∞−S0)
(10)
Here, r is the cell radius (μm), D is diffusivity (μm2 d-1), c is the organism’s speed of motion either due to swimming or sedimentation (μm d-1). The thickness of the boundary layer impacts upon the difference between S∞ and S0; the larger the cell, and the slower its motion through the water, the greater is the value of (S∞−S0) for a given value of T. By rearranging Eq 10, we obtain the value of S∞.
Cell motility (c, μm s-1) was configured using an empirical allometric equation using data from Sommer 1988 [81] and Visser & Kiørboe 2006 [82] according to [79] as:
c=38.542*(ESD)0.5424
(12)
Sedimentation rates (csed; μm s-1) were computed using Stoke’s law, from the cell radius (r; μm), cell density (ρorg; assumed here to be 1.0634 kg L-1), seawater density (ρw; assumed here to be 1.033 kg L-1), dynamic viscosity (η; assumed here to be 1.0846x 10−3 Pa s), and acceleration due to gravity (g; 9.8 m s-2).
In order to compute the value of the bulk-water nutrient concentration (S∞) that supports a given growth rate, the above equations were constructed to enable organism size, allometric parameters and motility to be altered. For given values of Gmax, NCmax and NCmin, the rate of N transport required to support a given G is computed. For a given cell size, cellular carbon density and N:C, we calculate the cell surface area, and the N-cell density at a given G. From these the rate of N-source transport per cell surface area is computed to support the given G; this is the value of T in Eqs 8 and 10.
|
Subsets and Splits